diff --git a/workspace/pytorch/SenondPointpillars.bash_scripts.ipynb b/workspace/pytorch/SenondPointpillars.bash_scripts.ipynb index 0cf3ff0..2822389 100644 --- a/workspace/pytorch/SenondPointpillars.bash_scripts.ipynb +++ b/workspace/pytorch/SenondPointpillars.bash_scripts.ipynb @@ -2,9 +2,17 @@ "cells": [ { "cell_type": "code", - "execution_count": null, + "execution_count": 1, "metadata": {}, - "outputs": [], + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "env: KITTI_ROOT=/data/kitti_lidar\n" + ] + } + ], "source": [ "%env KITTI_ROOT=/data/kitti_lidar\n", "!export KITTI_ROOT=/data/kitti_lidar" @@ -12,9 +20,17 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 2, "metadata": {}, - "outputs": [], + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "env: NUMBA_WARNINGS=0\n" + ] + } + ], "source": [ "%env NUMBA_WARNINGS=0" ] @@ -28,7 +44,7 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 3, "metadata": {}, "outputs": [], "source": [ @@ -40,7 +56,7 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 4, "metadata": {}, "outputs": [], "source": [ @@ -50,27 +66,65 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 5, "metadata": {}, - "outputs": [], + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "__init__.py create_data.py protos\t simple-inference.ipynb\n", + "builder data\t pytorch\t utils\n", + "configs framework\t script.py\n", + "core\t kittiviewer script_server.py\n" + ] + } + ], "source": [ "!ls ${SECOND_API}" ] }, { "cell_type": "code", - "execution_count": null, + "execution_count": 6, "metadata": {}, - "outputs": [], + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "testing training\n" + ] + } + ], "source": [ "!ls ${KITTI_ROOT}" ] }, { "cell_type": "code", - "execution_count": null, + "execution_count": 7, "metadata": {}, - "outputs": [], + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "/data/kitti_lidar\n", + "|-- testing\n", + "| |-- calib [7518 entries exceeds filelimit, not opening dir]\n", + "| |-- image_2 [7518 entries exceeds filelimit, not opening dir]\n", + "| `-- velodyne [7518 entries exceeds filelimit, not opening dir]\n", + "`-- training\n", + " |-- calib [7481 entries exceeds filelimit, not opening dir]\n", + " |-- image_2 [7481 entries exceeds filelimit, not opening dir]\n", + " |-- label_2 [7481 entries exceeds filelimit, not opening dir]\n", + " `-- velodyne [7481 entries exceeds filelimit, not opening dir]\n", + "\n", + "9 directories, 0 files\n" + ] + } + ], "source": [ "!tree --filelimit 10 ${KITTI_ROOT}" ] @@ -84,7 +138,7 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 8, "metadata": {}, "outputs": [], "source": [ @@ -94,18 +148,55 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 9, "metadata": {}, - "outputs": [], + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "Generate info. this may take several minutes.\n", + "Kitti info train file is saved to /data/kitti_lidar/kitti_infos_train.pkl\n", + "Kitti info val file is saved to /data/kitti_lidar/kitti_infos_val.pkl\n", + "Kitti info trainval file is saved to /data/kitti_lidar/kitti_infos_trainval.pkl\n", + "Kitti info test file is saved to /data/kitti_lidar/kitti_infos_test.pkl\n", + "[100.0%][===================>][184.77it/s][00:19>00:00] \n", + "[100.0%][===================>][195.29it/s][00:21>00:00] \n", + "[100.0%][===================>][189.51it/s][00:40>00:00] \n", + "remain number of infos: 3712\n", + "[100.0%][===================>][550.14it/s][00:06>00:00] \n", + "load 2207 Pedestrian database infos\n", + "load 14357 Car database infos\n", + "load 734 Cyclist database infos\n", + "load 1297 Van database infos\n", + "load 488 Truck database infos\n", + "load 224 Tram database infos\n", + "load 337 Misc database infos\n", + "load 56 Person_sitting database infos\n" + ] + } + ], "source": [ - "!python ${SECOND_API}/create_data.py kitti_data_prep ${KITTI_ROOT}" + "# !python ${SECOND_API}/create_data.py kitti_data_prep ${KITTI_ROOT}" ] }, { "cell_type": "code", - "execution_count": null, + "execution_count": 10, "metadata": {}, - "outputs": [], + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "-rw-r--r-- 1 jit jit 6M Jun 24 00:02 /data/kitti_lidar/kitti_dbinfos_train.pkl\n", + "-rw-r--r-- 1 jit jit 11M Jun 24 00:00 /data/kitti_lidar/kitti_infos_test.pkl\n", + "-rw-r--r-- 1 jit jit 13M Jun 23 23:59 /data/kitti_lidar/kitti_infos_train.pkl\n", + "-rw-r--r-- 1 jit jit 25M Jun 24 00:00 /data/kitti_lidar/kitti_infos_trainval.pkl\n", + "-rw-r--r-- 1 jit jit 13M Jun 24 00:00 /data/kitti_lidar/kitti_infos_val.pkl\n" + ] + } + ], "source": [ "!ls -l --block-size=M /data/kitti_lidar/*.pkl" ] @@ -119,18 +210,52 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 11, "metadata": {}, - "outputs": [], + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "/opt/second.pytorch/second/configs:\n", + "all.fhd.config\tcar.fhd.onestage.config nuscenes\t pointpillars\n", + "car.fhd.config\tcar.lite.config\t\t people.fhd.config\n", + "\n", + "/opt/second.pytorch/second/configs/nuscenes:\n", + "all.fhd.config\t\t all.pp.largea.config\tall.pp.mhead.config\n", + "all.pp.deprecated.config all.pp.lowa.config\tall.pp.mida.config\n", + "\n", + "/opt/second.pytorch/second/configs/pointpillars:\n", + "car ped_cycle\tpp_pretrain.config\n", + "\n", + "/opt/second.pytorch/second/configs/pointpillars/car:\n", + "xyres_16.config xyres_20.config xyres_24.config xyres_28.config\n", + "\n", + "/opt/second.pytorch/second/configs/pointpillars/ped_cycle:\n", + "xyres_16.config xyres_20.config xyres_24.config xyres_28.config\n" + ] + } + ], "source": [ "!ls -R ${SECOND_API}/configs" ] }, { "cell_type": "code", - "execution_count": null, + "execution_count": 12, "metadata": {}, - "outputs": [], + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "env: KITTI_ROOT=/data/kitti_lidar\n", + "env: MODEL_OUT_PATH=/data/kitti_lidar/trained_model\n", + "env: MODEL_NAME=car.lite\n", + "env: MODEL_NAME_FLAT=car_lite\n" + ] + } + ], "source": [ "%env KITTI_ROOT=/data/kitti_lidar\n", "%env MODEL_OUT_PATH=/data/kitti_lidar/trained_model\n", @@ -140,7 +265,7 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 13, "metadata": {}, "outputs": [], "source": [ @@ -155,9 +280,17 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 14, "metadata": {}, - "outputs": [], + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "car_lite\n" + ] + } + ], "source": [ "!echo $MODEL_NAME_FLAT" ] @@ -171,11 +304,225 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 15, "metadata": { "scrolled": true }, - "outputs": [], + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "model: {\n", + " second: {\n", + " network_class_name: \"VoxelNet\"\n", + " voxel_generator {\n", + " # point_cloud_range : [0, -40, -3, 70.4, 40, 1]\n", + " point_cloud_range : [0, -32.0, -3, 52.8, 32.0, 1]\n", + " # point_cloud_range : [-50, -50.0, -3, 50, 50, 1]\n", + " voxel_size : [0.05, 0.05, 0.1]\n", + " max_number_of_points_per_voxel : 1\n", + " block_filtering: false # filter voxels by block height\n", + " block_factor: 1 # block size: voxel_size * block_factor * block_size = 0.05 * 1 * 8 = 0.4\n", + " block_size: 8\n", + " height_threshold: 0.2\n", + " }\n", + "\n", + " voxel_feature_extractor: {\n", + " module_class_name: \"SimpleVoxelRadius\"\n", + " num_filters: [16]\n", + " with_distance: false\n", + " num_input_features: 4\n", + " }\n", + " middle_feature_extractor: {\n", + " module_class_name: \"SpMiddleFHDLite\"\n", + " # num_filters_down1: [] # protobuf don't support empty list.\n", + " # num_filters_down2: []\n", + " downsample_factor: 8\n", + " num_input_features: 3\n", + " }\n", + " rpn: {\n", + " module_class_name: \"RPNV2\"\n", + " layer_nums: [5]\n", + " layer_strides: [1]\n", + " num_filters: [128]\n", + " upsample_strides: [1]\n", + " num_upsample_filters: [128]\n", + " use_groupnorm: false\n", + " num_groups: 32\n", + " num_input_features: 128\n", + " }\n", + " loss: {\n", + " classification_loss: {\n", + " weighted_sigmoid_focal: {\n", + " alpha: 0.25\n", + " gamma: 2.0\n", + " anchorwise_output: true\n", + " }\n", + " }\n", + " localization_loss: {\n", + " weighted_smooth_l1: {\n", + " sigma: 3.0\n", + " code_weight: [1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0]\n", + " }\n", + " }\n", + " classification_weight: 1.0\n", + " localization_weight: 2.0\n", + " }\n", + " num_point_features: 4 # model's num point feature should be independent of dataset\n", + " # Outputs\n", + " use_sigmoid_score: true\n", + " encode_background_as_zeros: true\n", + " encode_rad_error_by_sin: true\n", + " sin_error_factor: 1.0\n", + "\n", + " use_direction_classifier: true # this can help for orientation benchmark\n", + " direction_loss_weight: 0.2 # enough.\n", + " num_direction_bins: 2\n", + " direction_limit_offset: 1\n", + "\n", + " # Loss\n", + " pos_class_weight: 1.0\n", + " neg_class_weight: 1.0\n", + "\n", + " loss_norm_type: NormByNumPositives\n", + " # Postprocess\n", + " post_center_limit_range: [0, -40, -2.2, 70.4, 40, 0.8]\n", + " nms_class_agnostic: false # only valid in multi-class nms\n", + " box_coder: {\n", + " ground_box3d_coder: {\n", + " linear_dim: false\n", + " encode_angle_vector: false\n", + " }\n", + " }\n", + " target_assigner: {\n", + " class_settings: {\n", + " anchor_generator_range: {\n", + " sizes: [1.6, 3.9, 1.56] # wlh\n", + " anchor_ranges: [0, -32.0, -1.0, 52.8, 32.0, -1.0]\n", + " rotations: [0, 1.57] # DON'T modify this unless you are very familiar with my code.\n", + " }\n", + " matched_threshold : 0.6\n", + " unmatched_threshold : 0.45\n", + " class_name: \"Car\"\n", + " use_rotate_nms: true\n", + " use_multi_class_nms: false\n", + " nms_pre_max_size: 1000\n", + " nms_post_max_size: 100\n", + " nms_score_threshold: 0.3 # 0.4 in submit, but 0.3 can get better hard performance\n", + " nms_iou_threshold: 0.01\n", + " region_similarity_calculator: {\n", + " nearest_iou_similarity: {\n", + " }\n", + " }\n", + " }\n", + " sample_positive_fraction : -1\n", + " sample_size : 512\n", + " assign_per_class: true\n", + " }\n", + " }\n", + "}\n", + "\n", + "train_input_reader: {\n", + " dataset: {\n", + " dataset_class_name: \"KittiDataset\"\n", + " kitti_info_path: \"/data/kitti_lidar/kitti_infos_train.pkl\"\n", + " kitti_root_path: \"/data/kitti_lidar\"\n", + " }\n", + " \n", + " batch_size: 12\n", + " preprocess: {\n", + " max_number_of_voxels: 15000\n", + " shuffle_points: true\n", + " num_workers: 3\n", + " groundtruth_localization_noise_std: [1.0, 1.0, 0.5]\n", + " # groundtruth_rotation_uniform_noise: [-0.3141592654, 0.3141592654]\n", + " groundtruth_rotation_uniform_noise: [-0.78539816, 0.78539816]\n", + " global_rotation_uniform_noise: [-0.78539816, 0.78539816]\n", + " global_scaling_uniform_noise: [0.95, 1.05]\n", + " global_random_rotation_range_per_object: [0, 0] # pi/4 ~ 3pi/4\n", + " global_translate_noise_std: [0, 0, 0]\n", + " anchor_area_threshold: -1\n", + " remove_points_after_sample: false\n", + " groundtruth_points_drop_percentage: 0.0\n", + " groundtruth_drop_max_keep_points: 15\n", + " remove_unknown_examples: false\n", + " sample_importance: 1.0\n", + " random_flip_x: false\n", + " random_flip_y: true\n", + " remove_environment: false\n", + "\n", + " database_sampler {\n", + " database_info_path: \"/data/kitti_lidar/kitti_dbinfos_train.pkl\"\n", + " sample_groups {\n", + " name_to_max_num {\n", + " key: \"Car\"\n", + " value: 15\n", + " }\n", + " }\n", + " database_prep_steps {\n", + " filter_by_min_num_points {\n", + " min_num_point_pairs {\n", + " key: \"Car\"\n", + " value: 5\n", + " }\n", + " }\n", + " }\n", + " database_prep_steps {\n", + " filter_by_difficulty {\n", + " removed_difficulties: [-1]\n", + " }\n", + " }\n", + " global_random_rotation_range_per_object: [0, 0]\n", + " rate: 1.0\n", + " }\n", + " }\n", + "}\n", + "\n", + "train_config: {\n", + " optimizer: {\n", + " adam_optimizer: {\n", + " learning_rate: {\n", + " one_cycle: {\n", + " lr_max: 3e-3\n", + " moms: [0.95, 0.85]\n", + " div_factor: 10.0\n", + " pct_start: 0.4\n", + " }\n", + " }\n", + " weight_decay: 0.01\n", + " }\n", + " fixed_weight_decay: true\n", + " use_moving_average: false\n", + " }\n", + " steps: 15500 # 310 * 50\n", + " steps_per_eval: 1550 # 310 * 5\n", + " save_checkpoints_secs : 1800 # half hour\n", + " save_summary_steps : 10\n", + " enable_mixed_precision: false \n", + " loss_scale_factor: -1\n", + " clear_metrics_every_epoch: true\n", + "}\n", + "\n", + "eval_input_reader: {\n", + " dataset: {\n", + " dataset_class_name: \"KittiDataset\"\n", + " kitti_info_path: \"/data/kitti_lidar/kitti_infos_val.pkl\"\n", + " # kitti_info_path: \"/data/kitti_lidar/kitti_infos_test.pkl\"\n", + " kitti_root_path: \"/data/kitti_lidar\"\n", + " }\n", + " batch_size: 12\n", + " preprocess: {\n", + " max_number_of_voxels: 30000\n", + " shuffle_points: false\n", + " num_workers: 3\n", + " anchor_area_threshold: -1\n", + " remove_environment: false\n", + " }\n", + "}" + ] + } + ], "source": [ "%%bash\n", "mkdir -p ${MODEL_OUT_PATH}\n", @@ -202,18 +549,70 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 16, "metadata": {}, - "outputs": [], + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "car_fhd car_lite car_onestage pp_model_for_nuscenes_pretrain\n" + ] + } + ], "source": [ "!ls /data/pretrained_models/Pointpillars_models_v1.5" ] }, { "cell_type": "code", - "execution_count": null, + "execution_count": 17, "metadata": {}, - "outputs": [], + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "[ 41 1280 1056]\n", + "Restoring parameters from /data/pretrained_models/Pointpillars_models_v1.5/car_lite/voxelnet-15500.tckpt\n", + "feature_map_size [1, 160, 132]\n", + "remain number of infos: 3769\n", + "Generate output labels...\n", + "[100.0%][===================>][12.27it/s][00:26>00:00] \n", + "generate label finished(140.19/s). start eval:\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typed_passes.py:314: NumbaPerformanceWarning: \n", + "The keyword argument 'parallel=True' was specified but no transformation for parallel execution was possible.\n", + "\n", + "To find out why, try turning on parallel diagnostics, see http://numba.pydata.org/numba-doc/latest/user/parallel.html#diagnostics for help.\n", + "\n", + "File \"../../opt/second.pytorch/second/utils/eval.py\", line 129:\n", + "@numba.jit(nopython=True, parallel=True)\n", + "def box3d_overlap_kernel(boxes,\n", + "^\n", + "\n", + " state.func_ir.loc))\n", + "Evaluation official\n", + "Car AP(Average Precision)@0.70, 0.70, 0.70:\n", + "bbox AP:90.64, 89.27, 87.86\n", + "bev AP:90.06, 87.28, 86.39\n", + "3d AP:88.06, 77.69, 75.41\n", + "aos AP:90.55, 88.69, 87.04\n", + "Car AP(Average Precision)@0.70, 0.50, 0.50:\n", + "bbox AP:90.64, 89.27, 87.86\n", + "bev AP:90.68, 89.79, 89.14\n", + "3d AP:90.67, 89.74, 88.99\n", + "aos AP:90.55, 88.69, 87.04\n", + "\n", + "Evaluation coco\n", + "Car coco AP@0.50:0.05:0.95:\n", + "bbox AP:72.30, 68.76, 67.12\n", + "bev AP:70.37, 67.05, 65.15\n", + "3d AP:60.75, 56.40, 54.51\n", + "aos AP:72.22, 68.36, 66.53\n", + "\n" + ] + } + ], "source": [ "!python ${SECOND_API}/pytorch/train.py evaluate --config_path=${MODEL_OUT_PATH}/train.config --model_dir=/data/pretrained_models/Pointpillars_models_v1.5/${MODEL_NAME_FLAT}" ] @@ -234,9 +633,30 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 18, "metadata": {}, - "outputs": [], + "outputs": [ + { + "data": { + "text/html": [ + "\n", + " \n", + " " + ], + "text/plain": [ + "" + ] + }, + "metadata": {}, + "output_type": "display_data" + } + ], "source": [ "%load_ext tensorboard.notebook\n", "%tensorboard --logdir /data/kitti_lidar/trained_model" @@ -251,9 +671,17 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 19, "metadata": {}, - "outputs": [], + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "/data/kitti_lidar/trained_model/car_lite\n" + ] + } + ], "source": [ "!echo ${MODEL_OUT_PATH}/${MODEL_NAME_FLAT}\n", "!rm -rf ${MODEL_OUT_PATH}/${MODEL_NAME_FLAT}" @@ -268,11 +696,1748 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 20, "metadata": { "scrolled": true }, - "outputs": [], + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "[ 41 1280 1056]\n", + "num parameters: 39\n", + "False _amp_stash\n", + "{'Car': 5}\n", + "[-1]\n", + "load 2207 Pedestrian database infos\n", + "load 14357 Car database infos\n", + "load 734 Cyclist database infos\n", + "load 1297 Van database infos\n", + "load 488 Truck database infos\n", + "load 224 Tram database infos\n", + "load 337 Misc database infos\n", + "load 56 Person_sitting database infos\n", + "After filter database:\n", + "load 2207 Pedestrian database infos\n", + "load 13442 Car database infos\n", + "load 734 Cyclist database infos\n", + "load 1297 Van database infos\n", + "load 488 Truck database infos\n", + "load 224 Tram database infos\n", + "load 337 Misc database infos\n", + "load 56 Person_sitting database infos\n", + "feature_map_size [1, 160, 132]\n", + "remain number of infos: 3712\n", + "feature_map_size [1, 160, 132]\n", + "remain number of infos: 3769\n", + "model: {\n", + " second: {\n", + " network_class_name: \"VoxelNet\"\n", + " voxel_generator {\n", + " # point_cloud_range : [0, -40, -3, 70.4, 40, 1]\n", + " point_cloud_range : [0, -32.0, -3, 52.8, 32.0, 1]\n", + " # point_cloud_range : [-50, -50.0, -3, 50, 50, 1]\n", + " voxel_size : [0.05, 0.05, 0.1]\n", + " max_number_of_points_per_voxel : 1\n", + " block_filtering: false # filter voxels by block height\n", + " block_factor: 1 # block size: voxel_size * block_factor * block_size = 0.05 * 1 * 8 = 0.4\n", + " block_size: 8\n", + " height_threshold: 0.2\n", + " }\n", + "\n", + " voxel_feature_extractor: {\n", + " module_class_name: \"SimpleVoxelRadius\"\n", + " num_filters: [16]\n", + " with_distance: false\n", + " num_input_features: 4\n", + " }\n", + " middle_feature_extractor: {\n", + " module_class_name: \"SpMiddleFHDLite\"\n", + " # num_filters_down1: [] # protobuf don't support empty list.\n", + " # num_filters_down2: []\n", + " downsample_factor: 8\n", + " num_input_features: 3\n", + " }\n", + " rpn: {\n", + " module_class_name: \"RPNV2\"\n", + " layer_nums: [5]\n", + " layer_strides: [1]\n", + " num_filters: [128]\n", + " upsample_strides: [1]\n", + " num_upsample_filters: [128]\n", + " use_groupnorm: false\n", + " num_groups: 32\n", + " num_input_features: 128\n", + " }\n", + " loss: {\n", + " classification_loss: {\n", + " weighted_sigmoid_focal: {\n", + " alpha: 0.25\n", + " gamma: 2.0\n", + " anchorwise_output: true\n", + " }\n", + " }\n", + " localization_loss: {\n", + " weighted_smooth_l1: {\n", + " sigma: 3.0\n", + " code_weight: [1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0]\n", + " }\n", + " }\n", + " classification_weight: 1.0\n", + " localization_weight: 2.0\n", + " }\n", + " num_point_features: 4 # model's num point feature should be independent of dataset\n", + " # Outputs\n", + " use_sigmoid_score: true\n", + " encode_background_as_zeros: true\n", + " encode_rad_error_by_sin: true\n", + " sin_error_factor: 1.0\n", + "\n", + " use_direction_classifier: true # this can help for orientation benchmark\n", + " direction_loss_weight: 0.2 # enough.\n", + " num_direction_bins: 2\n", + " direction_limit_offset: 1\n", + "\n", + " # Loss\n", + " pos_class_weight: 1.0\n", + " neg_class_weight: 1.0\n", + "\n", + " loss_norm_type: NormByNumPositives\n", + " # Postprocess\n", + " post_center_limit_range: [0, -40, -2.2, 70.4, 40, 0.8]\n", + " nms_class_agnostic: false # only valid in multi-class nms\n", + " box_coder: {\n", + " ground_box3d_coder: {\n", + " linear_dim: false\n", + " encode_angle_vector: false\n", + " }\n", + " }\n", + " target_assigner: {\n", + " class_settings: {\n", + " anchor_generator_range: {\n", + " sizes: [1.6, 3.9, 1.56] # wlh\n", + " anchor_ranges: [0, -32.0, -1.0, 52.8, 32.0, -1.0]\n", + " rotations: [0, 1.57] # DON'T modify this unless you are very familiar with my code.\n", + " }\n", + " matched_threshold : 0.6\n", + " unmatched_threshold : 0.45\n", + " class_name: \"Car\"\n", + " use_rotate_nms: true\n", + " use_multi_class_nms: false\n", + " nms_pre_max_size: 1000\n", + " nms_post_max_size: 100\n", + " nms_score_threshold: 0.3 # 0.4 in submit, but 0.3 can get better hard performance\n", + " nms_iou_threshold: 0.01\n", + " region_similarity_calculator: {\n", + " nearest_iou_similarity: {\n", + " }\n", + " }\n", + " }\n", + " sample_positive_fraction : -1\n", + " sample_size : 512\n", + " assign_per_class: true\n", + " }\n", + " }\n", + "}\n", + "\n", + "train_input_reader: {\n", + " dataset: {\n", + " dataset_class_name: \"KittiDataset\"\n", + " kitti_info_path: \"/data/kitti_lidar/kitti_infos_train.pkl\"\n", + " kitti_root_path: \"/data/kitti_lidar\"\n", + " }\n", + " \n", + " batch_size: 12\n", + " preprocess: {\n", + " max_number_of_voxels: 15000\n", + " shuffle_points: true\n", + " num_workers: 3\n", + " groundtruth_localization_noise_std: [1.0, 1.0, 0.5]\n", + " # groundtruth_rotation_uniform_noise: [-0.3141592654, 0.3141592654]\n", + " groundtruth_rotation_uniform_noise: [-0.78539816, 0.78539816]\n", + " global_rotation_uniform_noise: [-0.78539816, 0.78539816]\n", + " global_scaling_uniform_noise: [0.95, 1.05]\n", + " global_random_rotation_range_per_object: [0, 0] # pi/4 ~ 3pi/4\n", + " global_translate_noise_std: [0, 0, 0]\n", + " anchor_area_threshold: -1\n", + " remove_points_after_sample: false\n", + " groundtruth_points_drop_percentage: 0.0\n", + " groundtruth_drop_max_keep_points: 15\n", + " remove_unknown_examples: false\n", + " sample_importance: 1.0\n", + " random_flip_x: false\n", + " random_flip_y: true\n", + " remove_environment: false\n", + "\n", + " database_sampler {\n", + " database_info_path: \"/data/kitti_lidar/kitti_dbinfos_train.pkl\"\n", + " sample_groups {\n", + " name_to_max_num {\n", + " key: \"Car\"\n", + " value: 15\n", + " }\n", + " }\n", + " database_prep_steps {\n", + " filter_by_min_num_points {\n", + " min_num_point_pairs {\n", + " key: \"Car\"\n", + " value: 5\n", + " }\n", + " }\n", + " }\n", + " database_prep_steps {\n", + " filter_by_difficulty {\n", + " removed_difficulties: [-1]\n", + " }\n", + " }\n", + " global_random_rotation_range_per_object: [0, 0]\n", + " rate: 1.0\n", + " }\n", + " }\n", + "}\n", + "\n", + "train_config: {\n", + " optimizer: {\n", + " adam_optimizer: {\n", + " learning_rate: {\n", + " one_cycle: {\n", + " lr_max: 3e-3\n", + " moms: [0.95, 0.85]\n", + " div_factor: 10.0\n", + " pct_start: 0.4\n", + " }\n", + " }\n", + " weight_decay: 0.01\n", + " }\n", + " fixed_weight_decay: true\n", + " use_moving_average: false\n", + " }\n", + " steps: 15500 # 310 * 50\n", + " steps_per_eval: 1550 # 310 * 5\n", + " save_checkpoints_secs : 1800 # half hour\n", + " save_summary_steps : 10\n", + " enable_mixed_precision: false \n", + " loss_scale_factor: -1\n", + " clear_metrics_every_epoch: true\n", + "}\n", + "\n", + "eval_input_reader: {\n", + " dataset: {\n", + " dataset_class_name: \"KittiDataset\"\n", + " kitti_info_path: \"/data/kitti_lidar/kitti_infos_val.pkl\"\n", + " # kitti_info_path: \"/data/kitti_lidar/kitti_infos_test.pkl\"\n", + " kitti_root_path: \"/data/kitti_lidar\"\n", + " }\n", + " batch_size: 12\n", + " preprocess: {\n", + " max_number_of_voxels: 30000\n", + " shuffle_points: false\n", + " num_workers: 3\n", + " anchor_area_threshold: -1\n", + " remove_environment: false\n", + " }\n", + "}\n", + "\n", + "WORKER 0 seed: 1592956993\n", + "WORKER 1 seed: 1592956994\n", + "WORKER 2 seed: 1592956995\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "runtime.step=50, runtime.steptime=0.4401, runtime.voxel_gene_time=0.001134, runtime.prep_time=0.02529, loss.cls_loss=24.86, loss.cls_loss_rt=4.837, loss.loc_loss=2.464, loss.loc_loss_rt=1.247, loss.loc_elem=[0.03097, 0.01904, 0.1906, 0.0224, 0.04403, 0.02962, 0.2868], loss.cls_pos_rt=0.3025, loss.cls_neg_rt=4.535, loss.dir_rt=0.6855, rpn_acc=0.9656, pr.prec@10=0.001336, pr.rec@10=0.9996, pr.prec@30=0.00113, pr.rec@30=0.3379, pr.prec@50=0.00195, pr.rec@50=0.04843, pr.prec@70=0.007283, pr.rec@70=0.002041, pr.prec@80=0.009759, pr.rec@80=0.000503, pr.prec@90=0.009346, pr.rec@90=5.917e-05, pr.prec@95=0.03704, pr.rec@95=2.959e-05, misc.num_vox=175711, misc.num_pos=58, misc.num_neg=42092, misc.num_anchors=42240, misc.lr=0.0003004, misc.mem_usage=55.0\n", + "runtime.step=100, runtime.steptime=0.274, runtime.voxel_gene_time=0.001062, runtime.prep_time=0.0248, loss.cls_loss=13.62, loss.cls_loss_rt=1.094, loss.loc_loss=1.86, loss.loc_loss_rt=1.011, loss.loc_elem=[0.0269, 0.01716, 0.1313, 0.02447, 0.05143, 0.03828, 0.2158], loss.cls_pos_rt=0.3677, loss.cls_neg_rt=0.7261, loss.dir_rt=0.6889, rpn_acc=0.9821, pr.prec@10=0.001395, pr.rec@10=0.9881, pr.prec@30=0.00113, pr.rec@30=0.1691, pr.prec@50=0.00195, pr.rec@50=0.02423, pr.prec@70=0.007272, pr.rec@70=0.001021, pr.prec@80=0.009731, pr.rec@80=0.0002516, pr.prec@90=0.009259, pr.rec@90=2.96e-05, pr.prec@95=0.03704, pr.rec@95=1.48e-05, misc.num_vox=180000, misc.num_pos=65, misc.num_neg=42080, misc.num_anchors=42240, misc.lr=0.0003017, misc.mem_usage=55.0\n", + "runtime.step=150, runtime.steptime=0.2748, runtime.voxel_gene_time=0.001081, runtime.prep_time=0.02989, loss.cls_loss=9.331, loss.cls_loss_rt=0.6096, loss.loc_loss=1.532, loss.loc_loss_rt=0.8572, loss.loc_elem=[0.02011, 0.01966, 0.1229, 0.01862, 0.04352, 0.03373, 0.17], loss.cls_pos_rt=0.3852, loss.cls_neg_rt=0.2244, loss.dir_rt=0.684, rpn_acc=0.9876, pr.prec@10=0.001795, pr.rec@10=0.9047, pr.prec@30=0.00113, pr.rec@30=0.1127, pr.prec@50=0.00195, pr.rec@50=0.01615, pr.prec@70=0.007272, pr.rec@70=0.0006808, pr.prec@80=0.009731, pr.rec@80=0.0001677, pr.prec@90=0.009259, pr.rec@90=1.973e-05, pr.prec@95=0.03704, pr.rec@95=9.867e-06, misc.num_vox=179939, misc.num_pos=60, misc.num_neg=42085, misc.num_anchors=42240, misc.lr=0.0003038, misc.mem_usage=55.0\n", + "runtime.step=200, runtime.steptime=0.2742, runtime.voxel_gene_time=0.001052, runtime.prep_time=0.02944, loss.cls_loss=7.137, loss.cls_loss_rt=0.512, loss.loc_loss=1.337, loss.loc_loss_rt=0.684, loss.loc_elem=[0.01628, 0.0165, 0.07113, 0.01879, 0.04278, 0.02793, 0.1486], loss.cls_pos_rt=0.3948, loss.cls_neg_rt=0.1172, loss.dir_rt=0.6866, rpn_acc=0.9904, pr.prec@10=0.002235, pr.rec@10=0.8584, pr.prec@30=0.001156, pr.rec@30=0.08642, pr.prec@50=0.00195, pr.rec@50=0.0121, pr.prec@70=0.007272, pr.rec@70=0.0005099, pr.prec@80=0.009731, pr.rec@80=0.0001256, pr.prec@90=0.009259, pr.rec@90=1.478e-05, pr.prec@95=0.03704, pr.rec@95=7.39e-06, misc.num_vox=172972, misc.num_pos=49, misc.num_neg=42100, misc.num_anchors=42240, misc.lr=0.0003069, misc.mem_usage=55.0\n", + "runtime.step=250, runtime.steptime=0.2775, runtime.voxel_gene_time=0.001017, runtime.prep_time=0.02903, loss.cls_loss=5.807, loss.cls_loss_rt=0.4406, loss.loc_loss=1.208, loss.loc_loss_rt=0.6009, loss.loc_elem=[0.01337, 0.01388, 0.05812, 0.02102, 0.04376, 0.03144, 0.1189], loss.cls_pos_rt=0.3143, loss.cls_neg_rt=0.1263, loss.dir_rt=0.662, rpn_acc=0.9921, pr.prec@10=0.002689, pr.rec@10=0.8388, pr.prec@30=0.00139, pr.rec@30=0.08323, pr.prec@50=0.00195, pr.rec@50=0.009686, pr.prec@70=0.007272, pr.rec@70=0.0004082, pr.prec@80=0.009731, pr.rec@80=0.0001006, pr.prec@90=0.009259, pr.rec@90=1.183e-05, pr.prec@95=0.03704, pr.rec@95=5.917e-06, misc.num_vox=178528, misc.num_pos=57, misc.num_neg=42106, misc.num_anchors=42240, misc.lr=0.0003107, misc.mem_usage=54.9\n", + "reset Car\n", + "reset Car\n", + "reset Car\n", + "runtime.step=300, runtime.steptime=0.2771, runtime.voxel_gene_time=0.001551, runtime.prep_time=0.03547, loss.cls_loss=4.912, loss.cls_loss_rt=0.4106, loss.loc_loss=1.113, loss.loc_loss_rt=0.6446, loss.loc_elem=[0.01453, 0.01353, 0.07528, 0.01939, 0.04241, 0.04221, 0.1149], loss.cls_pos_rt=0.3135, loss.cls_neg_rt=0.09715, loss.dir_rt=0.6627, rpn_acc=0.9932, pr.prec@10=0.003163, pr.rec@10=0.8316, pr.prec@30=0.001905, pr.rec@30=0.09499, pr.prec@50=0.00195, pr.rec@50=0.00806, pr.prec@70=0.007272, pr.rec@70=0.0003397, pr.prec@80=0.009731, pr.rec@80=8.37e-05, pr.prec@90=0.009259, pr.rec@90=9.847e-06, pr.prec@95=0.03704, pr.rec@95=4.924e-06, misc.num_vox=179315, misc.num_pos=54, misc.num_neg=42100, misc.num_anchors=42240, misc.lr=0.0003155, misc.mem_usage=55.0\n", + "WORKER 0 seed: 1592957086\n", + "WORKER 1 seed: 1592957087\n", + "WORKER 2 seed: 1592957088\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "runtime.step=350, runtime.steptime=0.4426, runtime.voxel_gene_time=0.0009742, runtime.prep_time=0.02894, loss.cls_loss=0.4031, loss.cls_loss_rt=0.3886, loss.loc_loss=0.6038, loss.loc_loss_rt=0.617, loss.loc_elem=[0.01312, 0.01208, 0.06789, 0.02733, 0.04294, 0.03501, 0.1101], loss.cls_pos_rt=0.2941, loss.cls_neg_rt=0.09452, loss.dir_rt=0.6468, rpn_acc=0.9987, pr.prec@10=0.0451, pr.rec@10=0.805, pr.prec@30=0.5555, pr.rec@30=0.2386, pr.prec@50=0.0, pr.rec@50=0.0, pr.prec@70=0.0, pr.rec@70=0.0, pr.prec@80=0.0, pr.rec@80=0.0, pr.prec@90=0.0, pr.rec@90=0.0, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=178578, misc.num_pos=60, misc.num_neg=42084, misc.num_anchors=42240, misc.lr=0.0003211, misc.mem_usage=54.6\n", + "runtime.step=400, runtime.steptime=0.2737, runtime.voxel_gene_time=0.001112, runtime.prep_time=0.02501, loss.cls_loss=0.3892, loss.cls_loss_rt=0.3636, loss.loc_loss=0.5915, loss.loc_loss_rt=0.5685, loss.loc_elem=[0.01191, 0.01457, 0.05862, 0.01622, 0.04476, 0.03814, 0.1], loss.cls_pos_rt=0.277, loss.cls_neg_rt=0.08656, loss.dir_rt=0.6392, rpn_acc=0.9987, pr.prec@10=0.04764, pr.rec@10=0.817, pr.prec@30=0.5581, pr.rec@30=0.2649, pr.prec@50=0.0, pr.rec@50=0.0, pr.prec@70=0.0, pr.rec@70=0.0, pr.prec@80=0.0, pr.rec@80=0.0, pr.prec@90=0.0, pr.rec@90=0.0, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=173229, misc.num_pos=64, misc.num_neg=42078, misc.num_anchors=42240, misc.lr=0.0003275, misc.mem_usage=54.9\n", + "runtime.step=450, runtime.steptime=0.2783, runtime.voxel_gene_time=0.001094, runtime.prep_time=0.02822, loss.cls_loss=0.3754, loss.cls_loss_rt=0.3727, loss.loc_loss=0.5695, loss.loc_loss_rt=0.5917, loss.loc_elem=[0.01412, 0.01033, 0.08402, 0.02174, 0.03492, 0.0277, 0.103], loss.cls_pos_rt=0.2787, loss.cls_neg_rt=0.09402, loss.dir_rt=0.6283, rpn_acc=0.9987, pr.prec@10=0.0503, pr.rec@10=0.8287, pr.prec@30=0.5474, pr.rec@30=0.2953, pr.prec@50=1.0, pr.rec@50=0.001692, pr.prec@70=0.0, pr.rec@70=0.0, pr.prec@80=0.0, pr.rec@80=0.0, pr.prec@90=0.0, pr.rec@90=0.0, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=176904, misc.num_pos=57, misc.num_neg=42095, misc.num_anchors=42240, misc.lr=0.0003348, misc.mem_usage=54.9\n", + "runtime.step=500, runtime.steptime=0.2685, runtime.voxel_gene_time=0.001135, runtime.prep_time=0.03179, loss.cls_loss=0.3632, loss.cls_loss_rt=0.3386, loss.loc_loss=0.5592, loss.loc_loss_rt=0.5486, loss.loc_elem=[0.01289, 0.009706, 0.07254, 0.01846, 0.03503, 0.03191, 0.09378], loss.cls_pos_rt=0.2658, loss.cls_neg_rt=0.07287, loss.dir_rt=0.5804, rpn_acc=0.9987, pr.prec@10=0.05308, pr.rec@10=0.8358, pr.prec@30=0.5513, pr.rec@30=0.3236, pr.prec@50=0.9783, pr.rec@50=0.008013, pr.prec@70=0.0, pr.rec@70=0.0, pr.prec@80=0.0, pr.rec@80=0.0, pr.prec@90=0.0, pr.rec@90=0.0, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=174232, misc.num_pos=48, misc.num_neg=42107, misc.num_anchors=42240, misc.lr=0.0003429, misc.mem_usage=54.9\n", + "runtime.step=550, runtime.steptime=0.2719, runtime.voxel_gene_time=0.001074, runtime.prep_time=0.0241, loss.cls_loss=0.3547, loss.cls_loss_rt=0.3074, loss.loc_loss=0.5508, loss.loc_loss_rt=0.4928, loss.loc_elem=[0.0133, 0.008642, 0.06059, 0.01825, 0.03917, 0.03103, 0.0754], loss.cls_pos_rt=0.2399, loss.cls_neg_rt=0.06746, loss.dir_rt=0.5497, rpn_acc=0.9987, pr.prec@10=0.05511, pr.rec@10=0.8398, pr.prec@30=0.5494, pr.rec@30=0.3453, pr.prec@50=0.9686, pr.rec@50=0.01681, pr.prec@70=0.0, pr.rec@70=0.0, pr.prec@80=0.0, pr.rec@80=0.0, pr.prec@90=0.0, pr.rec@90=0.0, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=172272, misc.num_pos=53, misc.num_neg=42094, misc.num_anchors=42240, misc.lr=0.0003519, misc.mem_usage=55.0\n", + "runtime.step=600, runtime.steptime=0.2754, runtime.voxel_gene_time=0.001063, runtime.prep_time=0.02486, loss.cls_loss=0.347, loss.cls_loss_rt=0.2975, loss.loc_loss=0.5423, loss.loc_loss_rt=0.4531, loss.loc_elem=[0.01205, 0.01011, 0.04684, 0.01865, 0.04314, 0.02733, 0.06844], loss.cls_pos_rt=0.1901, loss.cls_neg_rt=0.1075, loss.dir_rt=0.5579, rpn_acc=0.9987, pr.prec@10=0.05696, pr.rec@10=0.8429, pr.prec@30=0.5483, pr.rec@30=0.3652, pr.prec@50=0.9543, pr.rec@50=0.03029, pr.prec@70=1.0, pr.rec@70=5.066e-06, pr.prec@80=0.0, pr.rec@80=0.0, pr.prec@90=0.0, pr.rec@90=0.0, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=179367, misc.num_pos=60, misc.num_neg=42082, misc.num_anchors=42240, misc.lr=0.0003617, misc.mem_usage=55.0\n", + "reset Car\n", + "reset Car\n", + "reset Car\n", + "WORKER 0 seed: 1592957179\n", + "WORKER 1 seed: 1592957180\n", + "WORKER 2 seed: 1592957181\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "runtime.step=650, runtime.steptime=0.449, runtime.voxel_gene_time=0.00128, runtime.prep_time=0.03682, loss.cls_loss=0.2962, loss.cls_loss_rt=0.3197, loss.loc_loss=0.4854, loss.loc_loss_rt=0.5052, loss.loc_elem=[0.01315, 0.01178, 0.05147, 0.02043, 0.04248, 0.03514, 0.07812], loss.cls_pos_rt=0.2368, loss.cls_neg_rt=0.08293, loss.dir_rt=0.5398, rpn_acc=0.9988, pr.prec@10=0.07231, pr.rec@10=0.8655, pr.prec@30=0.5701, pr.rec@30=0.4824, pr.prec@50=0.952, pr.rec@50=0.1299, pr.prec@70=0.0, pr.rec@70=0.0, pr.prec@80=0.0, pr.rec@80=0.0, pr.prec@90=0.0, pr.rec@90=0.0, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=176121, misc.num_pos=62, misc.num_neg=42083, misc.num_anchors=42240, misc.lr=0.0003723, misc.mem_usage=54.7\n", + "runtime.step=700, runtime.steptime=0.2785, runtime.voxel_gene_time=0.001573, runtime.prep_time=0.04145, loss.cls_loss=0.2909, loss.cls_loss_rt=0.2692, loss.loc_loss=0.4786, loss.loc_loss_rt=0.5018, loss.loc_elem=[0.01281, 0.009715, 0.06857, 0.0181, 0.04347, 0.02709, 0.07117], loss.cls_pos_rt=0.2004, loss.cls_neg_rt=0.06878, loss.dir_rt=0.467, rpn_acc=0.9988, pr.prec@10=0.07419, pr.rec@10=0.8699, pr.prec@30=0.5672, pr.rec@30=0.4984, pr.prec@50=0.9446, pr.rec@50=0.1369, pr.prec@70=1.0, pr.rec@70=0.0001257, pr.prec@80=0.0, pr.rec@80=0.0, pr.prec@90=0.0, pr.rec@90=0.0, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=176616, misc.num_pos=68, misc.num_neg=42082, misc.num_anchors=42240, misc.lr=0.0003838, misc.mem_usage=54.9\n", + "runtime.step=750, runtime.steptime=0.279, runtime.voxel_gene_time=0.0009129, runtime.prep_time=0.02881, loss.cls_loss=0.2861, loss.cls_loss_rt=0.302, loss.loc_loss=0.4774, loss.loc_loss_rt=0.4628, loss.loc_elem=[0.01044, 0.008871, 0.05359, 0.01719, 0.03499, 0.03517, 0.07115], loss.cls_pos_rt=0.2392, loss.cls_neg_rt=0.06281, loss.dir_rt=0.4707, rpn_acc=0.9988, pr.prec@10=0.0758, pr.rec@10=0.8725, pr.prec@30=0.5727, pr.rec@30=0.5067, pr.prec@50=0.9437, pr.rec@50=0.1497, pr.prec@70=0.9636, pr.rec@70=0.0005923, pr.prec@80=0.0, pr.rec@80=0.0, pr.prec@90=0.0, pr.rec@90=0.0, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=177819, misc.num_pos=55, misc.num_neg=42099, misc.num_anchors=42240, misc.lr=0.0003961, misc.mem_usage=55.0\n", + "runtime.step=800, runtime.steptime=0.2797, runtime.voxel_gene_time=0.001053, runtime.prep_time=0.02892, loss.cls_loss=0.2809, loss.cls_loss_rt=0.2734, loss.loc_loss=0.4714, loss.loc_loss_rt=0.456, loss.loc_elem=[0.01065, 0.008016, 0.05615, 0.01532, 0.04215, 0.03372, 0.06201], loss.cls_pos_rt=0.1982, loss.cls_neg_rt=0.07519, loss.dir_rt=0.434, rpn_acc=0.9989, pr.prec@10=0.07789, pr.rec@10=0.8751, pr.prec@30=0.5753, pr.rec@30=0.5182, pr.prec@50=0.9429, pr.rec@50=0.1669, pr.prec@70=0.9944, pr.rec@70=0.002898, pr.prec@80=0.0, pr.rec@80=0.0, pr.prec@90=0.0, pr.rec@90=0.0, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=178665, misc.num_pos=42, misc.num_neg=42121, misc.num_anchors=42240, misc.lr=0.0004091, misc.mem_usage=54.9\n", + "runtime.step=850, runtime.steptime=0.2744, runtime.voxel_gene_time=0.001728, runtime.prep_time=0.04476, loss.cls_loss=0.2777, loss.cls_loss_rt=0.2463, loss.loc_loss=0.4677, loss.loc_loss_rt=0.3962, loss.loc_elem=[0.01103, 0.00805, 0.04668, 0.01747, 0.03184, 0.02804, 0.05501], loss.cls_pos_rt=0.1617, loss.cls_neg_rt=0.0846, loss.dir_rt=0.4712, rpn_acc=0.9989, pr.prec@10=0.07942, pr.rec@10=0.8769, pr.prec@30=0.5758, pr.rec@30=0.5249, pr.prec@50=0.9394, pr.rec@50=0.1772, pr.prec@70=0.9952, pr.rec@70=0.005311, pr.prec@80=1.0, pr.rec@80=1.915e-05, pr.prec@90=0.0, pr.rec@90=0.0, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=178964, misc.num_pos=63, misc.num_neg=42075, misc.num_anchors=42240, misc.lr=0.000423, misc.mem_usage=54.9\n", + "runtime.step=900, runtime.steptime=0.28, runtime.voxel_gene_time=0.001042, runtime.prep_time=0.0316, loss.cls_loss=0.2738, loss.cls_loss_rt=0.2262, loss.loc_loss=0.4629, loss.loc_loss_rt=0.395, loss.loc_elem=[0.009199, 0.006592, 0.04198, 0.0168, 0.04713, 0.0256, 0.05018], loss.cls_pos_rt=0.1706, loss.cls_neg_rt=0.05565, loss.dir_rt=0.4577, rpn_acc=0.9989, pr.prec@10=0.08119, pr.rec@10=0.8791, pr.prec@30=0.5777, pr.rec@30=0.5346, pr.prec@50=0.9372, pr.rec@50=0.1914, pr.prec@70=0.9913, pr.rec@70=0.00835, pr.prec@80=1.0, pr.rec@80=2.094e-05, pr.prec@90=0.0, pr.rec@90=0.0, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=176369, misc.num_pos=56, misc.num_neg=42087, misc.num_anchors=42240, misc.lr=0.0004377, misc.mem_usage=54.7\n", + "reset Car\n", + "reset Car\n", + "reset Car\n", + "WORKER 0 seed: 1592957274\n", + "WORKER 1 seed: 1592957275\n", + "WORKER 2 seed: 1592957276\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "runtime.step=950, runtime.steptime=0.4422, runtime.voxel_gene_time=0.0009515, runtime.prep_time=0.02646, loss.cls_loss=0.259, loss.cls_loss_rt=0.235, loss.loc_loss=0.4283, loss.loc_loss_rt=0.448, loss.loc_elem=[0.007915, 0.007348, 0.04963, 0.02229, 0.03751, 0.03637, 0.06294], loss.cls_pos_rt=0.1812, loss.cls_neg_rt=0.05377, loss.dir_rt=0.5463, rpn_acc=0.999, pr.prec@10=0.08791, pr.rec@10=0.8849, pr.prec@30=0.5982, pr.rec@30=0.5628, pr.prec@50=0.9296, pr.rec@50=0.2518, pr.prec@70=0.9886, pr.rec@70=0.02769, pr.prec@80=1.0, pr.rec@80=0.0001276, pr.prec@90=0.0, pr.rec@90=0.0, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=179583, misc.num_pos=48, misc.num_neg=42131, misc.num_anchors=42240, misc.lr=0.0004531, misc.mem_usage=54.8\n", + "runtime.step=1000, runtime.steptime=0.2786, runtime.voxel_gene_time=0.001486, runtime.prep_time=0.0347, loss.cls_loss=0.2559, loss.cls_loss_rt=0.2483, loss.loc_loss=0.4436, loss.loc_loss_rt=0.4168, loss.loc_elem=[0.01066, 0.008584, 0.0417, 0.01674, 0.03731, 0.03558, 0.05781], loss.cls_pos_rt=0.1856, loss.cls_neg_rt=0.06276, loss.dir_rt=0.3831, rpn_acc=0.999, pr.prec@10=0.08954, pr.rec@10=0.8852, pr.prec@30=0.5977, pr.rec@30=0.5754, pr.prec@50=0.9282, pr.rec@50=0.2626, pr.prec@70=0.9945, pr.rec@70=0.03282, pr.prec@80=1.0, pr.rec@80=0.0006447, pr.prec@90=0.0, pr.rec@90=0.0, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=176796, misc.num_pos=50, misc.num_neg=42098, misc.num_anchors=42240, misc.lr=0.0004693, misc.mem_usage=55.1\n", + "runtime.step=1050, runtime.steptime=0.2794, runtime.voxel_gene_time=0.001048, runtime.prep_time=0.03681, loss.cls_loss=0.2487, loss.cls_loss_rt=0.2386, loss.loc_loss=0.4347, loss.loc_loss_rt=0.4283, loss.loc_elem=[0.009527, 0.007347, 0.0511, 0.01769, 0.04012, 0.03532, 0.05305], loss.cls_pos_rt=0.1869, loss.cls_neg_rt=0.05165, loss.dir_rt=0.3541, rpn_acc=0.999, pr.prec@10=0.09246, pr.rec@10=0.89, pr.prec@30=0.6032, pr.rec@30=0.5877, pr.prec@50=0.9326, pr.rec@50=0.2763, pr.prec@70=0.9936, pr.rec@70=0.03527, pr.prec@80=1.0, pr.rec@80=0.0006828, pr.prec@90=0.0, pr.rec@90=0.0, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=174874, misc.num_pos=59, misc.num_neg=42096, misc.num_anchors=42240, misc.lr=0.0004863, misc.mem_usage=55.0\n", + "runtime.step=1100, runtime.steptime=0.2787, runtime.voxel_gene_time=0.001334, runtime.prep_time=0.03353, loss.cls_loss=0.2448, loss.cls_loss_rt=0.2402, loss.loc_loss=0.4291, loss.loc_loss_rt=0.4625, loss.loc_elem=[0.009382, 0.007632, 0.06696, 0.02129, 0.03809, 0.03297, 0.05491], loss.cls_pos_rt=0.174, loss.cls_neg_rt=0.06623, loss.dir_rt=0.4463, rpn_acc=0.999, pr.prec@10=0.09478, pr.rec@10=0.893, pr.prec@30=0.6082, pr.rec@30=0.5954, pr.prec@50=0.9322, pr.rec@50=0.2873, pr.prec@70=0.9936, pr.rec@70=0.04105, pr.prec@80=1.0, pr.rec@80=0.001314, pr.prec@90=0.0, pr.rec@90=0.0, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=175451, misc.num_pos=41, misc.num_neg=42133, misc.num_anchors=42240, misc.lr=0.000504, misc.mem_usage=54.9\n", + "runtime.step=1150, runtime.steptime=0.2801, runtime.voxel_gene_time=0.001027, runtime.prep_time=0.02963, loss.cls_loss=0.2427, loss.cls_loss_rt=0.2349, loss.loc_loss=0.4252, loss.loc_loss_rt=0.4272, loss.loc_elem=[0.009603, 0.008624, 0.05688, 0.01823, 0.03166, 0.02734, 0.06127], loss.cls_pos_rt=0.1581, loss.cls_neg_rt=0.07681, loss.dir_rt=0.4317, rpn_acc=0.999, pr.prec@10=0.09669, pr.rec@10=0.8938, pr.prec@30=0.6079, pr.rec@30=0.5999, pr.prec@50=0.9302, pr.rec@50=0.2959, pr.prec@70=0.9939, pr.rec@70=0.04517, pr.prec@80=1.0, pr.rec@80=0.002, pr.prec@90=0.0, pr.rec@90=0.0, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=178219, misc.num_pos=49, misc.num_neg=42108, misc.num_anchors=42240, misc.lr=0.0005224, misc.mem_usage=54.9\n", + "runtime.step=1200, runtime.steptime=0.2811, runtime.voxel_gene_time=0.001138, runtime.prep_time=0.03431, loss.cls_loss=0.2391, loss.cls_loss_rt=0.2447, loss.loc_loss=0.4213, loss.loc_loss_rt=0.4232, loss.loc_elem=[0.007776, 0.006655, 0.05873, 0.01907, 0.03636, 0.02891, 0.05409], loss.cls_pos_rt=0.1918, loss.cls_neg_rt=0.05289, loss.dir_rt=0.4535, rpn_acc=0.999, pr.prec@10=0.09867, pr.rec@10=0.8959, pr.prec@30=0.613, pr.rec@30=0.6073, pr.prec@50=0.9312, pr.rec@50=0.3052, pr.prec@70=0.9945, pr.rec@70=0.05112, pr.prec@80=1.0, pr.rec@80=0.003553, pr.prec@90=0.0, pr.rec@90=0.0, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=176360, misc.num_pos=68, misc.num_neg=42073, misc.num_anchors=42240, misc.lr=0.0005416, misc.mem_usage=54.9\n", + "reset Car\n", + "reset Car\n", + "reset Car\n", + "WORKER 0 seed: 1592957368\n", + "WORKER 1 seed: 1592957369\n", + "WORKER 2 seed: 1592957370\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "runtime.step=1250, runtime.steptime=0.4469, runtime.voxel_gene_time=0.0009129, runtime.prep_time=0.02702, loss.cls_loss=0.2308, loss.cls_loss_rt=0.2179, loss.loc_loss=0.402, loss.loc_loss_rt=0.3893, loss.loc_elem=[0.008508, 0.007158, 0.03564, 0.01644, 0.03704, 0.03818, 0.05167], loss.cls_pos_rt=0.1453, loss.cls_neg_rt=0.07258, loss.dir_rt=0.4257, rpn_acc=0.9991, pr.prec@10=0.101, pr.rec@10=0.8963, pr.prec@30=0.6269, pr.rec@30=0.6297, pr.prec@50=0.9277, pr.rec@50=0.3535, pr.prec@70=0.9958, pr.rec@70=0.07422, pr.prec@80=1.0, pr.rec@80=0.006681, pr.prec@90=0.0, pr.rec@90=0.0, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=173062, misc.num_pos=50, misc.num_neg=42100, misc.num_anchors=42240, misc.lr=0.0005615, misc.mem_usage=54.7\n", + "runtime.step=1300, runtime.steptime=0.2762, runtime.voxel_gene_time=0.001238, runtime.prep_time=0.02776, loss.cls_loss=0.2291, loss.cls_loss_rt=0.2067, loss.loc_loss=0.4056, loss.loc_loss_rt=0.4095, loss.loc_elem=[0.009017, 0.008203, 0.05123, 0.01539, 0.03826, 0.02415, 0.05849], loss.cls_pos_rt=0.1695, loss.cls_neg_rt=0.03717, loss.dir_rt=0.3869, rpn_acc=0.9991, pr.prec@10=0.106, pr.rec@10=0.8954, pr.prec@30=0.6312, pr.rec@30=0.6329, pr.prec@50=0.9301, pr.rec@50=0.3492, pr.prec@70=0.9944, pr.rec@70=0.07403, pr.prec@80=1.0, pr.rec@80=0.008267, pr.prec@90=0.0, pr.rec@90=0.0, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=180000, misc.num_pos=50, misc.num_neg=42115, misc.num_anchors=42240, misc.lr=0.000582, misc.mem_usage=54.9\n", + "runtime.step=1350, runtime.steptime=0.2745, runtime.voxel_gene_time=0.001074, runtime.prep_time=0.03886, loss.cls_loss=0.2254, loss.cls_loss_rt=0.1962, loss.loc_loss=0.4015, loss.loc_loss_rt=0.3606, loss.loc_elem=[0.008158, 0.005257, 0.04004, 0.01836, 0.02925, 0.03048, 0.04874], loss.cls_pos_rt=0.138, loss.cls_neg_rt=0.05811, loss.dir_rt=0.3278, rpn_acc=0.9991, pr.prec@10=0.1072, pr.rec@10=0.8995, pr.prec@30=0.6321, pr.rec@30=0.64, pr.prec@50=0.931, pr.rec@50=0.3588, pr.prec@70=0.9944, pr.rec@70=0.08324, pr.prec@80=1.0, pr.rec@80=0.009438, pr.prec@90=0.0, pr.rec@90=0.0, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=175357, misc.num_pos=57, misc.num_neg=42101, misc.num_anchors=42240, misc.lr=0.0006033, misc.mem_usage=54.9\n", + "runtime.step=1400, runtime.steptime=0.2736, runtime.voxel_gene_time=0.001632, runtime.prep_time=0.04029, loss.cls_loss=0.2227, loss.cls_loss_rt=0.2426, loss.loc_loss=0.3984, loss.loc_loss_rt=0.409, loss.loc_elem=[0.008932, 0.008388, 0.03535, 0.01938, 0.04116, 0.028, 0.06329], loss.cls_pos_rt=0.1959, loss.cls_neg_rt=0.04671, loss.dir_rt=0.3454, rpn_acc=0.9991, pr.prec@10=0.1093, pr.rec@10=0.9011, pr.prec@30=0.6352, pr.rec@30=0.6456, pr.prec@50=0.9305, pr.rec@50=0.3666, pr.prec@70=0.9932, pr.rec@70=0.09112, pr.prec@80=0.9993, pr.rec@80=0.01241, pr.prec@90=1.0, pr.rec@90=9.04e-06, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=177361, misc.num_pos=50, misc.num_neg=42115, misc.num_anchors=42240, misc.lr=0.0006252, misc.mem_usage=55.0\n", + "runtime.step=1450, runtime.steptime=0.2767, runtime.voxel_gene_time=0.001601, runtime.prep_time=0.03736, loss.cls_loss=0.2201, loss.cls_loss_rt=0.1957, loss.loc_loss=0.3959, loss.loc_loss_rt=0.3712, loss.loc_elem=[0.008162, 0.007598, 0.02893, 0.01534, 0.0382, 0.03442, 0.05293], loss.cls_pos_rt=0.1459, loss.cls_neg_rt=0.04985, loss.dir_rt=0.3425, rpn_acc=0.9991, pr.prec@10=0.1105, pr.rec@10=0.9028, pr.prec@30=0.6395, pr.rec@30=0.6486, pr.prec@50=0.9305, pr.rec@50=0.3716, pr.prec@70=0.9918, pr.rec@70=0.09379, pr.prec@80=0.9984, pr.rec@80=0.01274, pr.prec@90=1.0, pr.rec@90=1.383e-05, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=173679, misc.num_pos=52, misc.num_neg=42124, misc.num_anchors=42240, misc.lr=0.0006478, misc.mem_usage=54.9\n", + "runtime.step=1500, runtime.steptime=0.2752, runtime.voxel_gene_time=0.001013, runtime.prep_time=0.03071, loss.cls_loss=0.2184, loss.cls_loss_rt=0.1811, loss.loc_loss=0.3949, loss.loc_loss_rt=0.3372, loss.loc_elem=[0.00755, 0.006407, 0.04505, 0.012, 0.02979, 0.02752, 0.04027], loss.cls_pos_rt=0.1292, loss.cls_neg_rt=0.05191, loss.dir_rt=0.3468, rpn_acc=0.9991, pr.prec@10=0.1116, pr.rec@10=0.9041, pr.prec@30=0.6405, pr.rec@30=0.6532, pr.prec@50=0.9304, pr.rec@50=0.3761, pr.prec@70=0.9916, pr.rec@70=0.09503, pr.prec@80=0.9982, pr.rec@80=0.01263, pr.prec@90=1.0, pr.rec@90=1.12e-05, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=179401, misc.num_pos=57, misc.num_neg=42093, misc.num_anchors=42240, misc.lr=0.0006711, misc.mem_usage=55.0\n", + "reset Car\n", + "reset Car\n", + "reset Car\n", + "WORKER 0 seed: 1592957462\n", + "WORKER 1 seed: 1592957463\n", + "WORKER 2 seed: 1592957464\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "runtime.step=1550, runtime.steptime=0.4378, runtime.voxel_gene_time=0.003899, runtime.prep_time=0.02941, loss.cls_loss=0.2239, loss.cls_loss_rt=0.2573, loss.loc_loss=0.3976, loss.loc_loss_rt=0.4408, loss.loc_elem=[0.01192, 0.007672, 0.05855, 0.01773, 0.03171, 0.04171, 0.05109], loss.cls_pos_rt=0.2011, loss.cls_neg_rt=0.05615, loss.dir_rt=0.4336, rpn_acc=0.9992, pr.prec@10=0.104, pr.rec@10=0.8985, pr.prec@30=0.6041, pr.rec@30=0.6603, pr.prec@50=0.9235, pr.rec@50=0.4156, pr.prec@70=0.9915, pr.rec@70=0.1382, pr.prec@80=1.0, pr.rec@80=0.03322, pr.prec@90=0.0, pr.rec@90=0.0, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=177447, misc.num_pos=41, misc.num_neg=42127, misc.num_anchors=42240, misc.lr=0.0006949, misc.mem_usage=54.8\n", + "#################################\n", + "# EVAL\n", + "#################################\n", + "Generate output labels...\n", + "[100.0%][===================>][11.94it/s][00:27>00:00] \n", + "generate label finished(134.56/s). start eval:\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typed_passes.py:314: NumbaPerformanceWarning: \n", + "The keyword argument 'parallel=True' was specified but no transformation for parallel execution was possible.\n", + "\n", + "To find out why, try turning on parallel diagnostics, see http://numba.pydata.org/numba-doc/latest/user/parallel.html#diagnostics for help.\n", + "\n", + "File \"../../opt/second.pytorch/second/utils/eval.py\", line 129:\n", + "@numba.jit(nopython=True, parallel=True)\n", + "def box3d_overlap_kernel(boxes,\n", + "^\n", + "\n", + " state.func_ir.loc))\n", + "Evaluation official\n", + "Car AP(Average Precision)@0.70, 0.70, 0.70:\n", + "bbox AP:56.03, 53.21, 52.34\n", + "bev AP:84.46, 71.07, 66.01\n", + "3d AP:18.40, 19.53, 19.02\n", + "aos AP:1.99, 2.69, 3.38\n", + "Car AP(Average Precision)@0.70, 0.50, 0.50:\n", + "bbox AP:56.03, 53.21, 52.34\n", + "bev AP:94.05, 86.37, 83.00\n", + "3d AP:88.08, 82.14, 76.95\n", + "aos AP:1.99, 2.69, 3.38\n", + "\n", + "Evaluation coco\n", + "Car coco AP@0.50:0.05:0.95:\n", + "bbox AP:46.02, 43.09, 41.89\n", + "bev AP:55.36, 49.90, 47.31\n", + "3d AP:30.11, 28.43, 27.43\n", + "aos AP:1.67, 2.39, 3.01\n", + "\n", + "eval.kitti.official.Car.bbox@0.70=[56.03, 53.21, 52.34], eval.kitti.official.Car.bev@0.70=[84.46, 71.07, 66.01], eval.kitti.official.Car.3d@0.70=[18.4, 19.53, 19.02], eval.kitti.official.Car.aos=[1.994, 2.692, 3.378], eval.kitti.official.Car.bev@0.50=[94.05, 86.37, 83.0], eval.kitti.official.Car.3d@0.50=[88.08, 82.14, 76.95], eval.kitti.coco.Car.bbox=[46.02, 43.09, 41.89], eval.kitti.coco.Car.bev=[55.36, 49.9, 47.31], eval.kitti.coco.Car.3d=[30.11, 28.43, 27.43], eval.kitti.coco.Car.aos=[1.667, 2.393, 3.012]\n", + "runtime.step=1600, runtime.steptime=1.174, runtime.voxel_gene_time=0.001097, runtime.prep_time=0.03469, loss.cls_loss=0.2126, loss.cls_loss_rt=0.189, loss.loc_loss=0.3865, loss.loc_loss_rt=0.3893, loss.loc_elem=[0.008575, 0.006055, 0.05608, 0.01807, 0.03049, 0.0317, 0.04369], loss.cls_pos_rt=0.1402, loss.cls_neg_rt=0.0488, loss.dir_rt=0.3277, rpn_acc=0.9992, pr.prec@10=0.115, pr.rec@10=0.9076, pr.prec@30=0.6464, pr.rec@30=0.6632, pr.prec@50=0.9295, pr.rec@50=0.3981, pr.prec@70=0.9918, pr.rec@70=0.1169, pr.prec@80=1.0, pr.rec@80=0.01763, pr.prec@90=0.0, pr.rec@90=0.0, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=175583, misc.num_pos=38, misc.num_neg=42127, misc.num_anchors=42240, misc.lr=0.0007194, misc.mem_usage=55.2\n", + "runtime.step=1650, runtime.steptime=0.2794, runtime.voxel_gene_time=0.0008893, runtime.prep_time=0.02841, loss.cls_loss=0.2074, loss.cls_loss_rt=0.1947, loss.loc_loss=0.3748, loss.loc_loss_rt=0.3307, loss.loc_elem=[0.006794, 0.005861, 0.03, 0.0154, 0.03561, 0.03241, 0.03929], loss.cls_pos_rt=0.1077, loss.cls_neg_rt=0.08702, loss.dir_rt=0.3097, rpn_acc=0.9992, pr.prec@10=0.1185, pr.rec@10=0.9112, pr.prec@30=0.6507, pr.rec@30=0.6722, pr.prec@50=0.9294, pr.rec@50=0.4103, pr.prec@70=0.9923, pr.rec@70=0.1257, pr.prec@80=0.9987, pr.rec@80=0.02135, pr.prec@90=1.0, pr.rec@90=2.814e-05, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=177021, misc.num_pos=46, misc.num_neg=42131, misc.num_anchors=42240, misc.lr=0.0007445, misc.mem_usage=55.2\n", + "runtime.step=1700, runtime.steptime=0.2782, runtime.voxel_gene_time=0.0009263, runtime.prep_time=0.02934, loss.cls_loss=0.2061, loss.cls_loss_rt=0.2219, loss.loc_loss=0.3737, loss.loc_loss_rt=0.3976, loss.loc_elem=[0.008844, 0.006015, 0.04677, 0.01775, 0.03542, 0.03172, 0.05226], loss.cls_pos_rt=0.146, loss.cls_neg_rt=0.07589, loss.dir_rt=0.2992, rpn_acc=0.9992, pr.prec@10=0.1196, pr.rec@10=0.9115, pr.prec@30=0.6528, pr.rec@30=0.6733, pr.prec@50=0.9307, pr.rec@50=0.4128, pr.prec@70=0.9925, pr.rec@70=0.1306, pr.prec@80=0.9988, pr.rec@80=0.02415, pr.prec@90=1.0, pr.rec@90=5.715e-05, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=172624, misc.num_pos=50, misc.num_neg=42106, misc.num_anchors=42240, misc.lr=0.0007701, misc.mem_usage=55.2\n", + "runtime.step=1750, runtime.steptime=0.2768, runtime.voxel_gene_time=0.0009844, runtime.prep_time=0.02914, loss.cls_loss=0.2042, loss.cls_loss_rt=0.1815, loss.loc_loss=0.3719, loss.loc_loss_rt=0.3407, loss.loc_elem=[0.007489, 0.006971, 0.03104, 0.02018, 0.034, 0.03333, 0.03735], loss.cls_pos_rt=0.1313, loss.cls_neg_rt=0.0502, loss.dir_rt=0.3005, rpn_acc=0.9992, pr.prec@10=0.1218, pr.rec@10=0.9129, pr.prec@30=0.6548, pr.rec@30=0.6768, pr.prec@50=0.93, pr.rec@50=0.4166, pr.prec@70=0.9918, pr.rec@70=0.1319, pr.prec@80=0.9985, pr.rec@80=0.02369, pr.prec@90=1.0, pr.rec@90=6.494e-05, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=178634, misc.num_pos=55, misc.num_neg=42115, misc.num_anchors=42240, misc.lr=0.0007963, misc.mem_usage=55.2\n", + "runtime.step=1800, runtime.steptime=0.2769, runtime.voxel_gene_time=0.00154, runtime.prep_time=0.03603, loss.cls_loss=0.2039, loss.cls_loss_rt=0.2395, loss.loc_loss=0.3712, loss.loc_loss_rt=0.3768, loss.loc_elem=[0.009941, 0.009086, 0.05032, 0.01627, 0.03403, 0.02479, 0.04397], loss.cls_pos_rt=0.1679, loss.cls_neg_rt=0.07158, loss.dir_rt=0.2938, rpn_acc=0.9992, pr.prec@10=0.1224, pr.rec@10=0.9124, pr.prec@30=0.6561, pr.rec@30=0.6785, pr.prec@50=0.93, pr.rec@50=0.4195, pr.prec@70=0.9924, pr.rec@70=0.1337, pr.prec@80=0.9986, pr.rec@80=0.02449, pr.prec@90=1.0, pr.rec@90=6.948e-05, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=178856, misc.num_pos=63, misc.num_neg=42085, misc.num_anchors=42240, misc.lr=0.0008231, misc.mem_usage=55.2\n", + "reset Car\n", + "reset Car\n", + "reset Car\n", + "runtime.step=1850, runtime.steptime=0.2779, runtime.voxel_gene_time=0.001382, runtime.prep_time=0.03604, loss.cls_loss=0.2027, loss.cls_loss_rt=0.225, loss.loc_loss=0.3705, loss.loc_loss_rt=0.3913, loss.loc_elem=[0.008464, 0.006591, 0.05613, 0.01576, 0.03012, 0.03514, 0.04343], loss.cls_pos_rt=0.1853, loss.cls_neg_rt=0.03974, loss.dir_rt=0.291, rpn_acc=0.9992, pr.prec@10=0.1233, pr.rec@10=0.9131, pr.prec@30=0.6585, pr.rec@30=0.6799, pr.prec@50=0.9315, pr.rec@50=0.4229, pr.prec@70=0.9922, pr.rec@70=0.1357, pr.prec@80=0.9988, pr.rec@80=0.02465, pr.prec@90=1.0, pr.rec@90=9.194e-05, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=179431, misc.num_pos=53, misc.num_neg=42120, misc.num_anchors=42240, misc.lr=0.0008504, misc.mem_usage=55.1\n", + "WORKER 0 seed: 1592957601\n", + "WORKER 1 seed: 1592957602\n", + "WORKER 2 seed: 1592957603\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "runtime.step=1900, runtime.steptime=0.4341, runtime.voxel_gene_time=0.000941, runtime.prep_time=0.02897, loss.cls_loss=0.2017, loss.cls_loss_rt=0.1844, loss.loc_loss=0.3676, loss.loc_loss_rt=0.332, loss.loc_elem=[0.00781, 0.004897, 0.03133, 0.01925, 0.0342, 0.03319, 0.0353], loss.cls_pos_rt=0.1324, loss.cls_neg_rt=0.05203, loss.dir_rt=0.2857, rpn_acc=0.9992, pr.prec@10=0.1237, pr.rec@10=0.9136, pr.prec@30=0.6617, pr.rec@30=0.6842, pr.prec@50=0.9292, pr.rec@50=0.4229, pr.prec@70=0.9914, pr.rec@70=0.1338, pr.prec@80=1.0, pr.rec@80=0.02375, pr.prec@90=1.0, pr.rec@90=3.205e-05, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=177172, misc.num_pos=59, misc.num_neg=42111, misc.num_anchors=42240, misc.lr=0.0008782, misc.mem_usage=55.2\n", + "runtime.step=1950, runtime.steptime=0.269, runtime.voxel_gene_time=0.0008852, runtime.prep_time=0.02698, loss.cls_loss=0.1963, loss.cls_loss_rt=0.1827, loss.loc_loss=0.3644, loss.loc_loss_rt=0.3593, loss.loc_elem=[0.006928, 0.005793, 0.04827, 0.01765, 0.03081, 0.02598, 0.04422], loss.cls_pos_rt=0.1439, loss.cls_neg_rt=0.03884, loss.dir_rt=0.2968, rpn_acc=0.9992, pr.prec@10=0.1282, pr.rec@10=0.9179, pr.prec@30=0.6679, pr.rec@30=0.6939, pr.prec@50=0.9306, pr.rec@50=0.4356, pr.prec@70=0.9919, pr.rec@70=0.1447, pr.prec@80=1.0, pr.rec@80=0.02582, pr.prec@90=1.0, pr.rec@90=4.581e-05, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=178325, misc.num_pos=61, misc.num_neg=42087, misc.num_anchors=42240, misc.lr=0.0009065, misc.mem_usage=55.2\n", + "runtime.step=2000, runtime.steptime=0.2696, runtime.voxel_gene_time=0.0008154, runtime.prep_time=0.02578, loss.cls_loss=0.1972, loss.cls_loss_rt=0.1858, loss.loc_loss=0.3654, loss.loc_loss_rt=0.3533, loss.loc_elem=[0.007375, 0.006603, 0.03846, 0.01391, 0.03059, 0.03307, 0.04667], loss.cls_pos_rt=0.1367, loss.cls_neg_rt=0.04903, loss.dir_rt=0.271, rpn_acc=0.9992, pr.prec@10=0.1272, pr.rec@10=0.9172, pr.prec@30=0.6678, pr.rec@30=0.6909, pr.prec@50=0.9296, pr.rec@50=0.4341, pr.prec@70=0.9915, pr.rec@70=0.1435, pr.prec@80=0.9996, pr.rec@80=0.02675, pr.prec@90=1.0, pr.rec@90=0.0001005, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=176815, misc.num_pos=77, misc.num_neg=42054, misc.num_anchors=42240, misc.lr=0.0009353, misc.mem_usage=55.3\n", + "runtime.step=2050, runtime.steptime=0.2703, runtime.voxel_gene_time=0.001031, runtime.prep_time=0.03021, loss.cls_loss=0.1956, loss.cls_loss_rt=0.1574, loss.loc_loss=0.3641, loss.loc_loss_rt=0.3223, loss.loc_elem=[0.00723, 0.005262, 0.03166, 0.01628, 0.02993, 0.02787, 0.04293], loss.cls_pos_rt=0.1014, loss.cls_neg_rt=0.05591, loss.dir_rt=0.3164, rpn_acc=0.9992, pr.prec@10=0.1281, pr.rec@10=0.9176, pr.prec@30=0.669, pr.rec@30=0.6948, pr.prec@50=0.9311, pr.rec@50=0.4397, pr.prec@70=0.9919, pr.rec@70=0.1482, pr.prec@80=0.9986, pr.rec@80=0.02725, pr.prec@90=1.0, pr.rec@90=0.0001201, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=177296, misc.num_pos=72, misc.num_neg=42067, misc.num_anchors=42240, misc.lr=0.0009646, misc.mem_usage=55.3\n", + "runtime.step=2100, runtime.steptime=0.2689, runtime.voxel_gene_time=0.001313, runtime.prep_time=0.03622, loss.cls_loss=0.1957, loss.cls_loss_rt=0.1947, loss.loc_loss=0.363, loss.loc_loss_rt=0.345, loss.loc_elem=[0.009121, 0.005181, 0.03865, 0.01885, 0.02742, 0.03349, 0.03981], loss.cls_pos_rt=0.1354, loss.cls_neg_rt=0.05932, loss.dir_rt=0.3125, rpn_acc=0.9992, pr.prec@10=0.1283, pr.rec@10=0.9176, pr.prec@30=0.6691, pr.rec@30=0.695, pr.prec@50=0.9308, pr.rec@50=0.4404, pr.prec@70=0.992, pr.rec@70=0.1492, pr.prec@80=0.9985, pr.rec@80=0.02768, pr.prec@90=1.0, pr.rec@90=0.0001561, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=177761, misc.num_pos=60, misc.num_neg=42092, misc.num_anchors=42240, misc.lr=0.0009942, misc.mem_usage=55.3\n", + "reset Car\n", + "runtime.step=2150, runtime.steptime=0.2701, runtime.voxel_gene_time=0.001044, runtime.prep_time=0.03007, loss.cls_loss=0.1948, loss.cls_loss_rt=0.1917, loss.loc_loss=0.3623, loss.loc_loss_rt=0.3465, loss.loc_elem=[0.008718, 0.005702, 0.0434, 0.01688, 0.03408, 0.03064, 0.03383], loss.cls_pos_rt=0.1607, loss.cls_neg_rt=0.031, loss.dir_rt=0.219, rpn_acc=0.9992, pr.prec@10=0.1291, pr.rec@10=0.9176, pr.prec@30=0.6709, pr.rec@30=0.6963, pr.prec@50=0.9315, pr.rec@50=0.4435, pr.prec@70=0.9921, pr.rec@70=0.1517, pr.prec@80=0.9981, pr.rec@80=0.02811, pr.prec@90=1.0, pr.rec@90=0.0001445, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=177373, misc.num_pos=59, misc.num_neg=42084, misc.num_anchors=42240, misc.lr=0.001024, misc.mem_usage=55.3\n", + "reset Car\n", + "reset Car\n", + "WORKER 0 seed: 1592957692\n", + "WORKER 1 seed: 1592957693\n", + "WORKER 2 seed: 1592957694\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "runtime.step=2200, runtime.steptime=0.433, runtime.voxel_gene_time=0.001105, runtime.prep_time=0.02894, loss.cls_loss=0.1981, loss.cls_loss_rt=0.2023, loss.loc_loss=0.3627, loss.loc_loss_rt=0.3657, loss.loc_elem=[0.007221, 0.008121, 0.03344, 0.01686, 0.03593, 0.02956, 0.05171], loss.cls_pos_rt=0.1425, loss.cls_neg_rt=0.05986, loss.dir_rt=0.3671, rpn_acc=0.9992, pr.prec@10=0.1277, pr.rec@10=0.9131, pr.prec@30=0.6682, pr.rec@30=0.6914, pr.prec@50=0.9291, pr.rec@50=0.4544, pr.prec@70=0.9898, pr.rec@70=0.1642, pr.prec@80=0.9972, pr.rec@80=0.02898, pr.prec@90=0.0, pr.rec@90=0.0, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=177917, misc.num_pos=64, misc.num_neg=42084, misc.num_anchors=42240, misc.lr=0.001055, misc.mem_usage=55.3\n", + "runtime.step=2250, runtime.steptime=0.2708, runtime.voxel_gene_time=0.0009317, runtime.prep_time=0.02267, loss.cls_loss=0.195, loss.cls_loss_rt=0.2002, loss.loc_loss=0.3631, loss.loc_loss_rt=0.3191, loss.loc_elem=[0.007177, 0.005857, 0.03059, 0.01556, 0.02779, 0.02942, 0.04313], loss.cls_pos_rt=0.1277, loss.cls_neg_rt=0.0725, loss.dir_rt=0.2846, rpn_acc=0.9992, pr.prec@10=0.1293, pr.rec@10=0.9178, pr.prec@30=0.6719, pr.rec@30=0.6961, pr.prec@50=0.9308, pr.rec@50=0.4523, pr.prec@70=0.9899, pr.rec@70=0.1602, pr.prec@80=0.9976, pr.rec@80=0.02827, pr.prec@90=1.0, pr.rec@90=3.41e-05, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=178422, misc.num_pos=49, misc.num_neg=42117, misc.num_anchors=42240, misc.lr=0.001086, misc.mem_usage=55.3\n", + "runtime.step=2300, runtime.steptime=0.2703, runtime.voxel_gene_time=0.001297, runtime.prep_time=0.03248, loss.cls_loss=0.1939, loss.cls_loss_rt=0.2019, loss.loc_loss=0.3623, loss.loc_loss_rt=0.3635, loss.loc_elem=[0.00859, 0.005805, 0.03522, 0.02296, 0.03322, 0.02476, 0.05118], loss.cls_pos_rt=0.1473, loss.cls_neg_rt=0.05455, loss.dir_rt=0.3202, rpn_acc=0.9992, pr.prec@10=0.1308, pr.rec@10=0.9175, pr.prec@30=0.6725, pr.rec@30=0.6995, pr.prec@50=0.9311, pr.rec@50=0.4549, pr.prec@70=0.9905, pr.rec@70=0.1595, pr.prec@80=0.9977, pr.rec@80=0.02829, pr.prec@90=1.0, pr.rec@90=2.165e-05, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=179378, misc.num_pos=71, misc.num_neg=42034, misc.num_anchors=42240, misc.lr=0.001117, misc.mem_usage=55.3\n", + "runtime.step=2350, runtime.steptime=0.2664, runtime.voxel_gene_time=0.00153, runtime.prep_time=0.03911, loss.cls_loss=0.192, loss.cls_loss_rt=0.1709, loss.loc_loss=0.3582, loss.loc_loss_rt=0.3273, loss.loc_elem=[0.006742, 0.005512, 0.03658, 0.01623, 0.0334, 0.02691, 0.03828], loss.cls_pos_rt=0.1316, loss.cls_neg_rt=0.03932, loss.dir_rt=0.2562, rpn_acc=0.9992, pr.prec@10=0.1319, pr.rec@10=0.9192, pr.prec@30=0.6741, pr.rec@30=0.7033, pr.prec@50=0.9323, pr.rec@50=0.459, pr.prec@70=0.9906, pr.rec@70=0.1594, pr.prec@80=0.9983, pr.rec@80=0.02751, pr.prec@90=1.0, pr.rec@90=1.582e-05, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=174103, misc.num_pos=48, misc.num_neg=42113, misc.num_anchors=42240, misc.lr=0.001149, misc.mem_usage=55.4\n", + "runtime.step=2400, runtime.steptime=0.2689, runtime.voxel_gene_time=0.002083, runtime.prep_time=0.04194, loss.cls_loss=0.1907, loss.cls_loss_rt=0.1554, loss.loc_loss=0.3583, loss.loc_loss_rt=0.398, loss.loc_elem=[0.008965, 0.005779, 0.04861, 0.01759, 0.04166, 0.03066, 0.04573], loss.cls_pos_rt=0.09801, loss.cls_neg_rt=0.05742, loss.dir_rt=0.2724, rpn_acc=0.9992, pr.prec@10=0.1333, pr.rec@10=0.9197, pr.prec@30=0.6762, pr.rec@30=0.7054, pr.prec@50=0.9315, pr.rec@50=0.4621, pr.prec@70=0.9912, pr.rec@70=0.1643, pr.prec@80=0.9987, pr.rec@80=0.02972, pr.prec@90=1.0, pr.rec@90=3.747e-05, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=176197, misc.num_pos=57, misc.num_neg=42098, misc.num_anchors=42240, misc.lr=0.00118, misc.mem_usage=55.4\n", + "runtime.step=2450, runtime.steptime=0.2692, runtime.voxel_gene_time=0.0008781, runtime.prep_time=0.02798, loss.cls_loss=0.1899, loss.cls_loss_rt=0.1818, loss.loc_loss=0.3568, loss.loc_loss_rt=0.3312, loss.loc_elem=[0.007957, 0.005163, 0.03387, 0.01922, 0.03192, 0.0325, 0.03497], loss.cls_pos_rt=0.1303, loss.cls_neg_rt=0.05146, loss.dir_rt=0.4154, rpn_acc=0.9992, pr.prec@10=0.1338, pr.rec@10=0.9204, pr.prec@30=0.6767, pr.rec@30=0.7069, pr.prec@50=0.9315, pr.rec@50=0.465, pr.prec@70=0.991, pr.rec@70=0.1662, pr.prec@80=0.9983, pr.rec@80=0.02993, pr.prec@90=1.0, pr.rec@90=5.146e-05, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=175471, misc.num_pos=52, misc.num_neg=42095, misc.num_anchors=42240, misc.lr=0.001213, misc.mem_usage=55.4\n", + "reset Car\n", + "reset Car\n", + "reset Car\n", + "WORKER 0 seed: 1592957783\n", + "WORKER 1 seed: 1592957784\n", + "WORKER 2 seed: 1592957785\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "runtime.step=2500, runtime.steptime=0.4317, runtime.voxel_gene_time=0.001042, runtime.prep_time=0.02982, loss.cls_loss=0.1912, loss.cls_loss_rt=0.2492, loss.loc_loss=0.3511, loss.loc_loss_rt=0.4118, loss.loc_elem=[0.01004, 0.006513, 0.05658, 0.01626, 0.03447, 0.02839, 0.05363], loss.cls_pos_rt=0.1853, loss.cls_neg_rt=0.06387, loss.dir_rt=0.2705, rpn_acc=0.9992, pr.prec@10=0.1334, pr.rec@10=0.9184, pr.prec@30=0.6831, pr.rec@30=0.7064, pr.prec@50=0.9294, pr.rec@50=0.4621, pr.prec@70=0.9894, pr.rec@70=0.1612, pr.prec@80=1.0, pr.rec@80=0.02228, pr.prec@90=0.0, pr.rec@90=0.0, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=176156, misc.num_pos=63, misc.num_neg=42071, misc.num_anchors=42240, misc.lr=0.001245, misc.mem_usage=55.2\n", + "runtime.step=2550, runtime.steptime=0.2692, runtime.voxel_gene_time=0.0009229, runtime.prep_time=0.02748, loss.cls_loss=0.1876, loss.cls_loss_rt=0.1628, loss.loc_loss=0.3522, loss.loc_loss_rt=0.3309, loss.loc_elem=[0.006722, 0.005388, 0.04044, 0.01714, 0.03027, 0.03076, 0.03475], loss.cls_pos_rt=0.1204, loss.cls_neg_rt=0.04239, loss.dir_rt=0.2792, rpn_acc=0.9992, pr.prec@10=0.1382, pr.rec@10=0.9198, pr.prec@30=0.6861, pr.rec@30=0.7148, pr.prec@50=0.9302, pr.rec@50=0.4718, pr.prec@70=0.9915, pr.rec@70=0.1671, pr.prec@80=0.9992, pr.rec@80=0.02412, pr.prec@90=0.0, pr.rec@90=0.0, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=177473, misc.num_pos=61, misc.num_neg=42083, misc.num_anchors=42240, misc.lr=0.001278, misc.mem_usage=55.4\n", + "runtime.step=2600, runtime.steptime=0.2684, runtime.voxel_gene_time=0.0009787, runtime.prep_time=0.02665, loss.cls_loss=0.1865, loss.cls_loss_rt=0.1798, loss.loc_loss=0.3505, loss.loc_loss_rt=0.3494, loss.loc_elem=[0.008362, 0.005124, 0.03879, 0.02068, 0.03097, 0.02628, 0.04447], loss.cls_pos_rt=0.1435, loss.cls_neg_rt=0.03633, loss.dir_rt=0.2178, rpn_acc=0.9993, pr.prec@10=0.139, pr.rec@10=0.9207, pr.prec@30=0.6864, pr.rec@30=0.7165, pr.prec@50=0.931, pr.rec@50=0.4762, pr.prec@70=0.9926, pr.rec@70=0.1747, pr.prec@80=0.9992, pr.rec@80=0.02961, pr.prec@90=1.0, pr.rec@90=3.468e-05, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=177242, misc.num_pos=46, misc.num_neg=42134, misc.num_anchors=42240, misc.lr=0.001311, misc.mem_usage=55.4\n", + "runtime.step=2650, runtime.steptime=0.2671, runtime.voxel_gene_time=0.0009937, runtime.prep_time=0.02444, loss.cls_loss=0.1842, loss.cls_loss_rt=0.1622, loss.loc_loss=0.3468, loss.loc_loss_rt=0.305, loss.loc_elem=[0.00628, 0.005862, 0.03203, 0.01418, 0.02741, 0.02601, 0.04072], loss.cls_pos_rt=0.1289, loss.cls_neg_rt=0.03337, loss.dir_rt=0.2873, rpn_acc=0.9993, pr.prec@10=0.1399, pr.rec@10=0.9227, pr.prec@30=0.686, pr.rec@30=0.7195, pr.prec@50=0.9309, pr.rec@50=0.479, pr.prec@70=0.9922, pr.rec@70=0.1748, pr.prec@80=0.9976, pr.rec@80=0.02818, pr.prec@90=1.0, pr.rec@90=3.326e-05, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=176499, misc.num_pos=46, misc.num_neg=42135, misc.num_anchors=42240, misc.lr=0.001344, misc.mem_usage=55.4\n", + "runtime.step=2700, runtime.steptime=0.2675, runtime.voxel_gene_time=0.0008812, runtime.prep_time=0.02617, loss.cls_loss=0.1854, loss.cls_loss_rt=0.197, loss.loc_loss=0.35, loss.loc_loss_rt=0.3919, loss.loc_elem=[0.008232, 0.006012, 0.03768, 0.01692, 0.04443, 0.03634, 0.04634], loss.cls_pos_rt=0.1422, loss.cls_neg_rt=0.05482, loss.dir_rt=0.2785, rpn_acc=0.9993, pr.prec@10=0.1383, pr.rec@10=0.9216, pr.prec@30=0.6841, pr.rec@30=0.7171, pr.prec@50=0.9309, pr.rec@50=0.4765, pr.prec@70=0.9922, pr.rec@70=0.1741, pr.prec@80=0.9977, pr.rec@80=0.02818, pr.prec@90=1.0, pr.rec@90=2.597e-05, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=176888, misc.num_pos=50, misc.num_neg=42115, misc.num_anchors=42240, misc.lr=0.001378, misc.mem_usage=55.4\n", + "runtime.step=2750, runtime.steptime=0.2698, runtime.voxel_gene_time=0.001066, runtime.prep_time=0.03678, loss.cls_loss=0.1841, loss.cls_loss_rt=0.1689, loss.loc_loss=0.3491, loss.loc_loss_rt=0.3121, loss.loc_elem=[0.00702, 0.005732, 0.03269, 0.01605, 0.03309, 0.02803, 0.03342], loss.cls_pos_rt=0.11, loss.cls_neg_rt=0.05886, loss.dir_rt=0.3559, rpn_acc=0.9993, pr.prec@10=0.1395, pr.rec@10=0.923, pr.prec@30=0.685, pr.rec@30=0.7189, pr.prec@50=0.9317, pr.rec@50=0.4795, pr.prec@70=0.9919, pr.rec@70=0.1761, pr.prec@80=0.9978, pr.rec@80=0.0287, pr.prec@90=1.0, pr.rec@90=2.656e-05, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=176840, misc.num_pos=50, misc.num_neg=42105, misc.num_anchors=42240, misc.lr=0.001411, misc.mem_usage=55.4\n", + "reset Car\n", + "reset Car\n", + "reset Car\n", + "WORKER 0 seed: 1592957875\n", + "WORKER 1 seed: 1592957876\n", + "WORKER 2 seed: 1592957877\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "runtime.step=2800, runtime.steptime=0.4337, runtime.voxel_gene_time=0.001147, runtime.prep_time=0.02529, loss.cls_loss=0.1923, loss.cls_loss_rt=0.1952, loss.loc_loss=0.3578, loss.loc_loss_rt=0.3126, loss.loc_elem=[0.007225, 0.006668, 0.02848, 0.01778, 0.03122, 0.02584, 0.03909], loss.cls_pos_rt=0.1503, loss.cls_neg_rt=0.04482, loss.dir_rt=0.3532, rpn_acc=0.9993, pr.prec@10=0.1317, pr.rec@10=0.9126, pr.prec@30=0.6897, pr.rec@30=0.7032, pr.prec@50=0.9287, pr.rec@50=0.4751, pr.prec@70=0.9907, pr.rec@70=0.1917, pr.prec@80=0.9964, pr.rec@80=0.04322, pr.prec@90=1.0, pr.rec@90=0.0003915, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=176998, misc.num_pos=63, misc.num_neg=42084, misc.num_anchors=42240, misc.lr=0.001445, misc.mem_usage=55.3\n", + "runtime.step=2850, runtime.steptime=0.2699, runtime.voxel_gene_time=0.00152, runtime.prep_time=0.04512, loss.cls_loss=0.1877, loss.cls_loss_rt=0.1869, loss.loc_loss=0.3538, loss.loc_loss_rt=0.3289, loss.loc_elem=[0.00813, 0.005686, 0.03827, 0.01286, 0.03006, 0.02338, 0.04606], loss.cls_pos_rt=0.1356, loss.cls_neg_rt=0.05122, loss.dir_rt=0.1916, rpn_acc=0.9993, pr.prec@10=0.1348, pr.rec@10=0.9184, pr.prec@30=0.6859, pr.rec@30=0.7111, pr.prec@50=0.9337, pr.rec@50=0.4799, pr.prec@70=0.9931, pr.rec@70=0.1882, pr.prec@80=0.9977, pr.rec@80=0.0373, pr.prec@90=1.0, pr.rec@90=0.0001721, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=175246, misc.num_pos=61, misc.num_neg=42097, misc.num_anchors=42240, misc.lr=0.001479, misc.mem_usage=55.5\n", + "runtime.step=2900, runtime.steptime=0.2715, runtime.voxel_gene_time=0.00107, runtime.prep_time=0.03021, loss.cls_loss=0.1831, loss.cls_loss_rt=0.1748, loss.loc_loss=0.3471, loss.loc_loss_rt=0.3754, loss.loc_elem=[0.007409, 0.005618, 0.04983, 0.01681, 0.03764, 0.02878, 0.04162], loss.cls_pos_rt=0.1276, loss.cls_neg_rt=0.04718, loss.dir_rt=0.1982, rpn_acc=0.9993, pr.prec@10=0.1408, pr.rec@10=0.9214, pr.prec@30=0.6912, pr.rec@30=0.7204, pr.prec@50=0.9333, pr.rec@50=0.4892, pr.prec@70=0.9926, pr.rec@70=0.1988, pr.prec@80=0.9982, pr.rec@80=0.04139, pr.prec@90=1.0, pr.rec@90=0.000137, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=176973, misc.num_pos=44, misc.num_neg=42122, misc.num_anchors=42240, misc.lr=0.001513, misc.mem_usage=55.5\n", + "runtime.step=2950, runtime.steptime=0.2703, runtime.voxel_gene_time=0.001223, runtime.prep_time=0.03993, loss.cls_loss=0.1811, loss.cls_loss_rt=0.18, loss.loc_loss=0.3428, loss.loc_loss_rt=0.3477, loss.loc_elem=[0.006849, 0.006245, 0.03443, 0.0149, 0.03362, 0.02866, 0.04916], loss.cls_pos_rt=0.1475, loss.cls_neg_rt=0.03249, loss.dir_rt=0.2861, rpn_acc=0.9993, pr.prec@10=0.1423, pr.rec@10=0.9233, pr.prec@30=0.6903, pr.rec@30=0.7228, pr.prec@50=0.9331, pr.rec@50=0.4939, pr.prec@70=0.9926, pr.rec@70=0.2023, pr.prec@80=0.9979, pr.rec@80=0.04236, pr.prec@90=1.0, pr.rec@90=0.000123, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=180000, misc.num_pos=51, misc.num_neg=42116, misc.num_anchors=42240, misc.lr=0.001547, misc.mem_usage=55.4\n", + "runtime.step=3000, runtime.steptime=0.2718, runtime.voxel_gene_time=0.0008559, runtime.prep_time=0.02394, loss.cls_loss=0.1803, loss.cls_loss_rt=0.1815, loss.loc_loss=0.3403, loss.loc_loss_rt=0.3095, loss.loc_elem=[0.008506, 0.005567, 0.03376, 0.01561, 0.02997, 0.0261, 0.03526], loss.cls_pos_rt=0.1301, loss.cls_neg_rt=0.05134, loss.dir_rt=0.2484, rpn_acc=0.9993, pr.prec@10=0.1436, pr.rec@10=0.9232, pr.prec@30=0.6919, pr.rec@30=0.724, pr.prec@50=0.9339, pr.rec@50=0.4966, pr.prec@70=0.9925, pr.rec@70=0.2066, pr.prec@80=0.9983, pr.rec@80=0.04449, pr.prec@90=1.0, pr.rec@90=0.0001284, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=176348, misc.num_pos=48, misc.num_neg=42127, misc.num_anchors=42240, misc.lr=0.001581, misc.mem_usage=55.4\n", + "runtime.step=3050, runtime.steptime=0.2736, runtime.voxel_gene_time=0.001004, runtime.prep_time=0.03084, loss.cls_loss=0.1791, loss.cls_loss_rt=0.1592, loss.loc_loss=0.3376, loss.loc_loss_rt=0.2948, loss.loc_elem=[0.006409, 0.005124, 0.0371, 0.0113, 0.03045, 0.03016, 0.02685], loss.cls_pos_rt=0.1229, loss.cls_neg_rt=0.03631, loss.dir_rt=0.2517, rpn_acc=0.9993, pr.prec@10=0.1445, pr.rec@10=0.9246, pr.prec@30=0.693, pr.rec@30=0.7265, pr.prec@50=0.9338, pr.rec@50=0.4992, pr.prec@70=0.9921, pr.rec@70=0.2081, pr.prec@80=0.9985, pr.rec@80=0.04486, pr.prec@90=1.0, pr.rec@90=0.0001099, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=178529, misc.num_pos=62, misc.num_neg=42087, misc.num_anchors=42240, misc.lr=0.001615, misc.mem_usage=55.4\n", + "reset Car\n", + "reset Car\n", + "reset Car\n", + "WORKER 0 seed: 1592957967\n", + "WORKER 1 seed: 1592957968\n", + "WORKER 2 seed: 1592957969\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "runtime.step=3100, runtime.steptime=0.4324, runtime.voxel_gene_time=0.001258, runtime.prep_time=0.03455, loss.cls_loss=0.1901, loss.cls_loss_rt=0.1708, loss.loc_loss=0.3438, loss.loc_loss_rt=0.3323, loss.loc_elem=[0.007137, 0.005903, 0.0387, 0.01637, 0.03241, 0.02421, 0.04143], loss.cls_pos_rt=0.137, loss.cls_neg_rt=0.03372, loss.dir_rt=0.2463, rpn_acc=0.9993, pr.prec@10=0.1339, pr.rec@10=0.9147, pr.prec@30=0.6828, pr.rec@30=0.7096, pr.prec@50=0.938, pr.rec@50=0.4713, pr.prec@70=0.9982, pr.rec@70=0.1647, pr.prec@80=1.0, pr.rec@80=0.01847, pr.prec@90=0.0, pr.rec@90=0.0, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=176903, misc.num_pos=74, misc.num_neg=42069, misc.num_anchors=42240, misc.lr=0.001649, misc.mem_usage=55.3\n", + "#################################\n", + "# EVAL\n", + "#################################\n", + "Generate output labels...\n", + "[100.0%][===================>][12.19it/s][00:26>00:00] \n", + "generate label finished(140.37/s). start eval:\n", + "Evaluation official\n", + "Car AP(Average Precision)@0.70, 0.70, 0.70:\n", + "bbox AP:89.92, 79.34, 77.90\n", + "bev AP:89.27, 78.74, 77.44\n", + "3d AP:74.25, 63.31, 57.01\n", + "aos AP:2.28, 3.41, 4.42\n", + "Car AP(Average Precision)@0.70, 0.50, 0.50:\n", + "bbox AP:89.92, 79.34, 77.90\n", + "bev AP:90.57, 88.39, 86.97\n", + "3d AP:90.51, 87.84, 86.03\n", + "aos AP:2.28, 3.41, 4.42\n", + "\n", + "Evaluation coco\n", + "Car coco AP@0.50:0.05:0.95:\n", + "bbox AP:65.22, 59.27, 57.11\n", + "bev AP:62.36, 57.91, 55.79\n", + "3d AP:50.53, 45.97, 43.30\n", + "aos AP:1.51, 2.65, 3.44\n", + "\n", + "eval.kitti.official.Car.bbox@0.70=[89.92, 79.34, 77.9], eval.kitti.official.Car.bev@0.70=[89.27, 78.74, 77.44], eval.kitti.official.Car.3d@0.70=[74.25, 63.31, 57.01], eval.kitti.official.Car.aos=[2.275, 3.412, 4.415], eval.kitti.official.Car.bev@0.50=[90.57, 88.39, 86.97], eval.kitti.official.Car.3d@0.50=[90.51, 87.84, 86.03], eval.kitti.coco.Car.bbox=[65.22, 59.27, 57.11], eval.kitti.coco.Car.bev=[62.36, 57.91, 55.79], eval.kitti.coco.Car.3d=[50.53, 45.97, 43.3], eval.kitti.coco.Car.aos=[1.51, 2.645, 3.444]\n", + "runtime.step=3150, runtime.steptime=1.042, runtime.voxel_gene_time=0.001326, runtime.prep_time=0.03203, loss.cls_loss=0.1782, loss.cls_loss_rt=0.1758, loss.loc_loss=0.3328, loss.loc_loss_rt=0.3178, loss.loc_elem=[0.0067, 0.006065, 0.03895, 0.01781, 0.02771, 0.02362, 0.03804], loss.cls_pos_rt=0.1219, loss.cls_neg_rt=0.0539, loss.dir_rt=0.2813, rpn_acc=0.9993, pr.prec@10=0.1433, pr.rec@10=0.9229, pr.prec@30=0.6979, pr.rec@30=0.7283, pr.prec@50=0.9341, pr.rec@50=0.5048, pr.prec@70=0.9915, pr.rec@70=0.2126, pr.prec@80=0.9984, pr.rec@80=0.04506, pr.prec@90=1.0, pr.rec@90=2.47e-05, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=176723, misc.num_pos=50, misc.num_neg=42118, misc.num_anchors=42240, misc.lr=0.001684, misc.mem_usage=55.5\n", + "runtime.step=3200, runtime.steptime=0.2656, runtime.voxel_gene_time=0.0009992, runtime.prep_time=0.0229, loss.cls_loss=0.1779, loss.cls_loss_rt=0.2055, loss.loc_loss=0.3332, loss.loc_loss_rt=0.4185, loss.loc_elem=[0.01002, 0.006465, 0.04947, 0.01673, 0.03532, 0.02935, 0.06188], loss.cls_pos_rt=0.1412, loss.cls_neg_rt=0.06423, loss.dir_rt=0.2364, rpn_acc=0.9993, pr.prec@10=0.1461, pr.rec@10=0.9235, pr.prec@30=0.6972, pr.rec@30=0.7301, pr.prec@50=0.9313, pr.rec@50=0.5053, pr.prec@70=0.9918, pr.rec@70=0.2142, pr.prec@80=0.9992, pr.rec@80=0.0476, pr.prec@90=1.0, pr.rec@90=5.383e-05, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=175266, misc.num_pos=65, misc.num_neg=42060, misc.num_anchors=42240, misc.lr=0.001718, misc.mem_usage=55.5\n", + "runtime.step=3250, runtime.steptime=0.2658, runtime.voxel_gene_time=0.0009212, runtime.prep_time=0.02603, loss.cls_loss=0.177, loss.cls_loss_rt=0.1863, loss.loc_loss=0.3317, loss.loc_loss_rt=0.3698, loss.loc_elem=[0.008355, 0.005682, 0.05524, 0.01946, 0.03317, 0.02609, 0.03691], loss.cls_pos_rt=0.1536, loss.cls_neg_rt=0.03274, loss.dir_rt=0.2652, rpn_acc=0.9993, pr.prec@10=0.1471, pr.rec@10=0.924, pr.prec@30=0.6986, pr.rec@30=0.7301, pr.prec@50=0.9323, pr.rec@50=0.5078, pr.prec@70=0.9923, pr.rec@70=0.2141, pr.prec@80=0.9992, pr.rec@80=0.04661, pr.prec@90=1.0, pr.rec@90=8.336e-05, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=179267, misc.num_pos=61, misc.num_neg=42093, misc.num_anchors=42240, misc.lr=0.001752, misc.mem_usage=55.5\n", + "runtime.step=3300, runtime.steptime=0.265, runtime.voxel_gene_time=0.000828, runtime.prep_time=0.03062, loss.cls_loss=0.176, loss.cls_loss_rt=0.1675, loss.loc_loss=0.3287, loss.loc_loss_rt=0.285, loss.loc_elem=[0.006138, 0.005508, 0.03134, 0.01624, 0.03239, 0.02031, 0.03058], loss.cls_pos_rt=0.1224, loss.cls_neg_rt=0.04506, loss.dir_rt=0.3234, rpn_acc=0.9993, pr.prec@10=0.1485, pr.rec@10=0.9256, pr.prec@30=0.6974, pr.rec@30=0.7325, pr.prec@50=0.9325, pr.rec@50=0.5094, pr.prec@70=0.9911, pr.rec@70=0.2168, pr.prec@80=0.9985, pr.rec@80=0.048, pr.prec@90=1.0, pr.rec@90=8.456e-05, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=176828, misc.num_pos=41, misc.num_neg=42136, misc.num_anchors=42240, misc.lr=0.001786, misc.mem_usage=55.5\n", + "runtime.step=3350, runtime.steptime=0.266, runtime.voxel_gene_time=0.0009387, runtime.prep_time=0.0245, loss.cls_loss=0.1757, loss.cls_loss_rt=0.1439, loss.loc_loss=0.3279, loss.loc_loss_rt=0.2994, loss.loc_elem=[0.005756, 0.006461, 0.03053, 0.01707, 0.03189, 0.02953, 0.02849], loss.cls_pos_rt=0.1025, loss.cls_neg_rt=0.04139, loss.dir_rt=0.2024, rpn_acc=0.9993, pr.prec@10=0.1491, pr.rec@10=0.9258, pr.prec@30=0.6983, pr.rec@30=0.7326, pr.prec@50=0.9329, pr.rec@50=0.5093, pr.prec@70=0.9915, pr.rec@70=0.2179, pr.prec@80=0.9986, pr.rec@80=0.04931, pr.prec@90=1.0, pr.rec@90=0.0001251, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=175032, misc.num_pos=55, misc.num_neg=42104, misc.num_anchors=42240, misc.lr=0.00182, misc.mem_usage=55.5\n", + "reset Car\n", + "reset Car\n", + "reset Car\n", + "WORKER 0 seed: 1592958096\n", + "WORKER 1 seed: 1592958097\n", + "WORKER 2 seed: 1592958098\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "runtime.step=3400, runtime.steptime=0.4258, runtime.voxel_gene_time=0.01746, runtime.prep_time=7.203, loss.cls_loss=0.1788, loss.cls_loss_rt=0.1788, loss.loc_loss=0.3436, loss.loc_loss_rt=0.3436, loss.loc_elem=[0.008111, 0.005193, 0.03874, 0.01538, 0.03048, 0.0294, 0.04447], loss.cls_pos_rt=0.1306, loss.cls_neg_rt=0.04814, loss.dir_rt=0.2354, rpn_acc=0.9993, pr.prec@10=0.1455, pr.rec@10=0.9164, pr.prec@30=0.6653, pr.rec@30=0.7209, pr.prec@50=0.9441, pr.rec@50=0.4791, pr.prec@70=1.0, pr.rec@70=0.2119, pr.prec@80=1.0, pr.rec@80=0.0597, pr.prec@90=1.0, pr.rec@90=0.001493, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=179824, misc.num_pos=42, misc.num_neg=42139, misc.num_anchors=42240, misc.lr=0.001854, misc.mem_usage=54.8\n", + "runtime.step=3450, runtime.steptime=0.2695, runtime.voxel_gene_time=0.00123, runtime.prep_time=0.03811, loss.cls_loss=0.1797, loss.cls_loss_rt=0.1831, loss.loc_loss=0.3414, loss.loc_loss_rt=0.338, loss.loc_elem=[0.00667, 0.006312, 0.03952, 0.01384, 0.028, 0.03162, 0.04303], loss.cls_pos_rt=0.1454, loss.cls_neg_rt=0.03768, loss.dir_rt=0.277, rpn_acc=0.9993, pr.prec@10=0.1431, pr.rec@10=0.9192, pr.prec@30=0.698, pr.rec@30=0.7251, pr.prec@50=0.9306, pr.rec@50=0.5077, pr.prec@70=0.994, pr.rec@70=0.2287, pr.prec@80=0.9995, pr.rec@80=0.05701, pr.prec@90=1.0, pr.rec@90=0.0004078, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=179215, misc.num_pos=60, misc.num_neg=42091, misc.num_anchors=42240, misc.lr=0.001887, misc.mem_usage=55.4\n", + "runtime.step=3500, runtime.steptime=0.2698, runtime.voxel_gene_time=0.0008967, runtime.prep_time=0.02948, loss.cls_loss=0.1747, loss.cls_loss_rt=0.1532, loss.loc_loss=0.3337, loss.loc_loss_rt=0.3301, loss.loc_elem=[0.007047, 0.005389, 0.03449, 0.02035, 0.03124, 0.02945, 0.03707], loss.cls_pos_rt=0.09929, loss.cls_neg_rt=0.05395, loss.dir_rt=0.2268, rpn_acc=0.9993, pr.prec@10=0.1485, pr.rec@10=0.9259, pr.prec@30=0.7028, pr.rec@30=0.7333, pr.prec@50=0.9314, pr.rec@50=0.5126, pr.prec@70=0.9936, pr.rec@70=0.2255, pr.prec@80=0.9989, pr.rec@80=0.05416, pr.prec@90=1.0, pr.rec@90=0.0002196, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=175018, misc.num_pos=48, misc.num_neg=42108, misc.num_anchors=42240, misc.lr=0.001921, misc.mem_usage=55.4\n", + "runtime.step=3550, runtime.steptime=0.2665, runtime.voxel_gene_time=0.001444, runtime.prep_time=0.03448, loss.cls_loss=0.1749, loss.cls_loss_rt=0.204, loss.loc_loss=0.3308, loss.loc_loss_rt=0.3575, loss.loc_elem=[0.006974, 0.005688, 0.04702, 0.01708, 0.03203, 0.03215, 0.0378], loss.cls_pos_rt=0.1416, loss.cls_neg_rt=0.06236, loss.dir_rt=0.2945, rpn_acc=0.9993, pr.prec@10=0.1494, pr.rec@10=0.9258, pr.prec@30=0.7018, pr.rec@30=0.7342, pr.prec@50=0.9294, pr.rec@50=0.5132, pr.prec@70=0.9923, pr.rec@70=0.2238, pr.prec@80=0.9984, pr.rec@80=0.05395, pr.prec@90=1.0, pr.rec@90=0.0002743, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=176446, misc.num_pos=59, misc.num_neg=42087, misc.num_anchors=42240, misc.lr=0.001954, misc.mem_usage=55.4\n", + "runtime.step=3600, runtime.steptime=0.2669, runtime.voxel_gene_time=0.0008738, runtime.prep_time=0.02539, loss.cls_loss=0.1734, loss.cls_loss_rt=0.1779, loss.loc_loss=0.3285, loss.loc_loss_rt=0.338, loss.loc_elem=[0.007058, 0.005292, 0.04933, 0.02026, 0.02646, 0.02434, 0.03624], loss.cls_pos_rt=0.1313, loss.cls_neg_rt=0.04657, loss.dir_rt=0.2723, rpn_acc=0.9993, pr.prec@10=0.1511, pr.rec@10=0.9276, pr.prec@30=0.7024, pr.rec@30=0.7358, pr.prec@50=0.93, pr.rec@50=0.5148, pr.prec@70=0.9924, pr.rec@70=0.222, pr.prec@80=0.9986, pr.rec@80=0.05101, pr.prec@90=1.0, pr.rec@90=0.0002135, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=175280, misc.num_pos=47, misc.num_neg=42106, misc.num_anchors=42240, misc.lr=0.001988, misc.mem_usage=55.4\n", + "runtime.step=3650, runtime.steptime=0.2681, runtime.voxel_gene_time=0.0009317, runtime.prep_time=0.02857, loss.cls_loss=0.1729, loss.cls_loss_rt=0.1575, loss.loc_loss=0.3274, loss.loc_loss_rt=0.2788, loss.loc_elem=[0.005489, 0.006361, 0.02717, 0.01581, 0.02683, 0.02859, 0.02917], loss.cls_pos_rt=0.08743, loss.cls_neg_rt=0.07011, loss.dir_rt=0.2628, rpn_acc=0.9993, pr.prec@10=0.1511, pr.rec@10=0.928, pr.prec@30=0.7023, pr.rec@30=0.7368, pr.prec@50=0.9315, pr.rec@50=0.5161, pr.prec@70=0.9923, pr.rec@70=0.2223, pr.prec@80=0.9988, pr.rec@80=0.05119, pr.prec@90=1.0, pr.rec@90=0.0001769, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=174397, misc.num_pos=68, misc.num_neg=42078, misc.num_anchors=42240, misc.lr=0.002021, misc.mem_usage=55.5\n", + "reset Car\n", + "reset Car\n", + "reset Car\n", + "runtime.step=3700, runtime.steptime=0.2748, runtime.voxel_gene_time=0.0009058, runtime.prep_time=0.03506, loss.cls_loss=0.1717, loss.cls_loss_rt=0.186, loss.loc_loss=0.3254, loss.loc_loss_rt=0.3197, loss.loc_elem=[0.006732, 0.007209, 0.03341, 0.01562, 0.03974, 0.0252, 0.03193], loss.cls_pos_rt=0.162, loss.cls_neg_rt=0.02395, loss.dir_rt=0.2086, rpn_acc=0.9993, pr.prec@10=0.1532, pr.rec@10=0.9287, pr.prec@30=0.7046, pr.rec@30=0.7385, pr.prec@50=0.9321, pr.rec@50=0.5197, pr.prec@70=0.9924, pr.rec@70=0.226, pr.prec@80=0.9991, pr.rec@80=0.05305, pr.prec@90=1.0, pr.rec@90=0.0001768, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=176311, misc.num_pos=34, misc.num_neg=42148, misc.num_anchors=42240, misc.lr=0.002053, misc.mem_usage=55.8\n", + "WORKER 0 seed: 1592958187\n", + "WORKER 1 seed: 1592958188\n", + "WORKER 2 seed: 1592958189\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "runtime.step=3750, runtime.steptime=0.4369, runtime.voxel_gene_time=0.001283, runtime.prep_time=0.03355, loss.cls_loss=0.1722, loss.cls_loss_rt=0.1759, loss.loc_loss=0.323, loss.loc_loss_rt=0.2944, loss.loc_elem=[0.008138, 0.004767, 0.03159, 0.01551, 0.02961, 0.02671, 0.03087], loss.cls_pos_rt=0.127, loss.cls_neg_rt=0.04887, loss.dir_rt=0.2427, rpn_acc=0.9993, pr.prec@10=0.154, pr.rec@10=0.9265, pr.prec@30=0.7049, pr.rec@30=0.7363, pr.prec@50=0.9364, pr.rec@50=0.5208, pr.prec@70=0.9925, pr.rec@70=0.2288, pr.prec@80=0.9975, pr.rec@80=0.05716, pr.prec@90=1.0, pr.rec@90=0.0002821, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=180000, misc.num_pos=39, misc.num_neg=42135, misc.num_anchors=42240, misc.lr=0.002086, misc.mem_usage=55.4\n", + "runtime.step=3800, runtime.steptime=0.2815, runtime.voxel_gene_time=0.001029, runtime.prep_time=0.02926, loss.cls_loss=0.1703, loss.cls_loss_rt=0.1569, loss.loc_loss=0.321, loss.loc_loss_rt=0.2982, loss.loc_elem=[0.006238, 0.00657, 0.03571, 0.01477, 0.02735, 0.02445, 0.034], loss.cls_pos_rt=0.1209, loss.cls_neg_rt=0.03598, loss.dir_rt=0.2366, rpn_acc=0.9993, pr.prec@10=0.1563, pr.rec@10=0.9281, pr.prec@30=0.7046, pr.rec@30=0.7411, pr.prec@50=0.9342, pr.rec@50=0.5261, pr.prec@70=0.9929, pr.rec@70=0.2367, pr.prec@80=0.9987, pr.rec@80=0.05945, pr.prec@90=1.0, pr.rec@90=0.0002249, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=177873, misc.num_pos=40, misc.num_neg=42125, misc.num_anchors=42240, misc.lr=0.002118, misc.mem_usage=55.3\n", + "runtime.step=3850, runtime.steptime=0.2803, runtime.voxel_gene_time=0.0009961, runtime.prep_time=0.02383, loss.cls_loss=0.1695, loss.cls_loss_rt=0.1778, loss.loc_loss=0.3191, loss.loc_loss_rt=0.3718, loss.loc_elem=[0.008724, 0.005892, 0.07458, 0.01362, 0.02599, 0.0256, 0.03148], loss.cls_pos_rt=0.1368, loss.cls_neg_rt=0.04099, loss.dir_rt=0.2839, rpn_acc=0.9993, pr.prec@10=0.1566, pr.rec@10=0.9291, pr.prec@30=0.7063, pr.rec@30=0.7431, pr.prec@50=0.935, pr.rec@50=0.5293, pr.prec@70=0.9936, pr.rec@70=0.2375, pr.prec@80=0.998, pr.rec@80=0.05805, pr.prec@90=1.0, pr.rec@90=0.0001982, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=178231, misc.num_pos=53, misc.num_neg=42112, misc.num_anchors=42240, misc.lr=0.00215, misc.mem_usage=55.2\n", + "runtime.step=3900, runtime.steptime=0.2774, runtime.voxel_gene_time=0.0007584, runtime.prep_time=0.01938, loss.cls_loss=0.1683, loss.cls_loss_rt=0.1637, loss.loc_loss=0.3184, loss.loc_loss_rt=0.3079, loss.loc_elem=[0.006664, 0.006301, 0.0306, 0.01514, 0.03189, 0.02736, 0.03597], loss.cls_pos_rt=0.1118, loss.cls_neg_rt=0.05191, loss.dir_rt=0.2786, rpn_acc=0.9993, pr.prec@10=0.158, pr.rec@10=0.931, pr.prec@30=0.7056, pr.rec@30=0.7463, pr.prec@50=0.9339, pr.rec@50=0.5314, pr.prec@70=0.9926, pr.rec@70=0.2348, pr.prec@80=0.9979, pr.rec@80=0.05618, pr.prec@90=1.0, pr.rec@90=0.0002088, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=173800, misc.num_pos=55, misc.num_neg=42098, misc.num_anchors=42240, misc.lr=0.002182, misc.mem_usage=55.2\n", + "runtime.step=3950, runtime.steptime=0.2813, runtime.voxel_gene_time=0.0009904, runtime.prep_time=0.03716, loss.cls_loss=0.1692, loss.cls_loss_rt=0.1629, loss.loc_loss=0.3188, loss.loc_loss_rt=0.3625, loss.loc_elem=[0.007418, 0.006731, 0.04279, 0.01874, 0.03711, 0.03378, 0.03471], loss.cls_pos_rt=0.1231, loss.cls_neg_rt=0.03975, loss.dir_rt=0.2235, rpn_acc=0.9993, pr.prec@10=0.1564, pr.rec@10=0.9304, pr.prec@30=0.7041, pr.rec@30=0.7445, pr.prec@50=0.9341, pr.rec@50=0.5286, pr.prec@70=0.9924, pr.rec@70=0.2315, pr.prec@80=0.9981, pr.rec@80=0.05476, pr.prec@90=1.0, pr.rec@90=0.0002452, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=177070, misc.num_pos=57, misc.num_neg=42099, misc.num_anchors=42240, misc.lr=0.002213, misc.mem_usage=55.2\n", + "runtime.step=4000, runtime.steptime=0.2807, runtime.voxel_gene_time=0.001264, runtime.prep_time=0.03286, loss.cls_loss=0.1685, loss.cls_loss_rt=0.1564, loss.loc_loss=0.3173, loss.loc_loss_rt=0.3085, loss.loc_elem=[0.007586, 0.003653, 0.03974, 0.01638, 0.03051, 0.02182, 0.03455], loss.cls_pos_rt=0.09378, loss.cls_neg_rt=0.06265, loss.dir_rt=0.2015, rpn_acc=0.9993, pr.prec@10=0.1575, pr.rec@10=0.9307, pr.prec@30=0.7059, pr.rec@30=0.7461, pr.prec@50=0.9334, pr.rec@50=0.531, pr.prec@70=0.9918, pr.rec@70=0.236, pr.prec@80=0.9982, pr.rec@80=0.05715, pr.prec@90=1.0, pr.rec@90=0.0002633, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=178593, misc.num_pos=58, misc.num_neg=42095, misc.num_anchors=42240, misc.lr=0.002244, misc.mem_usage=55.2\n", + "reset Car\n", + "reset Car\n", + "reset Car\n", + "WORKER 0 seed: 1592958281\n", + "WORKER 1 seed: 1592958282\n", + "WORKER 2 seed: 1592958283\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "runtime.step=4050, runtime.steptime=0.4401, runtime.voxel_gene_time=0.0008376, runtime.prep_time=0.02842, loss.cls_loss=0.1676, loss.cls_loss_rt=0.1707, loss.loc_loss=0.3157, loss.loc_loss_rt=0.3109, loss.loc_elem=[0.008119, 0.006587, 0.03384, 0.01423, 0.03182, 0.03226, 0.02862], loss.cls_pos_rt=0.1354, loss.cls_neg_rt=0.03528, loss.dir_rt=0.235, rpn_acc=0.9993, pr.prec@10=0.1535, pr.rec@10=0.9348, pr.prec@30=0.6995, pr.rec@30=0.7393, pr.prec@50=0.9358, pr.rec@50=0.5204, pr.prec@70=0.9923, pr.rec@70=0.2381, pr.prec@80=1.0, pr.rec@80=0.06411, pr.prec@90=1.0, pr.rec@90=0.0004477, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=175931, misc.num_pos=56, misc.num_neg=42104, misc.num_anchors=42240, misc.lr=0.002274, misc.mem_usage=55.1\n", + "runtime.step=4100, runtime.steptime=0.282, runtime.voxel_gene_time=0.001568, runtime.prep_time=0.04262, loss.cls_loss=0.1682, loss.cls_loss_rt=0.1471, loss.loc_loss=0.3174, loss.loc_loss_rt=0.2979, loss.loc_elem=[0.007062, 0.005866, 0.03042, 0.01489, 0.03238, 0.02397, 0.03434], loss.cls_pos_rt=0.1116, loss.cls_neg_rt=0.03545, loss.dir_rt=0.208, rpn_acc=0.9993, pr.prec@10=0.1559, pr.rec@10=0.93, pr.prec@30=0.7077, pr.rec@30=0.7458, pr.prec@50=0.9361, pr.rec@50=0.5327, pr.prec@70=0.9933, pr.rec@70=0.2425, pr.prec@80=0.9992, pr.rec@80=0.0656, pr.prec@90=1.0, pr.rec@90=0.0005135, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=176320, misc.num_pos=69, misc.num_neg=42071, misc.num_anchors=42240, misc.lr=0.002305, misc.mem_usage=55.2\n", + "runtime.step=4150, runtime.steptime=0.28, runtime.voxel_gene_time=0.001068, runtime.prep_time=0.02683, loss.cls_loss=0.1666, loss.cls_loss_rt=0.1721, loss.loc_loss=0.3141, loss.loc_loss_rt=0.2922, loss.loc_elem=[0.004924, 0.00556, 0.03373, 0.01905, 0.02555, 0.02353, 0.03373], loss.cls_pos_rt=0.1283, loss.cls_neg_rt=0.04376, loss.dir_rt=0.2, rpn_acc=0.9993, pr.prec@10=0.158, pr.rec@10=0.9305, pr.prec@30=0.712, pr.rec@30=0.7483, pr.prec@50=0.9368, pr.rec@50=0.5386, pr.prec@70=0.9932, pr.rec@70=0.2489, pr.prec@80=0.9989, pr.rec@80=0.06756, pr.prec@90=1.0, pr.rec@90=0.0004215, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=177388, misc.num_pos=46, misc.num_neg=42130, misc.num_anchors=42240, misc.lr=0.002334, misc.mem_usage=55.3\n", + "runtime.step=4200, runtime.steptime=0.2793, runtime.voxel_gene_time=0.001004, runtime.prep_time=0.02442, loss.cls_loss=0.1658, loss.cls_loss_rt=0.1641, loss.loc_loss=0.311, loss.loc_loss_rt=0.2622, loss.loc_elem=[0.005972, 0.005657, 0.02631, 0.0132, 0.02454, 0.02819, 0.02725], loss.cls_pos_rt=0.1078, loss.cls_neg_rt=0.05626, loss.dir_rt=0.2136, rpn_acc=0.9993, pr.prec@10=0.1587, pr.rec@10=0.9316, pr.prec@30=0.7131, pr.rec@30=0.7498, pr.prec@50=0.9364, pr.rec@50=0.54, pr.prec@70=0.9929, pr.rec@70=0.2501, pr.prec@80=0.9992, pr.rec@80=0.06842, pr.prec@90=1.0, pr.rec@90=0.0004198, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=176834, misc.num_pos=47, misc.num_neg=42116, misc.num_anchors=42240, misc.lr=0.002364, misc.mem_usage=55.3\n", + "runtime.step=4250, runtime.steptime=0.281, runtime.voxel_gene_time=0.001121, runtime.prep_time=0.03486, loss.cls_loss=0.1665, loss.cls_loss_rt=0.162, loss.loc_loss=0.311, loss.loc_loss_rt=0.2812, loss.loc_elem=[0.00672, 0.005539, 0.0364, 0.01112, 0.02461, 0.02158, 0.03463], loss.cls_pos_rt=0.1121, loss.cls_neg_rt=0.04987, loss.dir_rt=0.2731, rpn_acc=0.9993, pr.prec@10=0.1584, pr.rec@10=0.931, pr.prec@30=0.7121, pr.rec@30=0.7485, pr.prec@50=0.9354, pr.rec@50=0.537, pr.prec@70=0.9928, pr.rec@70=0.248, pr.prec@80=0.9991, pr.rec@80=0.06696, pr.prec@90=1.0, pr.rec@90=0.000362, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=175275, misc.num_pos=70, misc.num_neg=42075, misc.num_anchors=42240, misc.lr=0.002392, misc.mem_usage=55.3\n", + "runtime.step=4300, runtime.steptime=0.2783, runtime.voxel_gene_time=0.0009692, runtime.prep_time=0.02476, loss.cls_loss=0.1652, loss.cls_loss_rt=0.1845, loss.loc_loss=0.3104, loss.loc_loss_rt=0.3066, loss.loc_elem=[0.007976, 0.006201, 0.03279, 0.01655, 0.03486, 0.02226, 0.03265], loss.cls_pos_rt=0.1466, loss.cls_neg_rt=0.03795, loss.dir_rt=0.2874, rpn_acc=0.9993, pr.prec@10=0.1595, pr.rec@10=0.9317, pr.prec@30=0.7132, pr.rec@30=0.7502, pr.prec@50=0.9359, pr.rec@50=0.5404, pr.prec@70=0.9929, pr.rec@70=0.2528, pr.prec@80=0.9993, pr.rec@80=0.07013, pr.prec@90=1.0, pr.rec@90=0.0004275, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=174934, misc.num_pos=57, misc.num_neg=42080, misc.num_anchors=42240, misc.lr=0.002421, misc.mem_usage=55.3\n", + "reset Car\n", + "reset Car\n", + "reset Car\n", + "WORKER 0 seed: 1592958376\n", + "WORKER 1 seed: 1592958377\n", + "WORKER 2 seed: 1592958378\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "runtime.step=4350, runtime.steptime=0.4417, runtime.voxel_gene_time=0.0009875, runtime.prep_time=0.02699, loss.cls_loss=0.169, loss.cls_loss_rt=0.1725, loss.loc_loss=0.3237, loss.loc_loss_rt=0.3347, loss.loc_elem=[0.007962, 0.005845, 0.03159, 0.02109, 0.03247, 0.02771, 0.04068], loss.cls_pos_rt=0.1402, loss.cls_neg_rt=0.03225, loss.dir_rt=0.2893, rpn_acc=0.9993, pr.prec@10=0.1582, pr.rec@10=0.9282, pr.prec@30=0.7093, pr.rec@30=0.7472, pr.prec@50=0.9352, pr.rec@50=0.5394, pr.prec@70=0.9912, pr.rec@70=0.2608, pr.prec@80=0.9962, pr.rec@80=0.08018, pr.prec@90=1.0, pr.rec@90=0.001521, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=179809, misc.num_pos=51, misc.num_neg=42116, misc.num_anchors=42240, misc.lr=0.002448, misc.mem_usage=55.0\n", + "runtime.step=4400, runtime.steptime=0.2824, runtime.voxel_gene_time=0.001012, runtime.prep_time=0.03699, loss.cls_loss=0.1648, loss.cls_loss_rt=0.1449, loss.loc_loss=0.3113, loss.loc_loss_rt=0.2844, loss.loc_elem=[0.00676, 0.004977, 0.03211, 0.01856, 0.02576, 0.02447, 0.02957], loss.cls_pos_rt=0.115, loss.cls_neg_rt=0.02989, loss.dir_rt=0.2259, rpn_acc=0.9993, pr.prec@10=0.1608, pr.rec@10=0.9315, pr.prec@30=0.7153, pr.rec@30=0.753, pr.prec@50=0.9381, pr.rec@50=0.544, pr.prec@70=0.9921, pr.rec@70=0.2597, pr.prec@80=0.9977, pr.rec@80=0.07798, pr.prec@90=1.0, pr.rec@90=0.001212, pr.prec@95=1.0, pr.rec@95=1.986e-05, misc.num_vox=178313, misc.num_pos=68, misc.num_neg=42075, misc.num_anchors=42240, misc.lr=0.002476, misc.mem_usage=55.2\n", + "runtime.step=4450, runtime.steptime=0.2798, runtime.voxel_gene_time=0.001076, runtime.prep_time=0.02956, loss.cls_loss=0.163, loss.cls_loss_rt=0.181, loss.loc_loss=0.3098, loss.loc_loss_rt=0.302, loss.loc_elem=[0.007687, 0.006251, 0.02382, 0.01504, 0.03186, 0.02657, 0.03978], loss.cls_pos_rt=0.1379, loss.cls_neg_rt=0.04313, loss.dir_rt=0.2512, rpn_acc=0.9993, pr.prec@10=0.1635, pr.rec@10=0.9323, pr.prec@30=0.7149, pr.rec@30=0.7565, pr.prec@50=0.9365, pr.rec@50=0.547, pr.prec@70=0.9918, pr.rec@70=0.2591, pr.prec@80=0.9977, pr.rec@80=0.07601, pr.prec@90=1.0, pr.rec@90=0.0008558, pr.prec@95=1.0, pr.rec@95=1.189e-05, misc.num_vox=174220, misc.num_pos=60, misc.num_neg=42086, misc.num_anchors=42240, misc.lr=0.002503, misc.mem_usage=55.2\n", + "runtime.step=4500, runtime.steptime=0.2803, runtime.voxel_gene_time=0.001531, runtime.prep_time=0.03537, loss.cls_loss=0.1633, loss.cls_loss_rt=0.1805, loss.loc_loss=0.3083, loss.loc_loss_rt=0.3071, loss.loc_elem=[0.007511, 0.006693, 0.03097, 0.01293, 0.03337, 0.02682, 0.03523], loss.cls_pos_rt=0.1288, loss.cls_neg_rt=0.05173, loss.dir_rt=0.2695, rpn_acc=0.9993, pr.prec@10=0.1622, pr.rec@10=0.9322, pr.prec@30=0.7176, pr.rec@30=0.7542, pr.prec@50=0.9368, pr.rec@50=0.5471, pr.prec@70=0.9917, pr.rec@70=0.2584, pr.prec@80=0.9982, pr.rec@80=0.07518, pr.prec@90=1.0, pr.rec@90=0.0008333, pr.prec@95=1.0, pr.rec@95=8.503e-06, misc.num_vox=179621, misc.num_pos=58, misc.num_neg=42100, misc.num_anchors=42240, misc.lr=0.002529, misc.mem_usage=55.2\n", + "runtime.step=4550, runtime.steptime=0.2821, runtime.voxel_gene_time=0.0009477, runtime.prep_time=0.02974, loss.cls_loss=0.163, loss.cls_loss_rt=0.1761, loss.loc_loss=0.3065, loss.loc_loss_rt=0.341, loss.loc_elem=[0.007065, 0.005899, 0.03863, 0.01281, 0.03573, 0.03247, 0.03788], loss.cls_pos_rt=0.1435, loss.cls_neg_rt=0.0326, loss.dir_rt=0.227, rpn_acc=0.9993, pr.prec@10=0.1631, pr.rec@10=0.9322, pr.prec@30=0.7167, pr.rec@30=0.7544, pr.prec@50=0.9365, pr.rec@50=0.5477, pr.prec@70=0.9917, pr.rec@70=0.2607, pr.prec@80=0.9982, pr.rec@80=0.07864, pr.prec@90=1.0, pr.rec@90=0.0009108, pr.prec@95=1.0, pr.rec@95=6.6e-06, misc.num_vox=178850, misc.num_pos=65, misc.num_neg=42083, misc.num_anchors=42240, misc.lr=0.002555, misc.mem_usage=55.2\n", + "runtime.step=4600, runtime.steptime=0.2795, runtime.voxel_gene_time=0.001484, runtime.prep_time=0.03165, loss.cls_loss=0.162, loss.cls_loss_rt=0.1611, loss.loc_loss=0.3053, loss.loc_loss_rt=0.2999, loss.loc_elem=[0.005195, 0.006222, 0.02755, 0.01696, 0.0316, 0.03217, 0.03025], loss.cls_pos_rt=0.13, loss.cls_neg_rt=0.03109, loss.dir_rt=0.251, rpn_acc=0.9993, pr.prec@10=0.1636, pr.rec@10=0.933, pr.prec@30=0.7187, pr.rec@30=0.7565, pr.prec@50=0.9362, pr.rec@50=0.55, pr.prec@70=0.9919, pr.rec@70=0.2611, pr.prec@80=0.9985, pr.rec@80=0.07742, pr.prec@90=1.0, pr.rec@90=0.0007935, pr.prec@95=1.0, pr.rec@95=5.398e-06, misc.num_vox=179146, misc.num_pos=51, misc.num_neg=42094, misc.num_anchors=42240, misc.lr=0.00258, misc.mem_usage=55.3\n", + "reset Car\n", + "reset Car\n", + "reset Car\n", + "WORKER 0 seed: 1592958471\n", + "WORKER 1 seed: 1592958472\n", + "WORKER 2 seed: 1592958473\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "runtime.step=4650, runtime.steptime=0.4416, runtime.voxel_gene_time=0.001017, runtime.prep_time=0.03379, loss.cls_loss=0.1628, loss.cls_loss_rt=0.1466, loss.loc_loss=0.2959, loss.loc_loss_rt=0.2972, loss.loc_elem=[0.006551, 0.005554, 0.0296, 0.01748, 0.02921, 0.02698, 0.03325], loss.cls_pos_rt=0.1083, loss.cls_neg_rt=0.03825, loss.dir_rt=0.224, rpn_acc=0.9993, pr.prec@10=0.1625, pr.rec@10=0.9311, pr.prec@30=0.722, pr.rec@30=0.7471, pr.prec@50=0.943, pr.rec@50=0.5416, pr.prec@70=0.995, pr.rec@70=0.2545, pr.prec@80=0.9973, pr.rec@80=0.07262, pr.prec@90=1.0, pr.rec@90=0.0003899, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=179387, misc.num_pos=56, misc.num_neg=42093, misc.num_anchors=42240, misc.lr=0.002604, misc.mem_usage=55.3\n", + "#################################\n", + "# EVAL\n", + "#################################\n", + "Generate output labels...\n", + "[100.0%][===================>][11.63it/s][00:27>00:00] \n", + "generate label finished(135.01/s). start eval:\n", + "Evaluation official\n", + "Car AP(Average Precision)@0.70, 0.70, 0.70:\n", + "bbox AP:85.42, 75.98, 73.47\n", + "bev AP:88.40, 78.07, 76.52\n", + "3d AP:53.54, 48.87, 46.22\n", + "aos AP:1.92, 2.49, 3.20\n", + "Car AP(Average Precision)@0.70, 0.50, 0.50:\n", + "bbox AP:85.42, 75.98, 73.47\n", + "bev AP:90.38, 88.12, 86.62\n", + "3d AP:90.00, 87.29, 84.43\n", + "aos AP:1.92, 2.49, 3.20\n", + "\n", + "Evaluation coco\n", + "Car coco AP@0.50:0.05:0.95:\n", + "bbox AP:56.96, 54.02, 52.78\n", + "bev AP:64.17, 58.66, 57.25\n", + "3d AP:43.80, 40.38, 39.30\n", + "aos AP:1.25, 1.84, 2.45\n", + "\n", + "eval.kitti.official.Car.bbox@0.70=[85.42, 75.98, 73.47], eval.kitti.official.Car.bev@0.70=[88.4, 78.07, 76.52], eval.kitti.official.Car.3d@0.70=[53.54, 48.87, 46.22], eval.kitti.official.Car.aos=[1.915, 2.485, 3.196], eval.kitti.official.Car.bev@0.50=[90.38, 88.12, 86.62], eval.kitti.official.Car.3d@0.50=[90.0, 87.29, 84.43], eval.kitti.coco.Car.bbox=[56.96, 54.02, 52.78], eval.kitti.coco.Car.bev=[64.17, 58.66, 57.25], eval.kitti.coco.Car.3d=[43.8, 40.38, 39.3], eval.kitti.coco.Car.aos=[1.253, 1.836, 2.453]\n", + "runtime.step=4700, runtime.steptime=1.094, runtime.voxel_gene_time=0.00102, runtime.prep_time=0.03708, loss.cls_loss=0.161, loss.cls_loss_rt=0.1739, loss.loc_loss=0.3058, loss.loc_loss_rt=0.3319, loss.loc_elem=[0.00635, 0.006746, 0.04688, 0.01361, 0.02959, 0.02666, 0.03613], loss.cls_pos_rt=0.136, loss.cls_neg_rt=0.03788, loss.dir_rt=0.2302, rpn_acc=0.9994, pr.prec@10=0.1621, pr.rec@10=0.9326, pr.prec@30=0.7211, pr.rec@30=0.7546, pr.prec@50=0.9418, pr.rec@50=0.5519, pr.prec@70=0.9931, pr.rec@70=0.2684, pr.prec@80=0.9995, pr.rec@80=0.08698, pr.prec@90=1.0, pr.rec@90=0.0009321, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=179177, misc.num_pos=69, misc.num_neg=42077, misc.num_anchors=42240, misc.lr=0.002628, misc.mem_usage=55.3\n", + "runtime.step=4750, runtime.steptime=0.2698, runtime.voxel_gene_time=0.0009651, runtime.prep_time=0.03292, loss.cls_loss=0.1601, loss.cls_loss_rt=0.1641, loss.loc_loss=0.3047, loss.loc_loss_rt=0.2939, loss.loc_elem=[0.006725, 0.005595, 0.02949, 0.01537, 0.02839, 0.02239, 0.03902], loss.cls_pos_rt=0.116, loss.cls_neg_rt=0.04809, loss.dir_rt=0.319, rpn_acc=0.9994, pr.prec@10=0.1651, pr.rec@10=0.9329, pr.prec@30=0.7233, pr.rec@30=0.7584, pr.prec@50=0.9394, pr.rec@50=0.5567, pr.prec@70=0.9926, pr.rec@70=0.2714, pr.prec@80=0.9991, pr.rec@80=0.0853, pr.prec@90=1.0, pr.rec@90=0.0007842, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=174073, misc.num_pos=44, misc.num_neg=42126, misc.num_anchors=42240, misc.lr=0.002651, misc.mem_usage=55.4\n", + "runtime.step=4800, runtime.steptime=0.2676, runtime.voxel_gene_time=0.001013, runtime.prep_time=0.02718, loss.cls_loss=0.1582, loss.cls_loss_rt=0.1776, loss.loc_loss=0.3021, loss.loc_loss_rt=0.3579, loss.loc_elem=[0.00691, 0.006099, 0.04177, 0.01691, 0.03019, 0.02613, 0.05093], loss.cls_pos_rt=0.1259, loss.cls_neg_rt=0.05167, loss.dir_rt=0.2154, rpn_acc=0.9994, pr.prec@10=0.1668, pr.rec@10=0.9351, pr.prec@30=0.7262, pr.rec@30=0.7619, pr.prec@50=0.9391, pr.rec@50=0.5604, pr.prec@70=0.992, pr.rec@70=0.2707, pr.prec@80=0.9989, pr.rec@80=0.08388, pr.prec@90=1.0, pr.rec@90=0.0007892, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=173707, misc.num_pos=47, misc.num_neg=42108, misc.num_anchors=42240, misc.lr=0.002674, misc.mem_usage=55.4\n", + "runtime.step=4850, runtime.steptime=0.2699, runtime.voxel_gene_time=0.0009573, runtime.prep_time=0.02944, loss.cls_loss=0.1583, loss.cls_loss_rt=0.1392, loss.loc_loss=0.3008, loss.loc_loss_rt=0.251, loss.loc_elem=[0.00536, 0.004947, 0.02743, 0.01244, 0.0244, 0.02365, 0.02728], loss.cls_pos_rt=0.08771, loss.cls_neg_rt=0.0515, loss.dir_rt=0.1903, rpn_acc=0.9994, pr.prec@10=0.1674, pr.rec@10=0.9356, pr.prec@30=0.7244, pr.rec@30=0.7615, pr.prec@50=0.9377, pr.rec@50=0.559, pr.prec@70=0.9917, pr.rec@70=0.2707, pr.prec@80=0.9991, pr.rec@80=0.08444, pr.prec@90=1.0, pr.rec@90=0.0007486, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=176105, misc.num_pos=48, misc.num_neg=42110, misc.num_anchors=42240, misc.lr=0.002696, misc.mem_usage=55.4\n", + "runtime.step=4900, runtime.steptime=0.266, runtime.voxel_gene_time=0.0009518, runtime.prep_time=0.02539, loss.cls_loss=0.1579, loss.cls_loss_rt=0.1666, loss.loc_loss=0.2997, loss.loc_loss_rt=0.3054, loss.loc_elem=[0.006543, 0.005131, 0.04599, 0.01506, 0.02599, 0.02326, 0.03073], loss.cls_pos_rt=0.116, loss.cls_neg_rt=0.05056, loss.dir_rt=0.2381, rpn_acc=0.9994, pr.prec@10=0.1685, pr.rec@10=0.9358, pr.prec@30=0.7249, pr.rec@30=0.7633, pr.prec@50=0.9381, pr.rec@50=0.5613, pr.prec@70=0.9918, pr.rec@70=0.2713, pr.prec@80=0.9986, pr.rec@80=0.08489, pr.prec@90=1.0, pr.rec@90=0.0007298, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=176753, misc.num_pos=56, misc.num_neg=42095, misc.num_anchors=42240, misc.lr=0.002717, misc.mem_usage=55.4\n", + "reset Car\n", + "reset Car\n", + "reset Car\n", + "WORKER 0 seed: 1592958603\n", + "WORKER 1 seed: 1592958604\n", + "WORKER 2 seed: 1592958605\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "runtime.step=4950, runtime.steptime=0.4301, runtime.voxel_gene_time=0.002113, runtime.prep_time=0.0403, loss.cls_loss=0.1601, loss.cls_loss_rt=0.1424, loss.loc_loss=0.3241, loss.loc_loss_rt=0.2901, loss.loc_elem=[0.006821, 0.004862, 0.03272, 0.01644, 0.02963, 0.02379, 0.03078], loss.cls_pos_rt=0.1003, loss.cls_neg_rt=0.04212, loss.dir_rt=0.2312, rpn_acc=0.9994, pr.prec@10=0.1674, pr.rec@10=0.9265, pr.prec@30=0.7271, pr.rec@30=0.7597, pr.prec@50=0.9421, pr.rec@50=0.5541, pr.prec@70=0.9982, pr.rec@70=0.2673, pr.prec@80=1.0, pr.rec@80=0.08329, pr.prec@90=0.0, pr.rec@90=0.0, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=177489, misc.num_pos=58, misc.num_neg=42081, misc.num_anchors=42240, misc.lr=0.002738, misc.mem_usage=55.1\n", + "runtime.step=5000, runtime.steptime=0.2689, runtime.voxel_gene_time=0.001454, runtime.prep_time=0.03854, loss.cls_loss=0.1584, loss.cls_loss_rt=0.1932, loss.loc_loss=0.2986, loss.loc_loss_rt=0.2814, loss.loc_elem=[0.007073, 0.005995, 0.02681, 0.01842, 0.02859, 0.02183, 0.03197], loss.cls_pos_rt=0.1303, loss.cls_neg_rt=0.06288, loss.dir_rt=0.1976, rpn_acc=0.9994, pr.prec@10=0.1718, pr.rec@10=0.9334, pr.prec@30=0.7252, pr.rec@30=0.7637, pr.prec@50=0.9356, pr.rec@50=0.5548, pr.prec@70=0.9923, pr.rec@70=0.2742, pr.prec@80=0.9991, pr.rec@80=0.08904, pr.prec@90=1.0, pr.rec@90=0.0005025, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=177166, misc.num_pos=33, misc.num_neg=42165, misc.num_anchors=42240, misc.lr=0.002758, misc.mem_usage=55.3\n", + "runtime.step=5050, runtime.steptime=0.268, runtime.voxel_gene_time=0.0008159, runtime.prep_time=0.02865, loss.cls_loss=0.157, loss.cls_loss_rt=0.1902, loss.loc_loss=0.2956, loss.loc_loss_rt=0.3643, loss.loc_elem=[0.009441, 0.005478, 0.0335, 0.01836, 0.02942, 0.02698, 0.05894], loss.cls_pos_rt=0.1303, loss.cls_neg_rt=0.05989, loss.dir_rt=0.2133, rpn_acc=0.9994, pr.prec@10=0.1721, pr.rec@10=0.9349, pr.prec@30=0.7271, pr.rec@30=0.7665, pr.prec@50=0.9356, pr.rec@50=0.5598, pr.prec@70=0.9932, pr.rec@70=0.2741, pr.prec@80=0.9989, pr.rec@80=0.08807, pr.prec@90=1.0, pr.rec@90=0.0008749, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=178293, misc.num_pos=68, misc.num_neg=42066, misc.num_anchors=42240, misc.lr=0.002777, misc.mem_usage=55.4\n", + "runtime.step=5100, runtime.steptime=0.2663, runtime.voxel_gene_time=0.0009854, runtime.prep_time=0.02877, loss.cls_loss=0.1579, loss.cls_loss_rt=0.122, loss.loc_loss=0.2972, loss.loc_loss_rt=0.2433, loss.loc_elem=[0.004524, 0.005176, 0.02405, 0.01274, 0.0245, 0.02563, 0.02502], loss.cls_pos_rt=0.07601, loss.cls_neg_rt=0.046, loss.dir_rt=0.2044, rpn_acc=0.9994, pr.prec@10=0.1709, pr.rec@10=0.9347, pr.prec@30=0.7255, pr.rec@30=0.7656, pr.prec@50=0.9359, pr.rec@50=0.5601, pr.prec@70=0.992, pr.rec@70=0.2721, pr.prec@80=0.999, pr.rec@80=0.08516, pr.prec@90=1.0, pr.rec@90=0.0008124, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=172997, misc.num_pos=57, misc.num_neg=42091, misc.num_anchors=42240, misc.lr=0.002795, misc.mem_usage=55.4\n", + "runtime.step=5150, runtime.steptime=0.2668, runtime.voxel_gene_time=0.0009236, runtime.prep_time=0.02818, loss.cls_loss=0.1572, loss.cls_loss_rt=0.1321, loss.loc_loss=0.2974, loss.loc_loss_rt=0.2814, loss.loc_elem=[0.005974, 0.004364, 0.03398, 0.01357, 0.03087, 0.0227, 0.02922], loss.cls_pos_rt=0.09543, loss.cls_neg_rt=0.03666, loss.dir_rt=0.2266, rpn_acc=0.9994, pr.prec@10=0.1717, pr.rec@10=0.9354, pr.prec@30=0.7263, pr.rec@30=0.7662, pr.prec@50=0.9358, pr.rec@50=0.5619, pr.prec@70=0.9925, pr.rec@70=0.2724, pr.prec@80=0.999, pr.rec@80=0.08502, pr.prec@90=1.0, pr.rec@90=0.0007813, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=174248, misc.num_pos=34, misc.num_neg=42151, misc.num_anchors=42240, misc.lr=0.002813, misc.mem_usage=55.4\n", + "runtime.step=5200, runtime.steptime=0.268, runtime.voxel_gene_time=0.001402, runtime.prep_time=0.03264, loss.cls_loss=0.1562, loss.cls_loss_rt=0.1498, loss.loc_loss=0.2958, loss.loc_loss_rt=0.2688, loss.loc_elem=[0.004994, 0.004975, 0.02291, 0.01646, 0.02217, 0.02678, 0.0361], loss.cls_pos_rt=0.1109, loss.cls_neg_rt=0.03895, loss.dir_rt=0.1823, rpn_acc=0.9994, pr.prec@10=0.1727, pr.rec@10=0.936, pr.prec@30=0.7273, pr.rec@30=0.7678, pr.prec@50=0.9362, pr.rec@50=0.5637, pr.prec@70=0.9924, pr.rec@70=0.2759, pr.prec@80=0.9989, pr.rec@80=0.08862, pr.prec@90=1.0, pr.rec@90=0.0007386, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=179579, misc.num_pos=53, misc.num_neg=42108, misc.num_anchors=42240, misc.lr=0.00283, misc.mem_usage=55.4\n", + "reset Car\n", + "reset Car\n", + "reset Car\n", + "runtime.step=5250, runtime.steptime=0.2681, runtime.voxel_gene_time=0.0008457, runtime.prep_time=0.02598, loss.cls_loss=0.1548, loss.cls_loss_rt=0.1043, loss.loc_loss=0.2942, loss.loc_loss_rt=0.2645, loss.loc_elem=[0.004715, 0.004711, 0.02939, 0.01426, 0.02333, 0.02346, 0.0324], loss.cls_pos_rt=0.07778, loss.cls_neg_rt=0.02657, loss.dir_rt=0.1764, rpn_acc=0.9994, pr.prec@10=0.1741, pr.rec@10=0.9372, pr.prec@30=0.7284, pr.rec@30=0.7696, pr.prec@50=0.9369, pr.rec@50=0.5667, pr.prec@70=0.9923, pr.rec@70=0.2792, pr.prec@80=0.9988, pr.rec@80=0.09093, pr.prec@90=1.0, pr.rec@90=0.0007235, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=177469, misc.num_pos=47, misc.num_neg=42119, misc.num_anchors=42240, misc.lr=0.002846, misc.mem_usage=55.4\n", + "WORKER 0 seed: 1592958694\n", + "WORKER 1 seed: 1592958695\n", + "WORKER 2 seed: 1592958696\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "runtime.step=5300, runtime.steptime=0.4269, runtime.voxel_gene_time=0.0008807, runtime.prep_time=0.0263, loss.cls_loss=0.1595, loss.cls_loss_rt=0.1429, loss.loc_loss=0.2925, loss.loc_loss_rt=0.2662, loss.loc_elem=[0.006125, 0.005112, 0.02302, 0.01742, 0.02813, 0.02142, 0.03188], loss.cls_pos_rt=0.1138, loss.cls_neg_rt=0.02906, loss.dir_rt=0.1983, rpn_acc=0.9994, pr.prec@10=0.1671, pr.rec@10=0.9341, pr.prec@30=0.7205, pr.rec@30=0.7555, pr.prec@50=0.9398, pr.rec@50=0.5555, pr.prec@70=0.9931, pr.rec@70=0.2867, pr.prec@80=0.9988, pr.rec@80=0.1057, pr.prec@90=1.0, pr.rec@90=0.001237, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=179668, misc.num_pos=58, misc.num_neg=42066, misc.num_anchors=42240, misc.lr=0.002862, misc.mem_usage=55.2\n", + "runtime.step=5350, runtime.steptime=0.2653, runtime.voxel_gene_time=0.0009494, runtime.prep_time=0.02395, loss.cls_loss=0.1558, loss.cls_loss_rt=0.1625, loss.loc_loss=0.2899, loss.loc_loss_rt=0.3294, loss.loc_elem=[0.009363, 0.005284, 0.04793, 0.01811, 0.03269, 0.01991, 0.03144], loss.cls_pos_rt=0.1255, loss.cls_neg_rt=0.03701, loss.dir_rt=0.1434, rpn_acc=0.9994, pr.prec@10=0.1711, pr.rec@10=0.9362, pr.prec@30=0.7275, pr.rec@30=0.7638, pr.prec@50=0.9403, pr.rec@50=0.5648, pr.prec@70=0.9926, pr.rec@70=0.29, pr.prec@80=0.9991, pr.rec@80=0.1056, pr.prec@90=1.0, pr.rec@90=0.001227, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=178796, misc.num_pos=63, misc.num_neg=42093, misc.num_anchors=42240, misc.lr=0.002876, misc.mem_usage=55.4\n", + "runtime.step=5400, runtime.steptime=0.2644, runtime.voxel_gene_time=0.001033, runtime.prep_time=0.02772, loss.cls_loss=0.1562, loss.cls_loss_rt=0.1562, loss.loc_loss=0.2905, loss.loc_loss_rt=0.2688, loss.loc_elem=[0.005451, 0.005277, 0.03064, 0.01389, 0.02731, 0.01688, 0.03495], loss.cls_pos_rt=0.1196, loss.cls_neg_rt=0.03659, loss.dir_rt=0.1619, rpn_acc=0.9994, pr.prec@10=0.1705, pr.rec@10=0.9365, pr.prec@30=0.728, pr.rec@30=0.7655, pr.prec@50=0.9382, pr.rec@50=0.564, pr.prec@70=0.992, pr.rec@70=0.289, pr.prec@80=0.9989, pr.rec@80=0.1028, pr.prec@90=1.0, pr.rec@90=0.001202, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=176066, misc.num_pos=64, misc.num_neg=42067, misc.num_anchors=42240, misc.lr=0.00289, misc.mem_usage=55.4\n", + "runtime.step=5450, runtime.steptime=0.2639, runtime.voxel_gene_time=0.0009305, runtime.prep_time=0.02701, loss.cls_loss=0.1551, loss.cls_loss_rt=0.1453, loss.loc_loss=0.2883, loss.loc_loss_rt=0.249, loss.loc_elem=[0.006776, 0.006163, 0.02436, 0.01596, 0.02596, 0.02435, 0.02094], loss.cls_pos_rt=0.09857, loss.cls_neg_rt=0.0467, loss.dir_rt=0.2384, rpn_acc=0.9994, pr.prec@10=0.1725, pr.rec@10=0.9375, pr.prec@30=0.7293, pr.rec@30=0.7679, pr.prec@50=0.9376, pr.rec@50=0.5675, pr.prec@70=0.9919, pr.rec@70=0.2898, pr.prec@80=0.9988, pr.rec@80=0.1003, pr.prec@90=1.0, pr.rec@90=0.001032, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=174996, misc.num_pos=51, misc.num_neg=42107, misc.num_anchors=42240, misc.lr=0.002903, misc.mem_usage=55.4\n", + "runtime.step=5500, runtime.steptime=0.2635, runtime.voxel_gene_time=0.0008821, runtime.prep_time=0.02744, loss.cls_loss=0.154, loss.cls_loss_rt=0.1285, loss.loc_loss=0.2875, loss.loc_loss_rt=0.2427, loss.loc_elem=[0.005945, 0.00385, 0.02528, 0.01593, 0.02372, 0.02449, 0.02215], loss.cls_pos_rt=0.0829, loss.cls_neg_rt=0.0456, loss.dir_rt=0.2219, rpn_acc=0.9994, pr.prec@10=0.1735, pr.rec@10=0.9384, pr.prec@30=0.7292, pr.rec@30=0.7696, pr.prec@50=0.9376, pr.rec@50=0.5684, pr.prec@70=0.9919, pr.rec@70=0.2896, pr.prec@80=0.9987, pr.rec@80=0.09971, pr.prec@90=1.0, pr.rec@90=0.0009666, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=169981, misc.num_pos=75, misc.num_neg=42053, misc.num_anchors=42240, misc.lr=0.002916, misc.mem_usage=55.4\n", + "reset Car\n", + "reset Car\n", + "runtime.step=5550, runtime.steptime=0.2644, runtime.voxel_gene_time=0.001037, runtime.prep_time=0.0326, loss.cls_loss=0.1537, loss.cls_loss_rt=0.1412, loss.loc_loss=0.2882, loss.loc_loss_rt=0.2901, loss.loc_elem=[0.006298, 0.005945, 0.03199, 0.01521, 0.02838, 0.0262, 0.03101], loss.cls_pos_rt=0.1031, loss.cls_neg_rt=0.03809, loss.dir_rt=0.1756, rpn_acc=0.9994, pr.prec@10=0.1741, pr.rec@10=0.938, pr.prec@30=0.7292, pr.rec@30=0.7701, pr.prec@50=0.9379, pr.rec@50=0.5708, pr.prec@70=0.992, pr.rec@70=0.2949, pr.prec@80=0.9988, pr.rec@80=0.1041, pr.prec@90=1.0, pr.rec@90=0.001156, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=178355, misc.num_pos=54, misc.num_neg=42094, misc.num_anchors=42240, misc.lr=0.002927, misc.mem_usage=55.4\n", + "reset Car\n", + "WORKER 0 seed: 1592958784\n", + "WORKER 1 seed: 1592958785\n", + "WORKER 2 seed: 1592958786\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "runtime.step=5600, runtime.steptime=0.4271, runtime.voxel_gene_time=0.0008962, runtime.prep_time=0.02228, loss.cls_loss=0.1558, loss.cls_loss_rt=0.1545, loss.loc_loss=0.2859, loss.loc_loss_rt=0.3024, loss.loc_elem=[0.006894, 0.005138, 0.03879, 0.01537, 0.03028, 0.02567, 0.02906], loss.cls_pos_rt=0.1153, loss.cls_neg_rt=0.03919, loss.dir_rt=0.1959, rpn_acc=0.9994, pr.prec@10=0.171, pr.rec@10=0.9348, pr.prec@30=0.7253, pr.rec@30=0.7675, pr.prec@50=0.9354, pr.rec@50=0.5656, pr.prec@70=0.9904, pr.rec@70=0.2857, pr.prec@80=0.9984, pr.rec@80=0.09968, pr.prec@90=1.0, pr.rec@90=0.001049, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=172645, misc.num_pos=57, misc.num_neg=42100, misc.num_anchors=42240, misc.lr=0.002938, misc.mem_usage=55.4\n", + "runtime.step=5650, runtime.steptime=0.2659, runtime.voxel_gene_time=0.001528, runtime.prep_time=0.03843, loss.cls_loss=0.1534, loss.cls_loss_rt=0.1822, loss.loc_loss=0.285, loss.loc_loss_rt=0.2978, loss.loc_elem=[0.007028, 0.005551, 0.03429, 0.01593, 0.02764, 0.02346, 0.03501], loss.cls_pos_rt=0.1423, loss.cls_neg_rt=0.03988, loss.dir_rt=0.2267, rpn_acc=0.9994, pr.prec@10=0.1748, pr.rec@10=0.9366, pr.prec@30=0.7291, pr.rec@30=0.7718, pr.prec@50=0.937, pr.rec@50=0.5695, pr.prec@70=0.9916, pr.rec@70=0.2891, pr.prec@80=0.9988, pr.rec@80=0.102, pr.prec@90=1.0, pr.rec@90=0.0008411, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=174898, misc.num_pos=69, misc.num_neg=42065, misc.num_anchors=42240, misc.lr=0.002948, misc.mem_usage=55.4\n", + "runtime.step=5700, runtime.steptime=0.2666, runtime.voxel_gene_time=0.001201, runtime.prep_time=0.02988, loss.cls_loss=0.1526, loss.cls_loss_rt=0.1436, loss.loc_loss=0.2855, loss.loc_loss_rt=0.2917, loss.loc_elem=[0.007381, 0.005174, 0.02308, 0.01621, 0.02762, 0.02895, 0.03745], loss.cls_pos_rt=0.1121, loss.cls_neg_rt=0.03151, loss.dir_rt=0.2396, rpn_acc=0.9994, pr.prec@10=0.1758, pr.rec@10=0.9374, pr.prec@30=0.7331, pr.rec@30=0.7726, pr.prec@50=0.9381, pr.rec@50=0.5739, pr.prec@70=0.9919, pr.rec@70=0.2973, pr.prec@80=0.999, pr.rec@80=0.1105, pr.prec@90=1.0, pr.rec@90=0.001386, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=178190, misc.num_pos=65, misc.num_neg=42076, misc.num_anchors=42240, misc.lr=0.002957, misc.mem_usage=55.4\n", + "runtime.step=5750, runtime.steptime=0.2662, runtime.voxel_gene_time=0.001392, runtime.prep_time=0.03567, loss.cls_loss=0.151, loss.cls_loss_rt=0.1403, loss.loc_loss=0.2823, loss.loc_loss_rt=0.2553, loss.loc_elem=[0.006155, 0.005133, 0.02666, 0.0151, 0.0229, 0.02292, 0.02879], loss.cls_pos_rt=0.1042, loss.cls_neg_rt=0.03609, loss.dir_rt=0.2282, rpn_acc=0.9994, pr.prec@10=0.1783, pr.rec@10=0.939, pr.prec@30=0.7352, pr.rec@30=0.7754, pr.prec@50=0.9388, pr.rec@50=0.5783, pr.prec@70=0.9919, pr.rec@70=0.3019, pr.prec@80=0.9993, pr.rec@80=0.1132, pr.prec@90=1.0, pr.rec@90=0.001543, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=174869, misc.num_pos=59, misc.num_neg=42094, misc.num_anchors=42240, misc.lr=0.002965, misc.mem_usage=55.4\n", + "runtime.step=5800, runtime.steptime=0.2658, runtime.voxel_gene_time=0.0009131, runtime.prep_time=0.02628, loss.cls_loss=0.1509, loss.cls_loss_rt=0.1528, loss.loc_loss=0.2828, loss.loc_loss_rt=0.2487, loss.loc_elem=[0.006761, 0.005314, 0.02552, 0.01167, 0.02526, 0.02221, 0.02763], loss.cls_pos_rt=0.1063, loss.cls_neg_rt=0.04656, loss.dir_rt=0.198, rpn_acc=0.9994, pr.prec@10=0.1787, pr.rec@10=0.9388, pr.prec@30=0.7356, pr.rec@30=0.7767, pr.prec@50=0.939, pr.rec@50=0.5786, pr.prec@70=0.9918, pr.rec@70=0.302, pr.prec@80=0.9991, pr.rec@80=0.1141, pr.prec@90=1.0, pr.rec@90=0.001683, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=176311, misc.num_pos=72, misc.num_neg=42073, misc.num_anchors=42240, misc.lr=0.002972, misc.mem_usage=55.4\n", + "runtime.step=5850, runtime.steptime=0.2659, runtime.voxel_gene_time=0.001042, runtime.prep_time=0.03261, loss.cls_loss=0.1501, loss.cls_loss_rt=0.1449, loss.loc_loss=0.2806, loss.loc_loss_rt=0.2692, loss.loc_elem=[0.005708, 0.004159, 0.03133, 0.0178, 0.02551, 0.0248, 0.02528], loss.cls_pos_rt=0.1104, loss.cls_neg_rt=0.03452, loss.dir_rt=0.1702, rpn_acc=0.9994, pr.prec@10=0.1788, pr.rec@10=0.9395, pr.prec@30=0.7363, pr.rec@30=0.7776, pr.prec@50=0.9393, pr.rec@50=0.5801, pr.prec@70=0.9922, pr.rec@70=0.3047, pr.prec@80=0.999, pr.rec@80=0.116, pr.prec@90=1.0, pr.rec@90=0.001802, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=176536, misc.num_pos=48, misc.num_neg=42118, misc.num_anchors=42240, misc.lr=0.002979, misc.mem_usage=55.4\n", + "reset Car\n", + "reset Car\n", + "reset Car\n", + "WORKER 0 seed: 1592958874\n", + "WORKER 1 seed: 1592958875\n", + "WORKER 2 seed: 1592958876\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "runtime.step=5900, runtime.steptime=0.4296, runtime.voxel_gene_time=0.001031, runtime.prep_time=0.0284, loss.cls_loss=0.1548, loss.cls_loss_rt=0.1509, loss.loc_loss=0.2822, loss.loc_loss_rt=0.2916, loss.loc_elem=[0.007229, 0.004251, 0.03352, 0.01494, 0.02473, 0.02474, 0.03641], loss.cls_pos_rt=0.1053, loss.cls_neg_rt=0.04567, loss.dir_rt=0.1442, rpn_acc=0.9994, pr.prec@10=0.1723, pr.rec@10=0.938, pr.prec@30=0.7276, pr.rec@30=0.7633, pr.prec@50=0.9414, pr.rec@50=0.5637, pr.prec@70=0.9954, pr.rec@70=0.3014, pr.prec@80=0.9996, pr.rec@80=0.1195, pr.prec@90=1.0, pr.rec@90=0.002258, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=176804, misc.num_pos=57, misc.num_neg=42101, misc.num_anchors=42240, misc.lr=0.002984, misc.mem_usage=55.3\n", + "runtime.step=5950, runtime.steptime=0.2691, runtime.voxel_gene_time=0.001054, runtime.prep_time=0.0298, loss.cls_loss=0.1519, loss.cls_loss_rt=0.1417, loss.loc_loss=0.2766, loss.loc_loss_rt=0.2722, loss.loc_elem=[0.00596, 0.005257, 0.02535, 0.01521, 0.03505, 0.01941, 0.02985], loss.cls_pos_rt=0.1012, loss.cls_neg_rt=0.04051, loss.dir_rt=0.2302, rpn_acc=0.9994, pr.prec@10=0.1764, pr.rec@10=0.9389, pr.prec@30=0.7304, pr.rec@30=0.7734, pr.prec@50=0.939, pr.rec@50=0.5731, pr.prec@70=0.9918, pr.rec@70=0.3001, pr.prec@80=0.9985, pr.rec@80=0.1124, pr.prec@90=1.0, pr.rec@90=0.001933, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=178167, misc.num_pos=54, misc.num_neg=42097, misc.num_anchors=42240, misc.lr=0.002989, misc.mem_usage=55.4\n", + "runtime.step=6000, runtime.steptime=0.2666, runtime.voxel_gene_time=0.00155, runtime.prep_time=0.03847, loss.cls_loss=0.1503, loss.cls_loss_rt=0.1489, loss.loc_loss=0.2748, loss.loc_loss_rt=0.2579, loss.loc_elem=[0.007328, 0.004314, 0.03622, 0.01061, 0.0282, 0.01528, 0.02701], loss.cls_pos_rt=0.1147, loss.cls_neg_rt=0.03428, loss.dir_rt=0.1973, rpn_acc=0.9994, pr.prec@10=0.1798, pr.rec@10=0.939, pr.prec@30=0.7338, pr.rec@30=0.7756, pr.prec@50=0.9392, pr.rec@50=0.5785, pr.prec@70=0.9919, pr.rec@70=0.3037, pr.prec@80=0.9991, pr.rec@80=0.1156, pr.prec@90=1.0, pr.rec@90=0.001695, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=177710, misc.num_pos=63, misc.num_neg=42099, misc.num_anchors=42240, misc.lr=0.002993, misc.mem_usage=55.4\n", + "runtime.step=6050, runtime.steptime=0.2688, runtime.voxel_gene_time=0.0009775, runtime.prep_time=0.03015, loss.cls_loss=0.1498, loss.cls_loss_rt=0.1325, loss.loc_loss=0.2746, loss.loc_loss_rt=0.2699, loss.loc_elem=[0.006114, 0.004382, 0.03084, 0.01544, 0.03019, 0.02068, 0.02731], loss.cls_pos_rt=0.08187, loss.cls_neg_rt=0.05067, loss.dir_rt=0.2081, rpn_acc=0.9994, pr.prec@10=0.1793, pr.rec@10=0.9402, pr.prec@30=0.734, pr.rec@30=0.7765, pr.prec@50=0.9388, pr.rec@50=0.5792, pr.prec@70=0.9924, pr.rec@70=0.3029, pr.prec@80=0.9993, pr.rec@80=0.1145, pr.prec@90=1.0, pr.rec@90=0.001746, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=177819, misc.num_pos=59, misc.num_neg=42089, misc.num_anchors=42240, misc.lr=0.002996, misc.mem_usage=55.4\n", + "runtime.step=6100, runtime.steptime=0.2697, runtime.voxel_gene_time=0.0008156, runtime.prep_time=0.0225, loss.cls_loss=0.1499, loss.cls_loss_rt=0.1351, loss.loc_loss=0.2758, loss.loc_loss_rt=0.2786, loss.loc_elem=[0.005614, 0.004831, 0.03555, 0.01496, 0.02995, 0.02609, 0.02229], loss.cls_pos_rt=0.09899, loss.cls_neg_rt=0.03608, loss.dir_rt=0.1952, rpn_acc=0.9994, pr.prec@10=0.1789, pr.rec@10=0.9407, pr.prec@30=0.7339, pr.rec@30=0.7768, pr.prec@50=0.9383, pr.rec@50=0.5781, pr.prec@70=0.9918, pr.rec@70=0.2995, pr.prec@80=0.9985, pr.rec@80=0.1101, pr.prec@90=1.0, pr.rec@90=0.001405, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=178213, misc.num_pos=54, misc.num_neg=42082, misc.num_anchors=42240, misc.lr=0.002998, misc.mem_usage=55.4\n", + "runtime.step=6150, runtime.steptime=0.2676, runtime.voxel_gene_time=0.001159, runtime.prep_time=0.02806, loss.cls_loss=0.1494, loss.cls_loss_rt=0.1476, loss.loc_loss=0.2761, loss.loc_loss_rt=0.3027, loss.loc_elem=[0.007085, 0.004616, 0.04327, 0.01296, 0.03365, 0.01888, 0.03087], loss.cls_pos_rt=0.108, loss.cls_neg_rt=0.0396, loss.dir_rt=0.2297, rpn_acc=0.9994, pr.prec@10=0.1801, pr.rec@10=0.9409, pr.prec@30=0.7349, pr.rec@30=0.7781, pr.prec@50=0.9386, pr.rec@50=0.5811, pr.prec@70=0.9921, pr.rec@70=0.3024, pr.prec@80=0.9988, pr.rec@80=0.1108, pr.prec@90=1.0, pr.rec@90=0.001295, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=176474, misc.num_pos=58, misc.num_neg=42097, misc.num_anchors=42240, misc.lr=0.003, misc.mem_usage=55.4\n", + "reset Car\n", + "reset Car\n", + "reset Car\n", + "WORKER 0 seed: 1592958965\n", + "WORKER 1 seed: 1592958966\n", + "WORKER 2 seed: 1592958967\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "runtime.step=6200, runtime.steptime=0.4303, runtime.voxel_gene_time=0.001181, runtime.prep_time=0.03181, loss.cls_loss=0.1478, loss.cls_loss_rt=0.1576, loss.loc_loss=0.2798, loss.loc_loss_rt=0.2951, loss.loc_elem=[0.004981, 0.007064, 0.03492, 0.01739, 0.02403, 0.02001, 0.03914], loss.cls_pos_rt=0.1206, loss.cls_neg_rt=0.037, loss.dir_rt=0.2559, rpn_acc=0.9994, pr.prec@10=0.1817, pr.rec@10=0.9455, pr.prec@30=0.7323, pr.rec@30=0.7745, pr.prec@50=0.9354, pr.rec@50=0.5839, pr.prec@70=0.995, pr.rec@70=0.3233, pr.prec@80=0.999, pr.rec@80=0.1413, pr.prec@90=1.0, pr.rec@90=0.003234, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=176175, misc.num_pos=63, misc.num_neg=42085, misc.num_anchors=42240, misc.lr=0.003, misc.mem_usage=55.3\n", + "#################################\n", + "# EVAL\n", + "#################################\n", + "Generate output labels...\n", + "[100.0%][===================>][12.22it/s][00:26>00:00] \n", + "generate label finished(141.05/s). start eval:\n", + "Evaluation official\n", + "Car AP(Average Precision)@0.70, 0.70, 0.70:\n", + "bbox AP:93.68, 84.69, 78.98\n", + "bev AP:88.28, 83.66, 78.28\n", + "3d AP:70.45, 62.75, 60.78\n", + "aos AP:0.86, 2.11, 2.70\n", + "Car AP(Average Precision)@0.70, 0.50, 0.50:\n", + "bbox AP:93.68, 84.69, 78.98\n", + "bev AP:95.25, 88.71, 88.17\n", + "3d AP:95.05, 88.36, 87.41\n", + "aos AP:0.86, 2.11, 2.70\n", + "\n", + "Evaluation coco\n", + "Car coco AP@0.50:0.05:0.95:\n", + "bbox AP:65.24, 60.00, 58.62\n", + "bev AP:65.60, 60.79, 59.65\n", + "3d AP:50.03, 45.91, 45.51\n", + "aos AP:0.60, 1.60, 2.22\n", + "\n", + "eval.kitti.official.Car.bbox@0.70=[93.68, 84.69, 78.98], eval.kitti.official.Car.bev@0.70=[88.28, 83.66, 78.28], eval.kitti.official.Car.3d@0.70=[70.45, 62.75, 60.78], eval.kitti.official.Car.aos=[0.8634, 2.107, 2.699], eval.kitti.official.Car.bev@0.50=[95.25, 88.71, 88.17], eval.kitti.official.Car.3d@0.50=[95.05, 88.36, 87.41], eval.kitti.coco.Car.bbox=[65.24, 60.0, 58.62], eval.kitti.coco.Car.bev=[65.6, 60.79, 59.65], eval.kitti.coco.Car.3d=[50.03, 45.91, 45.51], eval.kitti.coco.Car.aos=[0.6027, 1.605, 2.22]\n", + "runtime.step=6250, runtime.steptime=1.04, runtime.voxel_gene_time=0.0009673, runtime.prep_time=0.02595, loss.cls_loss=0.1476, loss.cls_loss_rt=0.1505, loss.loc_loss=0.2764, loss.loc_loss_rt=0.2664, loss.loc_elem=[0.005189, 0.005378, 0.02731, 0.01414, 0.02342, 0.02512, 0.03266], loss.cls_pos_rt=0.1027, loss.cls_neg_rt=0.04787, loss.dir_rt=0.1619, rpn_acc=0.9994, pr.prec@10=0.1794, pr.rec@10=0.943, pr.prec@30=0.7363, pr.rec@30=0.7789, pr.prec@50=0.9404, pr.rec@50=0.5875, pr.prec@70=0.9935, pr.rec@70=0.3159, pr.prec@80=0.9984, pr.rec@80=0.1274, pr.prec@90=1.0, pr.rec@90=0.00172, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=177099, misc.num_pos=77, misc.num_neg=42058, misc.num_anchors=42240, misc.lr=0.003, misc.mem_usage=55.5\n", + "runtime.step=6300, runtime.steptime=0.2683, runtime.voxel_gene_time=0.001012, runtime.prep_time=0.03032, loss.cls_loss=0.1463, loss.cls_loss_rt=0.1389, loss.loc_loss=0.2728, loss.loc_loss_rt=0.28, loss.loc_elem=[0.005874, 0.005555, 0.03396, 0.01455, 0.03083, 0.02387, 0.02535], loss.cls_pos_rt=0.1114, loss.cls_neg_rt=0.02746, loss.dir_rt=0.1615, rpn_acc=0.9994, pr.prec@10=0.1829, pr.rec@10=0.9439, pr.prec@30=0.7383, pr.rec@30=0.782, pr.prec@50=0.9396, pr.rec@50=0.5887, pr.prec@70=0.993, pr.rec@70=0.3128, pr.prec@80=0.9988, pr.rec@80=0.1233, pr.prec@90=1.0, pr.rec@90=0.001685, pr.prec@95=1.0, pr.rec@95=2.459e-05, misc.num_vox=175153, misc.num_pos=65, misc.num_neg=42076, misc.num_anchors=42240, misc.lr=0.002999, misc.mem_usage=55.5\n", + "runtime.step=6350, runtime.steptime=0.2695, runtime.voxel_gene_time=0.0008872, runtime.prep_time=0.02677, loss.cls_loss=0.1457, loss.cls_loss_rt=0.1265, loss.loc_loss=0.2714, loss.loc_loss_rt=0.2249, loss.loc_elem=[0.005034, 0.004503, 0.01811, 0.01509, 0.02509, 0.02046, 0.02413], loss.cls_pos_rt=0.09789, loss.cls_neg_rt=0.02863, loss.dir_rt=0.2128, rpn_acc=0.9994, pr.prec@10=0.1842, pr.rec@10=0.944, pr.prec@30=0.7399, pr.rec@30=0.7838, pr.prec@50=0.9402, pr.rec@50=0.5914, pr.prec@70=0.9927, pr.rec@70=0.3162, pr.prec@80=0.9986, pr.rec@80=0.1216, pr.prec@90=1.0, pr.rec@90=0.001498, pr.prec@95=1.0, pr.rec@95=1.742e-05, misc.num_vox=178094, misc.num_pos=43, misc.num_neg=42134, misc.num_anchors=42240, misc.lr=0.002998, misc.mem_usage=55.5\n", + "runtime.step=6400, runtime.steptime=0.2643, runtime.voxel_gene_time=0.001256, runtime.prep_time=0.02947, loss.cls_loss=0.1472, loss.cls_loss_rt=0.1206, loss.loc_loss=0.2717, loss.loc_loss_rt=0.2458, loss.loc_elem=[0.007412, 0.004397, 0.02578, 0.01047, 0.02624, 0.02191, 0.02669], loss.cls_pos_rt=0.08723, loss.cls_neg_rt=0.03332, loss.dir_rt=0.2922, rpn_acc=0.9994, pr.prec@10=0.184, pr.rec@10=0.9428, pr.prec@30=0.7381, pr.rec@30=0.7813, pr.prec@50=0.9392, pr.rec@50=0.5884, pr.prec@70=0.9926, pr.rec@70=0.3142, pr.prec@80=0.9984, pr.rec@80=0.1214, pr.prec@90=1.0, pr.rec@90=0.001743, pr.prec@95=1.0, pr.rec@95=1.346e-05, misc.num_vox=177925, misc.num_pos=56, misc.num_neg=42113, misc.num_anchors=42240, misc.lr=0.002997, misc.mem_usage=55.5\n", + "runtime.step=6450, runtime.steptime=0.2689, runtime.voxel_gene_time=0.001418, runtime.prep_time=0.03711, loss.cls_loss=0.1463, loss.cls_loss_rt=0.1112, loss.loc_loss=0.2705, loss.loc_loss_rt=0.2425, loss.loc_elem=[0.005061, 0.005058, 0.02426, 0.01187, 0.02607, 0.02006, 0.02885], loss.cls_pos_rt=0.07916, loss.cls_neg_rt=0.03201, loss.dir_rt=0.1913, rpn_acc=0.9994, pr.prec@10=0.1861, pr.rec@10=0.9432, pr.prec@30=0.7383, pr.rec@30=0.7827, pr.prec@50=0.9396, pr.rec@50=0.5891, pr.prec@70=0.9927, pr.rec@70=0.3148, pr.prec@80=0.9984, pr.rec@80=0.1231, pr.prec@90=1.0, pr.rec@90=0.001881, pr.prec@95=1.0, pr.rec@95=1.097e-05, misc.num_vox=178970, misc.num_pos=47, misc.num_neg=42116, misc.num_anchors=42240, misc.lr=0.002995, misc.mem_usage=55.6\n", + "reset Car\n", + "reset Car\n", + "reset Car\n", + "WORKER 0 seed: 1592959095\n", + "WORKER 1 seed: 1592959096\n", + "WORKER 2 seed: 1592959097\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "runtime.step=6500, runtime.steptime=0.4328, runtime.voxel_gene_time=0.001182, runtime.prep_time=0.0328, loss.cls_loss=0.1523, loss.cls_loss_rt=0.1205, loss.loc_loss=0.2676, loss.loc_loss_rt=0.2381, loss.loc_elem=[0.00573, 0.003362, 0.02288, 0.01434, 0.02505, 0.02385, 0.02385], loss.cls_pos_rt=0.096, loss.cls_neg_rt=0.02448, loss.dir_rt=0.2098, rpn_acc=0.9994, pr.prec@10=0.1808, pr.rec@10=0.9395, pr.prec@30=0.7443, pr.rec@30=0.774, pr.prec@50=0.9409, pr.rec@50=0.5792, pr.prec@70=0.9902, pr.rec@70=0.3041, pr.prec@80=0.9975, pr.rec@80=0.1043, pr.prec@90=1.0, pr.rec@90=0.0002608, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=180000, misc.num_pos=56, misc.num_neg=42093, misc.num_anchors=42240, misc.lr=0.002992, misc.mem_usage=55.6\n", + "runtime.step=6550, runtime.steptime=0.2677, runtime.voxel_gene_time=0.0009253, runtime.prep_time=0.02411, loss.cls_loss=0.1484, loss.cls_loss_rt=0.1347, loss.loc_loss=0.2718, loss.loc_loss_rt=0.2745, loss.loc_elem=[0.004783, 0.00402, 0.03128, 0.01379, 0.02885, 0.02562, 0.02892], loss.cls_pos_rt=0.09012, loss.cls_neg_rt=0.04457, loss.dir_rt=0.2455, rpn_acc=0.9994, pr.prec@10=0.1839, pr.rec@10=0.9426, pr.prec@30=0.7362, pr.rec@30=0.7795, pr.prec@50=0.94, pr.rec@50=0.5847, pr.prec@70=0.992, pr.rec@70=0.3072, pr.prec@80=0.9986, pr.rec@80=0.1176, pr.prec@90=1.0, pr.rec@90=0.0006735, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=175905, misc.num_pos=53, misc.num_neg=42102, misc.num_anchors=42240, misc.lr=0.00299, misc.mem_usage=55.7\n", + "runtime.step=6600, runtime.steptime=0.2685, runtime.voxel_gene_time=0.0009332, runtime.prep_time=0.02856, loss.cls_loss=0.1454, loss.cls_loss_rt=0.1396, loss.loc_loss=0.2696, loss.loc_loss_rt=0.2682, loss.loc_elem=[0.005807, 0.004618, 0.03316, 0.01428, 0.02361, 0.02977, 0.02284], loss.cls_pos_rt=0.111, loss.cls_neg_rt=0.02868, loss.dir_rt=0.2013, rpn_acc=0.9994, pr.prec@10=0.1884, pr.rec@10=0.944, pr.prec@30=0.7368, pr.rec@30=0.7853, pr.prec@50=0.9417, pr.rec@50=0.592, pr.prec@70=0.9925, pr.rec@70=0.3177, pr.prec@80=0.9988, pr.rec@80=0.1281, pr.prec@90=1.0, pr.rec@90=0.00163, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=178282, misc.num_pos=56, misc.num_neg=42115, misc.num_anchors=42240, misc.lr=0.002986, misc.mem_usage=55.5\n", + "runtime.step=6650, runtime.steptime=0.2679, runtime.voxel_gene_time=0.0008872, runtime.prep_time=0.02884, loss.cls_loss=0.1452, loss.cls_loss_rt=0.15, loss.loc_loss=0.2671, loss.loc_loss_rt=0.281, loss.loc_elem=[0.007036, 0.006447, 0.03359, 0.01692, 0.03062, 0.02153, 0.02437], loss.cls_pos_rt=0.09923, loss.cls_neg_rt=0.05073, loss.dir_rt=0.19, rpn_acc=0.9994, pr.prec@10=0.1883, pr.rec@10=0.9429, pr.prec@30=0.7397, pr.rec@30=0.7847, pr.prec@50=0.9416, pr.rec@50=0.593, pr.prec@70=0.9925, pr.rec@70=0.3198, pr.prec@80=0.9985, pr.rec@80=0.1317, pr.prec@90=1.0, pr.rec@90=0.001796, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=172953, misc.num_pos=51, misc.num_neg=42109, misc.num_anchors=42240, misc.lr=0.002983, misc.mem_usage=55.6\n", + "runtime.step=6700, runtime.steptime=0.2692, runtime.voxel_gene_time=0.001446, runtime.prep_time=0.03334, loss.cls_loss=0.1451, loss.cls_loss_rt=0.1623, loss.loc_loss=0.2685, loss.loc_loss_rt=0.2996, loss.loc_elem=[0.006406, 0.005884, 0.03079, 0.01449, 0.02941, 0.02341, 0.03942], loss.cls_pos_rt=0.1144, loss.cls_neg_rt=0.04796, loss.dir_rt=0.2358, rpn_acc=0.9994, pr.prec@10=0.189, pr.rec@10=0.9422, pr.prec@30=0.7413, pr.rec@30=0.7858, pr.prec@50=0.942, pr.rec@50=0.5943, pr.prec@70=0.9925, pr.rec@70=0.3218, pr.prec@80=0.9981, pr.rec@80=0.1344, pr.prec@90=1.0, pr.rec@90=0.002067, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=178199, misc.num_pos=49, misc.num_neg=42111, misc.num_anchors=42240, misc.lr=0.002979, misc.mem_usage=55.6\n", + "runtime.step=6750, runtime.steptime=0.2661, runtime.voxel_gene_time=0.002273, runtime.prep_time=0.04269, loss.cls_loss=0.1442, loss.cls_loss_rt=0.1096, loss.loc_loss=0.2672, loss.loc_loss_rt=0.232, loss.loc_elem=[0.005471, 0.003616, 0.02679, 0.01227, 0.0203, 0.03132, 0.01623], loss.cls_pos_rt=0.07675, loss.cls_neg_rt=0.03288, loss.dir_rt=0.1761, rpn_acc=0.9994, pr.prec@10=0.1896, pr.rec@10=0.9425, pr.prec@30=0.7428, pr.rec@30=0.7871, pr.prec@50=0.942, pr.rec@50=0.5954, pr.prec@70=0.9922, pr.rec@70=0.3248, pr.prec@80=0.998, pr.rec@80=0.1355, pr.prec@90=1.0, pr.rec@90=0.002216, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=179278, misc.num_pos=63, misc.num_neg=42070, misc.num_anchors=42240, misc.lr=0.002974, misc.mem_usage=55.6\n", + "reset Car\n", + "reset Car\n", + "reset Car\n", + "WORKER 0 seed: 1592959186\n", + "WORKER 1 seed: 1592959187\n", + "WORKER 2 seed: 1592959188\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "runtime.step=6800, runtime.steptime=0.4305, runtime.voxel_gene_time=0.01948, runtime.prep_time=7.272, loss.cls_loss=0.1616, loss.cls_loss_rt=0.1552, loss.loc_loss=0.277, loss.loc_loss_rt=0.2771, loss.loc_elem=[0.005528, 0.004697, 0.03028, 0.01425, 0.02364, 0.02696, 0.0332], loss.cls_pos_rt=0.1185, loss.cls_neg_rt=0.03674, loss.dir_rt=0.1776, rpn_acc=0.9994, pr.prec@10=0.1996, pr.rec@10=0.9181, pr.prec@30=0.7456, pr.rec@30=0.7326, pr.prec@50=0.9543, pr.rec@50=0.5746, pr.prec@70=0.991, pr.rec@70=0.3181, pr.prec@80=1.0, pr.rec@80=0.1261, pr.prec@90=1.0, pr.rec@90=0.002174, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=179556, misc.num_pos=51, misc.num_neg=42113, misc.num_anchors=42240, misc.lr=0.002969, misc.mem_usage=54.7\n", + "runtime.step=6850, runtime.steptime=0.2667, runtime.voxel_gene_time=0.001562, runtime.prep_time=0.03722, loss.cls_loss=0.1413, loss.cls_loss_rt=0.1041, loss.loc_loss=0.2672, loss.loc_loss_rt=0.221, loss.loc_elem=[0.004566, 0.00361, 0.02182, 0.01413, 0.02561, 0.0199, 0.02088], loss.cls_pos_rt=0.08078, loss.cls_neg_rt=0.02333, loss.dir_rt=0.1567, rpn_acc=0.9994, pr.prec@10=0.192, pr.rec@10=0.9441, pr.prec@30=0.7481, pr.rec@30=0.7911, pr.prec@50=0.9449, pr.rec@50=0.6033, pr.prec@70=0.9933, pr.rec@70=0.3352, pr.prec@80=0.9992, pr.rec@80=0.1383, pr.prec@90=1.0, pr.rec@90=0.002661, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=178760, misc.num_pos=51, misc.num_neg=42092, misc.num_anchors=42240, misc.lr=0.002964, misc.mem_usage=55.5\n", + "runtime.step=6900, runtime.steptime=0.2648, runtime.voxel_gene_time=0.00154, runtime.prep_time=0.0399, loss.cls_loss=0.1408, loss.cls_loss_rt=0.1411, loss.loc_loss=0.2627, loss.loc_loss_rt=0.257, loss.loc_elem=[0.006734, 0.005546, 0.02446, 0.01522, 0.02652, 0.01851, 0.03153], loss.cls_pos_rt=0.1089, loss.cls_neg_rt=0.0322, loss.dir_rt=0.1774, rpn_acc=0.9994, pr.prec@10=0.1928, pr.rec@10=0.9433, pr.prec@30=0.7468, pr.rec@30=0.7941, pr.prec@50=0.9437, pr.rec@50=0.6069, pr.prec@70=0.9927, pr.rec@70=0.3326, pr.prec@80=0.9987, pr.rec@80=0.1347, pr.prec@90=1.0, pr.rec@90=0.002062, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=178110, misc.num_pos=57, misc.num_neg=42098, misc.num_anchors=42240, misc.lr=0.002958, misc.mem_usage=55.4\n", + "runtime.step=6950, runtime.steptime=0.2652, runtime.voxel_gene_time=0.001096, runtime.prep_time=0.02666, loss.cls_loss=0.1406, loss.cls_loss_rt=0.1499, loss.loc_loss=0.2624, loss.loc_loss_rt=0.2915, loss.loc_elem=[0.004928, 0.005191, 0.03135, 0.01546, 0.03078, 0.03013, 0.02795], loss.cls_pos_rt=0.09387, loss.cls_neg_rt=0.05601, loss.dir_rt=0.2284, rpn_acc=0.9994, pr.prec@10=0.1936, pr.rec@10=0.9443, pr.prec@30=0.7466, pr.rec@30=0.794, pr.prec@50=0.9419, pr.rec@50=0.6053, pr.prec@70=0.9921, pr.rec@70=0.3331, pr.prec@80=0.9983, pr.rec@80=0.1393, pr.prec@90=1.0, pr.rec@90=0.002717, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=177350, misc.num_pos=51, misc.num_neg=42111, misc.num_anchors=42240, misc.lr=0.002952, misc.mem_usage=55.4\n", + "runtime.step=7000, runtime.steptime=0.2654, runtime.voxel_gene_time=0.0009232, runtime.prep_time=0.02576, loss.cls_loss=0.1398, loss.cls_loss_rt=0.1406, loss.loc_loss=0.261, loss.loc_loss_rt=0.2579, loss.loc_elem=[0.005725, 0.005064, 0.01979, 0.01177, 0.02815, 0.02028, 0.03817], loss.cls_pos_rt=0.1032, loss.cls_neg_rt=0.03745, loss.dir_rt=0.1341, rpn_acc=0.9994, pr.prec@10=0.1949, pr.rec@10=0.9449, pr.prec@30=0.7459, pr.rec@30=0.7956, pr.prec@50=0.9428, pr.rec@50=0.6056, pr.prec@70=0.9923, pr.rec@70=0.3332, pr.prec@80=0.9984, pr.rec@80=0.1391, pr.prec@90=1.0, pr.rec@90=0.002581, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=173362, misc.num_pos=48, misc.num_neg=42106, misc.num_anchors=42240, misc.lr=0.002946, misc.mem_usage=55.5\n", + "runtime.step=7050, runtime.steptime=0.2658, runtime.voxel_gene_time=0.0006766, runtime.prep_time=0.01924, loss.cls_loss=0.1399, loss.cls_loss_rt=0.1503, loss.loc_loss=0.2595, loss.loc_loss_rt=0.275, loss.loc_elem=[0.006317, 0.00616, 0.03368, 0.01615, 0.02556, 0.02084, 0.0288], loss.cls_pos_rt=0.106, loss.cls_neg_rt=0.04424, loss.dir_rt=0.1955, rpn_acc=0.9994, pr.prec@10=0.1951, pr.rec@10=0.9448, pr.prec@30=0.7459, pr.rec@30=0.7943, pr.prec@50=0.9423, pr.rec@50=0.6045, pr.prec@70=0.9922, pr.rec@70=0.3344, pr.prec@80=0.9983, pr.rec@80=0.1416, pr.prec@90=1.0, pr.rec@90=0.002739, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=169823, misc.num_pos=58, misc.num_neg=42093, misc.num_anchors=42240, misc.lr=0.002939, misc.mem_usage=55.5\n", + "reset Car\n", + "reset Car\n", + "reset Car\n", + "runtime.step=7100, runtime.steptime=0.2651, runtime.voxel_gene_time=0.001477, runtime.prep_time=0.03288, loss.cls_loss=0.1395, loss.cls_loss_rt=0.1296, loss.loc_loss=0.2579, loss.loc_loss_rt=0.2754, loss.loc_elem=[0.005735, 0.005018, 0.02371, 0.01338, 0.03033, 0.02337, 0.03618], loss.cls_pos_rt=0.08717, loss.cls_neg_rt=0.04238, loss.dir_rt=0.1284, rpn_acc=0.9994, pr.prec@10=0.1948, pr.rec@10=0.9454, pr.prec@30=0.7454, pr.rec@30=0.7951, pr.prec@50=0.9423, pr.rec@50=0.6056, pr.prec@70=0.9922, pr.rec@70=0.3371, pr.prec@80=0.9983, pr.rec@80=0.1441, pr.prec@90=1.0, pr.rec@90=0.002815, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=175330, misc.num_pos=48, misc.num_neg=42131, misc.num_anchors=42240, misc.lr=0.002931, misc.mem_usage=55.6\n", + "WORKER 0 seed: 1592959276\n", + "WORKER 1 seed: 1592959277\n", + "WORKER 2 seed: 1592959278\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "runtime.step=7150, runtime.steptime=0.4309, runtime.voxel_gene_time=0.001368, runtime.prep_time=0.03236, loss.cls_loss=0.1441, loss.cls_loss_rt=0.1564, loss.loc_loss=0.2706, loss.loc_loss_rt=0.2675, loss.loc_elem=[0.006107, 0.004607, 0.03033, 0.01183, 0.02788, 0.02123, 0.03177], loss.cls_pos_rt=0.1282, loss.cls_neg_rt=0.02819, loss.dir_rt=0.229, rpn_acc=0.9994, pr.prec@10=0.192, pr.rec@10=0.94, pr.prec@30=0.7452, pr.rec@30=0.7898, pr.prec@50=0.9422, pr.rec@50=0.6022, pr.prec@70=0.9927, pr.rec@70=0.3426, pr.prec@80=0.9986, pr.rec@80=0.1534, pr.prec@90=1.0, pr.rec@90=0.006481, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=177843, misc.num_pos=42, misc.num_neg=42124, misc.num_anchors=42240, misc.lr=0.002924, misc.mem_usage=55.5\n", + "runtime.step=7200, runtime.steptime=0.2674, runtime.voxel_gene_time=0.0009248, runtime.prep_time=0.02787, loss.cls_loss=0.1395, loss.cls_loss_rt=0.151, loss.loc_loss=0.2631, loss.loc_loss_rt=0.2777, loss.loc_elem=[0.007386, 0.005585, 0.02461, 0.009773, 0.02853, 0.02024, 0.04273], loss.cls_pos_rt=0.1205, loss.cls_neg_rt=0.03046, loss.dir_rt=0.1384, rpn_acc=0.9994, pr.prec@10=0.1974, pr.rec@10=0.9438, pr.prec@30=0.7509, pr.rec@30=0.7966, pr.prec@50=0.9428, pr.rec@50=0.6097, pr.prec@70=0.9928, pr.rec@70=0.3425, pr.prec@80=0.9985, pr.rec@80=0.1527, pr.prec@90=1.0, pr.rec@90=0.00518, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=178256, misc.num_pos=66, misc.num_neg=42088, misc.num_anchors=42240, misc.lr=0.002915, misc.mem_usage=55.5\n", + "runtime.step=7250, runtime.steptime=0.2693, runtime.voxel_gene_time=0.001183, runtime.prep_time=0.02835, loss.cls_loss=0.1384, loss.cls_loss_rt=0.1088, loss.loc_loss=0.2576, loss.loc_loss_rt=0.2027, loss.loc_elem=[0.00545, 0.003326, 0.01836, 0.01256, 0.0218, 0.02059, 0.01928], loss.cls_pos_rt=0.07579, loss.cls_neg_rt=0.03303, loss.dir_rt=0.1247, rpn_acc=0.9994, pr.prec@10=0.1967, pr.rec@10=0.9449, pr.prec@30=0.7512, pr.rec@30=0.797, pr.prec@50=0.943, pr.rec@50=0.6111, pr.prec@70=0.9928, pr.rec@70=0.3469, pr.prec@80=0.9987, pr.rec@80=0.1579, pr.prec@90=1.0, pr.rec@90=0.005558, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=179239, misc.num_pos=62, misc.num_neg=42088, misc.num_anchors=42240, misc.lr=0.002907, misc.mem_usage=55.5\n", + "runtime.step=7300, runtime.steptime=0.2661, runtime.voxel_gene_time=0.001002, runtime.prep_time=0.02847, loss.cls_loss=0.1379, loss.cls_loss_rt=0.1678, loss.loc_loss=0.2566, loss.loc_loss_rt=0.29, loss.loc_elem=[0.004792, 0.005734, 0.03815, 0.01453, 0.02642, 0.02811, 0.02728], loss.cls_pos_rt=0.1352, loss.cls_neg_rt=0.03259, loss.dir_rt=0.231, rpn_acc=0.9994, pr.prec@10=0.1985, pr.rec@10=0.9452, pr.prec@30=0.7517, pr.rec@30=0.7973, pr.prec@50=0.9432, pr.rec@50=0.6125, pr.prec@70=0.9928, pr.rec@70=0.3526, pr.prec@80=0.9988, pr.rec@80=0.1654, pr.prec@90=1.0, pr.rec@90=0.006854, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=176555, misc.num_pos=66, misc.num_neg=42092, misc.num_anchors=42240, misc.lr=0.002898, misc.mem_usage=55.5\n", + "runtime.step=7350, runtime.steptime=0.2684, runtime.voxel_gene_time=0.001136, runtime.prep_time=0.04156, loss.cls_loss=0.1379, loss.cls_loss_rt=0.1104, loss.loc_loss=0.2571, loss.loc_loss_rt=0.2157, loss.loc_elem=[0.005357, 0.004418, 0.01856, 0.0105, 0.02233, 0.02004, 0.02664], loss.cls_pos_rt=0.07415, loss.cls_neg_rt=0.03623, loss.dir_rt=0.218, rpn_acc=0.9994, pr.prec@10=0.1973, pr.rec@10=0.9461, pr.prec@30=0.7507, pr.rec@30=0.7973, pr.prec@50=0.9429, pr.rec@50=0.6094, pr.prec@70=0.9922, pr.rec@70=0.3469, pr.prec@80=0.9988, pr.rec@80=0.1613, pr.prec@90=1.0, pr.rec@90=0.006061, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=177219, misc.num_pos=66, misc.num_neg=42076, misc.num_anchors=42240, misc.lr=0.002888, misc.mem_usage=55.5\n", + "runtime.step=7400, runtime.steptime=0.2679, runtime.voxel_gene_time=0.00148, runtime.prep_time=0.03702, loss.cls_loss=0.1374, loss.cls_loss_rt=0.1121, loss.loc_loss=0.256, loss.loc_loss_rt=0.2108, loss.loc_elem=[0.005107, 0.004795, 0.01736, 0.01554, 0.02735, 0.01652, 0.01874], loss.cls_pos_rt=0.07981, loss.cls_neg_rt=0.03229, loss.dir_rt=0.1224, rpn_acc=0.9994, pr.prec@10=0.1986, pr.rec@10=0.9463, pr.prec@30=0.7509, pr.rec@30=0.7979, pr.prec@50=0.9429, pr.rec@50=0.6106, pr.prec@70=0.9925, pr.rec@70=0.3493, pr.prec@80=0.9989, pr.rec@80=0.1633, pr.prec@90=1.0, pr.rec@90=0.006219, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=177904, misc.num_pos=47, misc.num_neg=42116, misc.num_anchors=42240, misc.lr=0.002879, misc.mem_usage=55.5\n", + "reset Car\n", + "reset Car\n", + "reset Car\n", + "WORKER 0 seed: 1592959367\n", + "WORKER 1 seed: 1592959368\n", + "WORKER 2 seed: 1592959369\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "runtime.step=7450, runtime.steptime=0.4316, runtime.voxel_gene_time=0.001396, runtime.prep_time=0.03169, loss.cls_loss=0.1424, loss.cls_loss_rt=0.1491, loss.loc_loss=0.2616, loss.loc_loss_rt=0.2856, loss.loc_elem=[0.006868, 0.004587, 0.0282, 0.01292, 0.03302, 0.02305, 0.03417], loss.cls_pos_rt=0.1084, loss.cls_neg_rt=0.04075, loss.dir_rt=0.2207, rpn_acc=0.9994, pr.prec@10=0.1881, pr.rec@10=0.9435, pr.prec@30=0.7449, pr.rec@30=0.7883, pr.prec@50=0.943, pr.rec@50=0.6004, pr.prec@70=0.9906, pr.rec@70=0.3458, pr.prec@80=0.9982, pr.rec@80=0.168, pr.prec@90=1.0, pr.rec@90=0.007104, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=176522, misc.num_pos=61, misc.num_neg=42100, misc.num_anchors=42240, misc.lr=0.002868, misc.mem_usage=55.6\n", + "runtime.step=7500, runtime.steptime=0.2687, runtime.voxel_gene_time=0.001393, runtime.prep_time=0.03346, loss.cls_loss=0.1382, loss.cls_loss_rt=0.1303, loss.loc_loss=0.2544, loss.loc_loss_rt=0.2687, loss.loc_elem=[0.006468, 0.004164, 0.02469, 0.01995, 0.02959, 0.02089, 0.02862], loss.cls_pos_rt=0.0966, loss.cls_neg_rt=0.03373, loss.dir_rt=0.1873, rpn_acc=0.9994, pr.prec@10=0.1954, pr.rec@10=0.9474, pr.prec@30=0.7493, pr.rec@30=0.7936, pr.prec@50=0.9448, pr.rec@50=0.6084, pr.prec@70=0.9923, pr.rec@70=0.3541, pr.prec@80=0.9991, pr.rec@80=0.1688, pr.prec@90=1.0, pr.rec@90=0.006597, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=176934, misc.num_pos=65, misc.num_neg=42074, misc.num_anchors=42240, misc.lr=0.002858, misc.mem_usage=55.6\n", + "runtime.step=7550, runtime.steptime=0.271, runtime.voxel_gene_time=0.001846, runtime.prep_time=0.03867, loss.cls_loss=0.1368, loss.cls_loss_rt=0.1167, loss.loc_loss=0.2532, loss.loc_loss_rt=0.223, loss.loc_elem=[0.005278, 0.00481, 0.02208, 0.01571, 0.02275, 0.01861, 0.02227], loss.cls_pos_rt=0.08021, loss.cls_neg_rt=0.03644, loss.dir_rt=0.1901, rpn_acc=0.9994, pr.prec@10=0.1974, pr.rec@10=0.9479, pr.prec@30=0.7525, pr.rec@30=0.7974, pr.prec@50=0.9443, pr.rec@50=0.6145, pr.prec@70=0.9924, pr.rec@70=0.3572, pr.prec@80=0.9992, pr.rec@80=0.1721, pr.prec@90=1.0, pr.rec@90=0.007319, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=178496, misc.num_pos=75, misc.num_neg=42056, misc.num_anchors=42240, misc.lr=0.002847, misc.mem_usage=55.6\n", + "runtime.step=7600, runtime.steptime=0.268, runtime.voxel_gene_time=0.001206, runtime.prep_time=0.03053, loss.cls_loss=0.1367, loss.cls_loss_rt=0.1457, loss.loc_loss=0.2535, loss.loc_loss_rt=0.2379, loss.loc_elem=[0.0054, 0.005618, 0.02347, 0.01176, 0.02266, 0.02613, 0.0239], loss.cls_pos_rt=0.1054, loss.cls_neg_rt=0.04036, loss.dir_rt=0.1844, rpn_acc=0.9994, pr.prec@10=0.1985, pr.rec@10=0.9464, pr.prec@30=0.7525, pr.rec@30=0.7973, pr.prec@50=0.9456, pr.rec@50=0.6161, pr.prec@70=0.9924, pr.rec@70=0.3602, pr.prec@80=0.9991, pr.rec@80=0.175, pr.prec@90=1.0, pr.rec@90=0.007867, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=178115, misc.num_pos=61, misc.num_neg=42087, misc.num_anchors=42240, misc.lr=0.002836, misc.mem_usage=55.5\n", + "runtime.step=7650, runtime.steptime=0.2678, runtime.voxel_gene_time=0.001545, runtime.prep_time=0.03675, loss.cls_loss=0.1372, loss.cls_loss_rt=0.1256, loss.loc_loss=0.2549, loss.loc_loss_rt=0.2032, loss.loc_elem=[0.005143, 0.004013, 0.01775, 0.01419, 0.02078, 0.01934, 0.02036], loss.cls_pos_rt=0.08992, loss.cls_neg_rt=0.03564, loss.dir_rt=0.1396, rpn_acc=0.9994, pr.prec@10=0.1969, pr.rec@10=0.9463, pr.prec@30=0.752, pr.rec@30=0.7972, pr.prec@50=0.945, pr.rec@50=0.612, pr.prec@70=0.9922, pr.rec@70=0.3548, pr.prec@80=0.999, pr.rec@80=0.1718, pr.prec@90=1.0, pr.rec@90=0.007418, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=177980, misc.num_pos=45, misc.num_neg=42132, misc.num_anchors=42240, misc.lr=0.002824, misc.mem_usage=55.5\n", + "runtime.step=7700, runtime.steptime=0.2696, runtime.voxel_gene_time=0.0008907, runtime.prep_time=0.02802, loss.cls_loss=0.1364, loss.cls_loss_rt=0.1689, loss.loc_loss=0.253, loss.loc_loss_rt=0.2724, loss.loc_elem=[0.005488, 0.005987, 0.02306, 0.01586, 0.02554, 0.02149, 0.03876], loss.cls_pos_rt=0.1443, loss.cls_neg_rt=0.02459, loss.dir_rt=0.1709, rpn_acc=0.9994, pr.prec@10=0.1989, pr.rec@10=0.9465, pr.prec@30=0.7543, pr.rec@30=0.7989, pr.prec@50=0.9458, pr.rec@50=0.6159, pr.prec@70=0.9925, pr.rec@70=0.3583, pr.prec@80=0.9988, pr.rec@80=0.1737, pr.prec@90=1.0, pr.rec@90=0.007511, pr.prec@95=1.0, pr.rec@95=5.209e-06, misc.num_vox=178227, misc.num_pos=43, misc.num_neg=42140, misc.num_anchors=42240, misc.lr=0.002812, misc.mem_usage=55.6\n", + "reset Car\n", + "reset Car\n", + "reset Car\n", + "WORKER 0 seed: 1592959458\n", + "WORKER 1 seed: 1592959459\n", + "WORKER 2 seed: 1592959460\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "runtime.step=7750, runtime.steptime=0.433, runtime.voxel_gene_time=0.0009587, runtime.prep_time=0.02955, loss.cls_loss=0.1436, loss.cls_loss_rt=0.1309, loss.loc_loss=0.2583, loss.loc_loss_rt=0.2542, loss.loc_elem=[0.004313, 0.004785, 0.02606, 0.01791, 0.02603, 0.02889, 0.01912], loss.cls_pos_rt=0.08783, loss.cls_neg_rt=0.04302, loss.dir_rt=0.2198, rpn_acc=0.9994, pr.prec@10=0.1875, pr.rec@10=0.9463, pr.prec@30=0.7427, pr.rec@30=0.7861, pr.prec@50=0.9455, pr.rec@50=0.5942, pr.prec@70=0.9934, pr.rec@70=0.3371, pr.prec@80=0.9989, pr.rec@80=0.1683, pr.prec@90=1.0, pr.rec@90=0.01153, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=178049, misc.num_pos=72, misc.num_neg=42064, misc.num_anchors=42240, misc.lr=0.002799, misc.mem_usage=55.3\n", + "#################################\n", + "# EVAL\n", + "#################################\n", + "Generate output labels...\n", + "[100.0%][===================>][12.18it/s][00:26>00:00] \n", + "generate label finished(141.39/s). start eval:\n", + "Evaluation official\n", + "Car AP(Average Precision)@0.70, 0.70, 0.70:\n", + "bbox AP:89.72, 86.32, 79.63\n", + "bev AP:89.14, 79.47, 78.70\n", + "3d AP:80.50, 66.87, 65.49\n", + "aos AP:0.67, 1.64, 2.32\n", + "Car AP(Average Precision)@0.70, 0.50, 0.50:\n", + "bbox AP:89.72, 86.32, 79.63\n", + "bev AP:90.26, 88.44, 87.73\n", + "3d AP:90.19, 88.15, 87.38\n", + "aos AP:0.67, 1.64, 2.32\n", + "\n", + "Evaluation coco\n", + "Car coco AP@0.50:0.05:0.95:\n", + "bbox AP:67.01, 63.63, 62.32\n", + "bev AP:66.55, 62.23, 60.78\n", + "3d AP:54.78, 50.17, 48.64\n", + "aos AP:0.48, 1.22, 1.90\n", + "\n", + "eval.kitti.official.Car.bbox@0.70=[89.72, 86.32, 79.63], eval.kitti.official.Car.bev@0.70=[89.14, 79.47, 78.7], eval.kitti.official.Car.3d@0.70=[80.5, 66.87, 65.49], eval.kitti.official.Car.aos=[0.6667, 1.64, 2.324], eval.kitti.official.Car.bev@0.50=[90.26, 88.44, 87.73], eval.kitti.official.Car.3d@0.50=[90.19, 88.15, 87.38], eval.kitti.coco.Car.bbox=[67.01, 63.63, 62.32], eval.kitti.coco.Car.bev=[66.55, 62.23, 60.78], eval.kitti.coco.Car.3d=[54.78, 50.17, 48.64], eval.kitti.coco.Car.aos=[0.4752, 1.221, 1.904]\n", + "runtime.step=7800, runtime.steptime=1.045, runtime.voxel_gene_time=0.001081, runtime.prep_time=0.03168, loss.cls_loss=0.1388, loss.cls_loss_rt=0.1129, loss.loc_loss=0.255, loss.loc_loss_rt=0.2521, loss.loc_elem=[0.005743, 0.004296, 0.02382, 0.01422, 0.03091, 0.02137, 0.02572], loss.cls_pos_rt=0.07879, loss.cls_neg_rt=0.03411, loss.dir_rt=0.1622, rpn_acc=0.9994, pr.prec@10=0.1949, pr.rec@10=0.9461, pr.prec@30=0.7521, pr.rec@30=0.796, pr.prec@50=0.9448, pr.rec@50=0.6085, pr.prec@70=0.9928, pr.rec@70=0.3486, pr.prec@80=0.9989, pr.rec@80=0.16, pr.prec@90=1.0, pr.rec@90=0.007667, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=177810, misc.num_pos=52, misc.num_neg=42112, misc.num_anchors=42240, misc.lr=0.002786, misc.mem_usage=55.6\n", + "runtime.step=7850, runtime.steptime=0.2693, runtime.voxel_gene_time=0.0009058, runtime.prep_time=0.02906, loss.cls_loss=0.1351, loss.cls_loss_rt=0.1156, loss.loc_loss=0.2511, loss.loc_loss_rt=0.237, loss.loc_elem=[0.004949, 0.003789, 0.03224, 0.01329, 0.02467, 0.02067, 0.01889], loss.cls_pos_rt=0.07312, loss.cls_neg_rt=0.04253, loss.dir_rt=0.197, rpn_acc=0.9994, pr.prec@10=0.2008, pr.rec@10=0.9473, pr.prec@30=0.757, pr.rec@30=0.803, pr.prec@50=0.9458, pr.rec@50=0.6187, pr.prec@70=0.9928, pr.rec@70=0.358, pr.prec@80=0.9988, pr.rec@80=0.1681, pr.prec@90=1.0, pr.rec@90=0.006737, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=171219, misc.num_pos=66, misc.num_neg=42072, misc.num_anchors=42240, misc.lr=0.002773, misc.mem_usage=55.6\n", + "runtime.step=7900, runtime.steptime=0.2664, runtime.voxel_gene_time=0.001014, runtime.prep_time=0.02815, loss.cls_loss=0.1341, loss.cls_loss_rt=0.09497, loss.loc_loss=0.2497, loss.loc_loss_rt=0.2162, loss.loc_elem=[0.00519, 0.004209, 0.02176, 0.01137, 0.0229, 0.02193, 0.02073], loss.cls_pos_rt=0.07118, loss.cls_neg_rt=0.02379, loss.dir_rt=0.22, rpn_acc=0.9994, pr.prec@10=0.2029, pr.rec@10=0.9477, pr.prec@30=0.7589, pr.rec@30=0.8042, pr.prec@50=0.9455, pr.rec@50=0.6215, pr.prec@70=0.9929, pr.rec@70=0.3628, pr.prec@80=0.999, pr.rec@80=0.1725, pr.prec@90=1.0, pr.rec@90=0.006303, pr.prec@95=1.0, pr.rec@95=8.494e-06, misc.num_vox=179011, misc.num_pos=57, misc.num_neg=42106, misc.num_anchors=42240, misc.lr=0.00276, misc.mem_usage=55.6\n", + "runtime.step=7950, runtime.steptime=0.2684, runtime.voxel_gene_time=0.0009534, runtime.prep_time=0.02854, loss.cls_loss=0.1341, loss.cls_loss_rt=0.1691, loss.loc_loss=0.249, loss.loc_loss_rt=0.2865, loss.loc_elem=[0.005522, 0.004641, 0.03026, 0.01379, 0.03207, 0.02597, 0.03098], loss.cls_pos_rt=0.1268, loss.cls_neg_rt=0.04231, loss.dir_rt=0.1763, rpn_acc=0.9994, pr.prec@10=0.203, pr.rec@10=0.9482, pr.prec@30=0.7564, pr.rec@30=0.8034, pr.prec@50=0.9446, pr.rec@50=0.62, pr.prec@70=0.9928, pr.rec@70=0.3618, pr.prec@80=0.999, pr.rec@80=0.1722, pr.prec@90=1.0, pr.rec@90=0.006242, pr.prec@95=1.0, pr.rec@95=6.585e-06, misc.num_vox=178048, misc.num_pos=70, misc.num_neg=42067, misc.num_anchors=42240, misc.lr=0.002746, misc.mem_usage=55.6\n", + "runtime.step=8000, runtime.steptime=0.2693, runtime.voxel_gene_time=0.001362, runtime.prep_time=0.03812, loss.cls_loss=0.1346, loss.cls_loss_rt=0.1254, loss.loc_loss=0.2488, loss.loc_loss_rt=0.2307, loss.loc_elem=[0.005113, 0.003782, 0.02822, 0.0131, 0.02537, 0.01783, 0.02191], loss.cls_pos_rt=0.08862, loss.cls_neg_rt=0.03674, loss.dir_rt=0.1911, rpn_acc=0.9994, pr.prec@10=0.2015, pr.rec@10=0.9473, pr.prec@30=0.7566, pr.rec@30=0.8028, pr.prec@50=0.9457, pr.rec@50=0.619, pr.prec@70=0.9929, pr.rec@70=0.3619, pr.prec@80=0.9992, pr.rec@80=0.1718, pr.prec@90=1.0, pr.rec@90=0.006164, pr.prec@95=1.0, pr.rec@95=3.769e-05, misc.num_vox=179835, misc.num_pos=40, misc.num_neg=42132, misc.num_anchors=42240, misc.lr=0.002731, misc.mem_usage=55.6\n", + "reset Car\n", + "reset Car\n", + "reset Car\n", + "WORKER 0 seed: 1592959588\n", + "WORKER 1 seed: 1592959589\n", + "WORKER 2 seed: 1592959590\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "runtime.step=8050, runtime.steptime=0.4306, runtime.voxel_gene_time=0.001719, runtime.prep_time=0.03883, loss.cls_loss=0.1426, loss.cls_loss_rt=0.1192, loss.loc_loss=0.2516, loss.loc_loss_rt=0.2494, loss.loc_elem=[0.005217, 0.003857, 0.02952, 0.01297, 0.02421, 0.02502, 0.02388], loss.cls_pos_rt=0.08164, loss.cls_neg_rt=0.03759, loss.dir_rt=0.2129, rpn_acc=0.9994, pr.prec@10=0.1841, pr.rec@10=0.9394, pr.prec@30=0.7342, pr.rec@30=0.7912, pr.prec@50=0.9406, pr.rec@50=0.6135, pr.prec@70=0.995, pr.rec@70=0.3651, pr.prec@80=0.998, pr.rec@80=0.1852, pr.prec@90=1.0, pr.rec@90=0.01467, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=176582, misc.num_pos=69, misc.num_neg=42064, misc.num_anchors=42240, misc.lr=0.002717, misc.mem_usage=55.4\n", + "runtime.step=8100, runtime.steptime=0.2652, runtime.voxel_gene_time=0.0008757, runtime.prep_time=0.02679, loss.cls_loss=0.1381, loss.cls_loss_rt=0.1471, loss.loc_loss=0.2459, loss.loc_loss_rt=0.2221, loss.loc_elem=[0.004875, 0.005357, 0.02116, 0.01193, 0.02226, 0.02217, 0.02331], loss.cls_pos_rt=0.1071, loss.cls_neg_rt=0.03999, loss.dir_rt=0.1737, rpn_acc=0.9994, pr.prec@10=0.1927, pr.rec@10=0.9436, pr.prec@30=0.7547, pr.rec@30=0.7966, pr.prec@50=0.9433, pr.rec@50=0.6144, pr.prec@70=0.9922, pr.rec@70=0.3632, pr.prec@80=0.9982, pr.rec@80=0.1787, pr.prec@90=1.0, pr.rec@90=0.01094, pr.prec@95=1.0, pr.rec@95=2.256e-05, misc.num_vox=176301, misc.num_pos=50, misc.num_neg=42094, misc.num_anchors=42240, misc.lr=0.002702, misc.mem_usage=55.4\n", + "runtime.step=8150, runtime.steptime=0.2672, runtime.voxel_gene_time=0.0007286, runtime.prep_time=0.02324, loss.cls_loss=0.1354, loss.cls_loss_rt=0.1093, loss.loc_loss=0.2491, loss.loc_loss_rt=0.2146, loss.loc_elem=[0.004841, 0.004194, 0.02211, 0.01477, 0.0225, 0.0208, 0.01806], loss.cls_pos_rt=0.07426, loss.cls_neg_rt=0.03506, loss.dir_rt=0.1646, rpn_acc=0.9994, pr.prec@10=0.197, pr.rec@10=0.9469, pr.prec@30=0.7551, pr.rec@30=0.8006, pr.prec@50=0.9436, pr.rec@50=0.617, pr.prec@70=0.993, pr.rec@70=0.3645, pr.prec@80=0.9988, pr.rec@80=0.1768, pr.prec@90=1.0, pr.rec@90=0.008358, pr.prec@95=1.0, pr.rec@95=1.278e-05, misc.num_vox=176624, misc.num_pos=56, misc.num_neg=42103, misc.num_anchors=42240, misc.lr=0.002686, misc.mem_usage=55.4\n", + "runtime.step=8200, runtime.steptime=0.2708, runtime.voxel_gene_time=0.001492, runtime.prep_time=0.0359, loss.cls_loss=0.1347, loss.cls_loss_rt=0.1371, loss.loc_loss=0.2471, loss.loc_loss_rt=0.2054, loss.loc_elem=[0.004916, 0.004016, 0.02043, 0.01321, 0.02185, 0.02078, 0.01749], loss.cls_pos_rt=0.09135, loss.cls_neg_rt=0.04577, loss.dir_rt=0.109, rpn_acc=0.9994, pr.prec@10=0.1999, pr.rec@10=0.9475, pr.prec@30=0.7568, pr.rec@30=0.8028, pr.prec@50=0.9446, pr.rec@50=0.6197, pr.prec@70=0.9924, pr.rec@70=0.3666, pr.prec@80=0.9988, pr.rec@80=0.1806, pr.prec@90=1.0, pr.rec@90=0.009076, pr.prec@95=1.0, pr.rec@95=3.57e-05, misc.num_vox=179308, misc.num_pos=54, misc.num_neg=42108, misc.num_anchors=42240, misc.lr=0.002671, misc.mem_usage=55.4\n", + "runtime.step=8250, runtime.steptime=0.2699, runtime.voxel_gene_time=0.001533, runtime.prep_time=0.03905, loss.cls_loss=0.1334, loss.cls_loss_rt=0.1398, loss.loc_loss=0.2451, loss.loc_loss_rt=0.2376, loss.loc_elem=[0.005546, 0.004345, 0.0268, 0.01756, 0.02288, 0.0224, 0.01924], loss.cls_pos_rt=0.09146, loss.cls_neg_rt=0.04835, loss.dir_rt=0.1432, rpn_acc=0.9994, pr.prec@10=0.2017, pr.rec@10=0.9489, pr.prec@30=0.7565, pr.rec@30=0.8035, pr.prec@50=0.9449, pr.rec@50=0.6207, pr.prec@70=0.9924, pr.rec@70=0.3696, pr.prec@80=0.9988, pr.rec@80=0.1837, pr.prec@90=1.0, pr.rec@90=0.009159, pr.prec@95=1.0, pr.rec@95=4.11e-05, misc.num_vox=178278, misc.num_pos=65, misc.num_neg=42092, misc.num_anchors=42240, misc.lr=0.002655, misc.mem_usage=55.4\n", + "runtime.step=8300, runtime.steptime=0.2692, runtime.voxel_gene_time=0.0009396, runtime.prep_time=0.03178, loss.cls_loss=0.1328, loss.cls_loss_rt=0.1098, loss.loc_loss=0.2441, loss.loc_loss_rt=0.1813, loss.loc_elem=[0.005137, 0.004116, 0.01322, 0.0127, 0.02236, 0.02027, 0.01284], loss.cls_pos_rt=0.08435, loss.cls_neg_rt=0.02549, loss.dir_rt=0.1296, rpn_acc=0.9994, pr.prec@10=0.203, pr.rec@10=0.9486, pr.prec@30=0.7583, pr.rec@30=0.8051, pr.prec@50=0.9453, pr.rec@50=0.6225, pr.prec@70=0.9926, pr.rec@70=0.3715, pr.prec@80=0.9987, pr.rec@80=0.1859, pr.prec@90=1.0, pr.rec@90=0.01017, pr.prec@95=1.0, pr.rec@95=3.883e-05, misc.num_vox=176891, misc.num_pos=55, misc.num_neg=42112, misc.num_anchors=42240, misc.lr=0.002638, misc.mem_usage=55.4\n", + "reset Car\n", + "reset Car\n", + "reset Car\n", + "WORKER 0 seed: 1592959679\n", + "WORKER 1 seed: 1592959680\n", + "WORKER 2 seed: 1592959681\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "runtime.step=8350, runtime.steptime=0.4275, runtime.voxel_gene_time=0.001007, runtime.prep_time=0.02782, loss.cls_loss=0.1551, loss.cls_loss_rt=0.1303, loss.loc_loss=0.269, loss.loc_loss_rt=0.2952, loss.loc_elem=[0.01306, 0.007339, 0.03689, 0.01697, 0.02943, 0.02172, 0.02219], loss.cls_pos_rt=0.09598, loss.cls_neg_rt=0.03436, loss.dir_rt=0.136, rpn_acc=0.9993, pr.prec@10=0.166, pr.rec@10=0.9434, pr.prec@30=0.747, pr.rec@30=0.771, pr.prec@50=0.9363, pr.rec@50=0.5669, pr.prec@70=0.9888, pr.rec@70=0.3042, pr.prec@80=1.0, pr.rec@80=0.1251, pr.prec@90=1.0, pr.rec@90=0.004264, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=179271, misc.num_pos=73, misc.num_neg=42052, misc.num_anchors=42240, misc.lr=0.002622, misc.mem_usage=55.2\n", + "runtime.step=8400, runtime.steptime=0.265, runtime.voxel_gene_time=0.001473, runtime.prep_time=0.03732, loss.cls_loss=0.1393, loss.cls_loss_rt=0.1323, loss.loc_loss=0.2573, loss.loc_loss_rt=0.2188, loss.loc_elem=[0.004735, 0.004623, 0.02096, 0.01411, 0.02216, 0.01743, 0.02537], loss.cls_pos_rt=0.09673, loss.cls_neg_rt=0.03561, loss.dir_rt=0.175, rpn_acc=0.9994, pr.prec@10=0.1949, pr.rec@10=0.9439, pr.prec@30=0.7526, pr.rec@30=0.796, pr.prec@50=0.948, pr.rec@50=0.6094, pr.prec@70=0.9924, pr.rec@70=0.3513, pr.prec@80=0.9992, pr.rec@80=0.1621, pr.prec@90=1.0, pr.rec@90=0.005875, pr.prec@95=1.0, pr.rec@95=5.154e-05, misc.num_vox=173712, misc.num_pos=41, misc.num_neg=42128, misc.num_anchors=42240, misc.lr=0.002605, misc.mem_usage=55.5\n", + "runtime.step=8450, runtime.steptime=0.267, runtime.voxel_gene_time=0.0009604, runtime.prep_time=0.02927, loss.cls_loss=0.1369, loss.cls_loss_rt=0.1078, loss.loc_loss=0.2493, loss.loc_loss_rt=0.2215, loss.loc_elem=[0.005755, 0.005272, 0.01536, 0.01481, 0.03237, 0.01851, 0.01866], loss.cls_pos_rt=0.07539, loss.cls_neg_rt=0.03241, loss.dir_rt=0.1832, rpn_acc=0.9994, pr.prec@10=0.2014, pr.rec@10=0.9446, pr.prec@30=0.7543, pr.rec@30=0.801, pr.prec@50=0.9477, pr.rec@50=0.6161, pr.prec@70=0.993, pr.rec@70=0.3612, pr.prec@80=0.9989, pr.rec@80=0.173, pr.prec@90=1.0, pr.rec@90=0.006799, pr.prec@95=1.0, pr.rec@95=2.753e-05, misc.num_vox=177429, misc.num_pos=61, misc.num_neg=42091, misc.num_anchors=42240, misc.lr=0.002588, misc.mem_usage=55.6\n", + "runtime.step=8500, runtime.steptime=0.2663, runtime.voxel_gene_time=0.0009782, runtime.prep_time=0.03243, loss.cls_loss=0.1338, loss.cls_loss_rt=0.0921, loss.loc_loss=0.2456, loss.loc_loss_rt=0.1784, loss.loc_elem=[0.003941, 0.004001, 0.01876, 0.01257, 0.02193, 0.01641, 0.01157], loss.cls_pos_rt=0.05974, loss.cls_neg_rt=0.03237, loss.dir_rt=0.1514, rpn_acc=0.9994, pr.prec@10=0.2048, pr.rec@10=0.9474, pr.prec@30=0.7581, pr.rec@30=0.8046, pr.prec@50=0.9476, pr.rec@50=0.6218, pr.prec@70=0.9934, pr.rec@70=0.3696, pr.prec@80=0.9991, pr.rec@80=0.1838, pr.prec@90=1.0, pr.rec@90=0.01003, pr.prec@95=1.0, pr.rec@95=6.565e-05, misc.num_vox=171922, misc.num_pos=75, misc.num_neg=42054, misc.num_anchors=42240, misc.lr=0.00257, misc.mem_usage=55.4\n", + "runtime.step=8550, runtime.steptime=0.2655, runtime.voxel_gene_time=0.00103, runtime.prep_time=0.0359, loss.cls_loss=0.1329, loss.cls_loss_rt=0.1446, loss.loc_loss=0.2438, loss.loc_loss_rt=0.2615, loss.loc_elem=[0.006645, 0.004866, 0.02001, 0.01601, 0.02295, 0.02612, 0.03417], loss.cls_pos_rt=0.1201, loss.cls_neg_rt=0.02454, loss.dir_rt=0.1871, rpn_acc=0.9995, pr.prec@10=0.2057, pr.rec@10=0.9476, pr.prec@30=0.7592, pr.rec@30=0.8058, pr.prec@50=0.9479, pr.rec@50=0.6244, pr.prec@70=0.9933, pr.rec@70=0.3735, pr.prec@80=0.9991, pr.rec@80=0.1867, pr.prec@90=1.0, pr.rec@90=0.01062, pr.prec@95=1.0, pr.rec@95=6.421e-05, misc.num_vox=179339, misc.num_pos=64, misc.num_neg=42097, misc.num_anchors=42240, misc.lr=0.002552, misc.mem_usage=55.4\n", + "runtime.step=8600, runtime.steptime=0.2643, runtime.voxel_gene_time=0.001473, runtime.prep_time=0.04164, loss.cls_loss=0.1329, loss.cls_loss_rt=0.1376, loss.loc_loss=0.2431, loss.loc_loss_rt=0.2216, loss.loc_elem=[0.005512, 0.004107, 0.0215, 0.01244, 0.0252, 0.01945, 0.0226], loss.cls_pos_rt=0.09855, loss.cls_neg_rt=0.03906, loss.dir_rt=0.2347, rpn_acc=0.9994, pr.prec@10=0.2061, pr.rec@10=0.9478, pr.prec@30=0.7581, pr.rec@30=0.8058, pr.prec@50=0.9475, pr.rec@50=0.624, pr.prec@70=0.9931, pr.rec@70=0.3712, pr.prec@80=0.9988, pr.rec@80=0.1854, pr.prec@90=1.0, pr.rec@90=0.01013, pr.prec@95=1.0, pr.rec@95=5.155e-05, misc.num_vox=179523, misc.num_pos=54, misc.num_neg=42109, misc.num_anchors=42240, misc.lr=0.002534, misc.mem_usage=55.4\n", + "reset Car\n", + "reset Car\n", + "reset Car\n", + "runtime.step=8650, runtime.steptime=0.2663, runtime.voxel_gene_time=0.001, runtime.prep_time=0.03485, loss.cls_loss=0.1311, loss.cls_loss_rt=0.1195, loss.loc_loss=0.2406, loss.loc_loss_rt=0.2032, loss.loc_elem=[0.005716, 0.003578, 0.01981, 0.01123, 0.0205, 0.01797, 0.02282], loss.cls_pos_rt=0.08475, loss.cls_neg_rt=0.03479, loss.dir_rt=0.1189, rpn_acc=0.9995, pr.prec@10=0.2078, pr.rec@10=0.9492, pr.prec@30=0.7609, pr.rec@30=0.8074, pr.prec@50=0.9476, pr.rec@50=0.6272, pr.prec@70=0.9932, pr.rec@70=0.3759, pr.prec@80=0.9989, pr.rec@80=0.191, pr.prec@90=1.0, pr.rec@90=0.01151, pr.prec@95=1.0, pr.rec@95=4.796e-05, misc.num_vox=178598, misc.num_pos=48, misc.num_neg=42121, misc.num_anchors=42240, misc.lr=0.002515, misc.mem_usage=55.5\n", + "WORKER 0 seed: 1592959769\n", + "WORKER 1 seed: 1592959770\n", + "WORKER 2 seed: 1592959771\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "runtime.step=8700, runtime.steptime=0.4287, runtime.voxel_gene_time=0.000993, runtime.prep_time=0.02795, loss.cls_loss=0.1317, loss.cls_loss_rt=0.1139, loss.loc_loss=0.2414, loss.loc_loss_rt=0.2318, loss.loc_elem=[0.005301, 0.003543, 0.02444, 0.01491, 0.02022, 0.02179, 0.02568], loss.cls_pos_rt=0.08232, loss.cls_neg_rt=0.03161, loss.dir_rt=0.1232, rpn_acc=0.9995, pr.prec@10=0.2017, pr.rec@10=0.9474, pr.prec@30=0.7633, pr.rec@30=0.8041, pr.prec@50=0.946, pr.rec@50=0.625, pr.prec@70=0.9936, pr.rec@70=0.3788, pr.prec@80=0.9989, pr.rec@80=0.2039, pr.prec@90=1.0, pr.rec@90=0.01408, pr.prec@95=1.0, pr.rec@95=3.088e-05, misc.num_vox=177272, misc.num_pos=42, misc.num_neg=42152, misc.num_anchors=42240, misc.lr=0.002497, misc.mem_usage=55.4\n", + "runtime.step=8750, runtime.steptime=0.2674, runtime.voxel_gene_time=0.001325, runtime.prep_time=0.03448, loss.cls_loss=0.1301, loss.cls_loss_rt=0.1358, loss.loc_loss=0.2376, loss.loc_loss_rt=0.2343, loss.loc_elem=[0.004812, 0.004267, 0.02096, 0.01367, 0.02412, 0.01649, 0.03283], loss.cls_pos_rt=0.1013, loss.cls_neg_rt=0.03445, loss.dir_rt=0.1342, rpn_acc=0.9995, pr.prec@10=0.2075, pr.rec@10=0.949, pr.prec@30=0.7647, pr.rec@30=0.8089, pr.prec@50=0.9463, pr.rec@50=0.63, pr.prec@70=0.993, pr.rec@70=0.3856, pr.prec@80=0.9987, pr.rec@80=0.2045, pr.prec@90=1.0, pr.rec@90=0.01366, pr.prec@95=1.0, pr.rec@95=3.018e-05, misc.num_vox=179162, misc.num_pos=57, misc.num_neg=42102, misc.num_anchors=42240, misc.lr=0.002477, misc.mem_usage=55.4\n", + "runtime.step=8800, runtime.steptime=0.2656, runtime.voxel_gene_time=0.0009766, runtime.prep_time=0.02533, loss.cls_loss=0.1312, loss.cls_loss_rt=0.1432, loss.loc_loss=0.2393, loss.loc_loss_rt=0.2247, loss.loc_elem=[0.003719, 0.006456, 0.02215, 0.01359, 0.0224, 0.01741, 0.02664], loss.cls_pos_rt=0.0923, loss.cls_neg_rt=0.0509, loss.dir_rt=0.1781, rpn_acc=0.9995, pr.prec@10=0.2051, pr.rec@10=0.9492, pr.prec@30=0.7641, pr.rec@30=0.8068, pr.prec@50=0.9466, pr.rec@50=0.6278, pr.prec@70=0.9924, pr.rec@70=0.3843, pr.prec@80=0.9986, pr.rec@80=0.2033, pr.prec@90=1.0, pr.rec@90=0.0142, pr.prec@95=1.0, pr.rec@95=2.997e-05, misc.num_vox=175149, misc.num_pos=63, misc.num_neg=42084, misc.num_anchors=42240, misc.lr=0.002458, misc.mem_usage=55.4\n", + "runtime.step=8850, runtime.steptime=0.2663, runtime.voxel_gene_time=0.0008874, runtime.prep_time=0.02414, loss.cls_loss=0.1305, loss.cls_loss_rt=0.1138, loss.loc_loss=0.2386, loss.loc_loss_rt=0.2137, loss.loc_elem=[0.00511, 0.003584, 0.01997, 0.01396, 0.01977, 0.0231, 0.02133], loss.cls_pos_rt=0.06376, loss.cls_neg_rt=0.05008, loss.dir_rt=0.1565, rpn_acc=0.9995, pr.prec@10=0.2075, pr.rec@10=0.9493, pr.prec@30=0.7626, pr.rec@30=0.8082, pr.prec@50=0.9465, pr.rec@50=0.6298, pr.prec@70=0.993, pr.rec@70=0.3851, pr.prec@80=0.9988, pr.rec@80=0.2036, pr.prec@90=1.0, pr.rec@90=0.01416, pr.prec@95=1.0, pr.rec@95=2.242e-05, misc.num_vox=177044, misc.num_pos=55, misc.num_neg=42088, misc.num_anchors=42240, misc.lr=0.002438, misc.mem_usage=55.4\n", + "runtime.step=8900, runtime.steptime=0.2656, runtime.voxel_gene_time=0.001331, runtime.prep_time=0.03326, loss.cls_loss=0.1301, loss.cls_loss_rt=0.1317, loss.loc_loss=0.2376, loss.loc_loss_rt=0.2279, loss.loc_elem=[0.005101, 0.003697, 0.02143, 0.01747, 0.01932, 0.02359, 0.02335], loss.cls_pos_rt=0.1013, loss.cls_neg_rt=0.03048, loss.dir_rt=0.134, rpn_acc=0.9995, pr.prec@10=0.2089, pr.rec@10=0.9496, pr.prec@30=0.7619, pr.rec@30=0.8087, pr.prec@50=0.9467, pr.rec@50=0.6298, pr.prec@70=0.9929, pr.rec@70=0.3839, pr.prec@80=0.9988, pr.rec@80=0.2023, pr.prec@90=1.0, pr.rec@90=0.0138, pr.prec@95=1.0, pr.rec@95=1.791e-05, misc.num_vox=179846, misc.num_pos=58, misc.num_neg=42103, misc.num_anchors=42240, misc.lr=0.002419, misc.mem_usage=55.4\n", + "reset Car\n", + "reset Car\n", + "runtime.step=8950, runtime.steptime=0.2654, runtime.voxel_gene_time=0.00102, runtime.prep_time=0.03191, loss.cls_loss=0.129, loss.cls_loss_rt=0.1566, loss.loc_loss=0.2358, loss.loc_loss_rt=0.254, loss.loc_elem=[0.005066, 0.004396, 0.02712, 0.01535, 0.03446, 0.01958, 0.02103], loss.cls_pos_rt=0.1028, loss.cls_neg_rt=0.0538, loss.dir_rt=0.2663, rpn_acc=0.9995, pr.prec@10=0.2105, pr.rec@10=0.9502, pr.prec@30=0.7637, pr.rec@30=0.8104, pr.prec@50=0.9474, pr.rec@50=0.6333, pr.prec@70=0.9929, pr.rec@70=0.3872, pr.prec@80=0.9986, pr.rec@80=0.205, pr.prec@90=1.0, pr.rec@90=0.01446, pr.prec@95=1.0, pr.rec@95=2.983e-05, misc.num_vox=177289, misc.num_pos=53, misc.num_neg=42117, misc.num_anchors=42240, misc.lr=0.002398, misc.mem_usage=55.5\n", + "reset Car\n", + "WORKER 0 seed: 1592959859\n", + "WORKER 1 seed: 1592959860\n", + "WORKER 2 seed: 1592959861\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "runtime.step=9000, runtime.steptime=0.4269, runtime.voxel_gene_time=0.001355, runtime.prep_time=0.03395, loss.cls_loss=0.1355, loss.cls_loss_rt=0.1306, loss.loc_loss=0.2452, loss.loc_loss_rt=0.2038, loss.loc_elem=[0.004405, 0.00507, 0.02034, 0.01225, 0.02214, 0.01831, 0.0194], loss.cls_pos_rt=0.0932, loss.cls_neg_rt=0.03742, loss.dir_rt=0.1329, rpn_acc=0.9994, pr.prec@10=0.1974, pr.rec@10=0.9461, pr.prec@30=0.7558, pr.rec@30=0.7974, pr.prec@50=0.9474, pr.rec@50=0.6176, pr.prec@70=0.9935, pr.rec@70=0.3735, pr.prec@80=0.9998, pr.rec@80=0.1962, pr.prec@90=1.0, pr.rec@90=0.0138, pr.prec@95=1.0, pr.rec@95=3.782e-05, misc.num_vox=178232, misc.num_pos=59, misc.num_neg=42099, misc.num_anchors=42240, misc.lr=0.002378, misc.mem_usage=55.4\n", + "runtime.step=9050, runtime.steptime=0.2682, runtime.voxel_gene_time=0.0009632, runtime.prep_time=0.0295, loss.cls_loss=0.1325, loss.cls_loss_rt=0.1718, loss.loc_loss=0.2416, loss.loc_loss_rt=0.2787, loss.loc_elem=[0.006094, 0.005002, 0.02777, 0.01602, 0.02522, 0.02189, 0.03736], loss.cls_pos_rt=0.1195, loss.cls_neg_rt=0.05233, loss.dir_rt=0.1893, rpn_acc=0.9995, pr.prec@10=0.2058, pr.rec@10=0.9476, pr.prec@30=0.7601, pr.rec@30=0.8075, pr.prec@50=0.9482, pr.rec@50=0.6288, pr.prec@70=0.993, pr.rec@70=0.3821, pr.prec@80=0.9992, pr.rec@80=0.2011, pr.prec@90=1.0, pr.rec@90=0.01587, pr.prec@95=1.0, pr.rec@95=3.316e-05, misc.num_vox=177085, misc.num_pos=65, misc.num_neg=42090, misc.num_anchors=42240, misc.lr=0.002357, misc.mem_usage=55.4\n", + "runtime.step=9100, runtime.steptime=0.2694, runtime.voxel_gene_time=0.001013, runtime.prep_time=0.02939, loss.cls_loss=0.131, loss.cls_loss_rt=0.134, loss.loc_loss=0.2388, loss.loc_loss_rt=0.2836, loss.loc_elem=[0.006293, 0.004763, 0.03089, 0.01343, 0.02731, 0.02246, 0.03667], loss.cls_pos_rt=0.09616, loss.cls_neg_rt=0.03788, loss.dir_rt=0.223, rpn_acc=0.9995, pr.prec@10=0.2092, pr.rec@10=0.9485, pr.prec@30=0.7654, pr.rec@30=0.8086, pr.prec@50=0.9489, pr.rec@50=0.6318, pr.prec@70=0.9933, pr.rec@70=0.3854, pr.prec@80=0.9989, pr.rec@80=0.2053, pr.prec@90=1.0, pr.rec@90=0.01578, pr.prec@95=1.0, pr.rec@95=4.25e-05, misc.num_vox=179148, misc.num_pos=58, misc.num_neg=42082, misc.num_anchors=42240, misc.lr=0.002336, misc.mem_usage=55.4\n", + "runtime.step=9150, runtime.steptime=0.2689, runtime.voxel_gene_time=0.001562, runtime.prep_time=0.03711, loss.cls_loss=0.1287, loss.cls_loss_rt=0.1449, loss.loc_loss=0.2353, loss.loc_loss_rt=0.2571, loss.loc_elem=[0.005172, 0.004318, 0.02788, 0.01214, 0.0228, 0.02417, 0.03208], loss.cls_pos_rt=0.1089, loss.cls_neg_rt=0.03598, loss.dir_rt=0.2037, rpn_acc=0.9995, pr.prec@10=0.2134, pr.rec@10=0.9498, pr.prec@30=0.7663, pr.rec@30=0.8125, pr.prec@50=0.9491, pr.rec@50=0.6366, pr.prec@70=0.9931, pr.rec@70=0.3923, pr.prec@80=0.9986, pr.rec@80=0.2116, pr.prec@90=1.0, pr.rec@90=0.01738, pr.prec@95=1.0, pr.rec@95=5.486e-05, misc.num_vox=178765, misc.num_pos=59, misc.num_neg=42085, misc.num_anchors=42240, misc.lr=0.002315, misc.mem_usage=55.4\n", + "runtime.step=9200, runtime.steptime=0.2664, runtime.voxel_gene_time=0.001189, runtime.prep_time=0.03149, loss.cls_loss=0.1283, loss.cls_loss_rt=0.1141, loss.loc_loss=0.235, loss.loc_loss_rt=0.2292, loss.loc_elem=[0.005255, 0.004428, 0.02249, 0.01597, 0.02765, 0.02094, 0.01786], loss.cls_pos_rt=0.08106, loss.cls_neg_rt=0.03301, loss.dir_rt=0.1377, rpn_acc=0.9995, pr.prec@10=0.2133, pr.rec@10=0.9504, pr.prec@30=0.7665, pr.rec@30=0.8126, pr.prec@50=0.9482, pr.rec@50=0.6368, pr.prec@70=0.9932, pr.rec@70=0.3929, pr.prec@80=0.9986, pr.rec@80=0.2127, pr.prec@90=0.9997, pr.rec@90=0.01837, pr.prec@95=1.0, pr.rec@95=4.965e-05, misc.num_vox=178110, misc.num_pos=64, misc.num_neg=42086, misc.num_anchors=42240, misc.lr=0.002294, misc.mem_usage=55.4\n", + "runtime.step=9250, runtime.steptime=0.2697, runtime.voxel_gene_time=0.0009308, runtime.prep_time=0.02407, loss.cls_loss=0.128, loss.cls_loss_rt=0.1468, loss.loc_loss=0.2343, loss.loc_loss_rt=0.2548, loss.loc_elem=[0.006257, 0.004051, 0.03344, 0.01219, 0.02615, 0.02036, 0.02494], loss.cls_pos_rt=0.1126, loss.cls_neg_rt=0.03424, loss.dir_rt=0.154, rpn_acc=0.9995, pr.prec@10=0.2141, pr.rec@10=0.9503, pr.prec@30=0.7665, pr.rec@30=0.8125, pr.prec@50=0.9487, pr.rec@50=0.6375, pr.prec@70=0.9935, pr.rec@70=0.3953, pr.prec@80=0.9987, pr.rec@80=0.215, pr.prec@90=0.9997, pr.rec@90=0.01882, pr.prec@95=1.0, pr.rec@95=5.123e-05, misc.num_vox=173496, misc.num_pos=52, misc.num_neg=42118, misc.num_anchors=42240, misc.lr=0.002272, misc.mem_usage=55.6\n", + "reset Car\n", + "reset Car\n", + "reset Car\n", + "WORKER 0 seed: 1592959950\n", + "WORKER 1 seed: 1592959951\n", + "WORKER 2 seed: 1592959952\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "runtime.step=9300, runtime.steptime=0.4277, runtime.voxel_gene_time=0.00136, runtime.prep_time=0.02955, loss.cls_loss=0.1323, loss.cls_loss_rt=0.1324, loss.loc_loss=0.2401, loss.loc_loss_rt=0.2084, loss.loc_elem=[0.004551, 0.005078, 0.02168, 0.01287, 0.02097, 0.01886, 0.02021], loss.cls_pos_rt=0.08976, loss.cls_neg_rt=0.04261, loss.dir_rt=0.1408, rpn_acc=0.9995, pr.prec@10=0.2005, pr.rec@10=0.9539, pr.prec@30=0.7642, pr.rec@30=0.807, pr.prec@50=0.9473, pr.rec@50=0.6264, pr.prec@70=0.9869, pr.rec@70=0.3824, pr.prec@80=0.9942, pr.rec@80=0.1983, pr.prec@90=1.0, pr.rec@90=0.01305, pr.prec@95=1.0, pr.rec@95=5.038e-05, misc.num_vox=177894, misc.num_pos=45, misc.num_neg=42130, misc.num_anchors=42240, misc.lr=0.00225, misc.mem_usage=55.3\n", + "#################################\n", + "# EVAL\n", + "#################################\n", + "Generate output labels...\n", + "[100.0%][===================>][12.32it/s][00:26>00:00] \n", + "generate label finished(141.92/s). start eval:\n", + "Evaluation official\n", + "Car AP(Average Precision)@0.70, 0.70, 0.70:\n", + "bbox AP:89.12, 85.46, 78.86\n", + "bev AP:89.36, 85.78, 79.36\n", + "3d AP:72.54, 63.46, 57.00\n", + "aos AP:0.38, 1.78, 2.45\n", + "Car AP(Average Precision)@0.70, 0.50, 0.50:\n", + "bbox AP:89.12, 85.46, 78.86\n", + "bev AP:90.51, 89.21, 88.69\n", + "3d AP:90.44, 88.84, 88.01\n", + "aos AP:0.38, 1.78, 2.45\n", + "\n", + "Evaluation coco\n", + "Car coco AP@0.50:0.05:0.95:\n", + "bbox AP:66.68, 62.67, 60.92\n", + "bev AP:65.50, 62.53, 61.21\n", + "3d AP:50.35, 46.37, 44.99\n", + "aos AP:0.28, 1.30, 2.01\n", + "\n", + "eval.kitti.official.Car.bbox@0.70=[89.12, 85.46, 78.86], eval.kitti.official.Car.bev@0.70=[89.36, 85.78, 79.36], eval.kitti.official.Car.3d@0.70=[72.54, 63.46, 57.0], eval.kitti.official.Car.aos=[0.375, 1.78, 2.452], eval.kitti.official.Car.bev@0.50=[90.51, 89.21, 88.69], eval.kitti.official.Car.3d@0.50=[90.44, 88.84, 88.01], eval.kitti.coco.Car.bbox=[66.68, 62.67, 60.92], eval.kitti.coco.Car.bev=[65.5, 62.53, 61.21], eval.kitti.coco.Car.3d=[50.35, 46.37, 44.99], eval.kitti.coco.Car.aos=[0.2771, 1.305, 2.01]\n", + "runtime.step=9350, runtime.steptime=1.034, runtime.voxel_gene_time=0.001412, runtime.prep_time=0.04092, loss.cls_loss=0.1319, loss.cls_loss_rt=0.111, loss.loc_loss=0.2384, loss.loc_loss_rt=0.2344, loss.loc_elem=[0.005504, 0.004724, 0.01919, 0.01522, 0.02577, 0.02194, 0.02483], loss.cls_pos_rt=0.06712, loss.cls_neg_rt=0.04386, loss.dir_rt=0.1927, rpn_acc=0.9995, pr.prec@10=0.2044, pr.rec@10=0.9501, pr.prec@30=0.7641, pr.rec@30=0.8058, pr.prec@50=0.9477, pr.rec@50=0.6291, pr.prec@70=0.9916, pr.rec@70=0.3775, pr.prec@80=0.9971, pr.rec@80=0.1849, pr.prec@90=1.0, pr.rec@90=0.007672, pr.prec@95=1.0, pr.rec@95=1.853e-05, misc.num_vox=177416, misc.num_pos=51, misc.num_neg=42110, misc.num_anchors=42240, misc.lr=0.002228, misc.mem_usage=55.4\n", + "runtime.step=9400, runtime.steptime=0.265, runtime.voxel_gene_time=0.001101, runtime.prep_time=0.03554, loss.cls_loss=0.1291, loss.cls_loss_rt=0.1338, loss.loc_loss=0.2362, loss.loc_loss_rt=0.2234, loss.loc_elem=[0.005617, 0.003561, 0.0219, 0.01506, 0.02332, 0.02114, 0.0211], loss.cls_pos_rt=0.1031, loss.cls_neg_rt=0.0307, loss.dir_rt=0.1511, rpn_acc=0.9995, pr.prec@10=0.2099, pr.rec@10=0.9503, pr.prec@30=0.7669, pr.rec@30=0.8105, pr.prec@50=0.9494, pr.rec@50=0.6366, pr.prec@70=0.992, pr.rec@70=0.3889, pr.prec@80=0.9981, pr.rec@80=0.1998, pr.prec@90=1.0, pr.rec@90=0.009931, pr.prec@95=1.0, pr.rec@95=1.143e-05, misc.num_vox=178594, misc.num_pos=56, misc.num_neg=42096, misc.num_anchors=42240, misc.lr=0.002206, misc.mem_usage=55.4\n", + "runtime.step=9450, runtime.steptime=0.2623, runtime.voxel_gene_time=0.001225, runtime.prep_time=0.03428, loss.cls_loss=0.1276, loss.cls_loss_rt=0.1478, loss.loc_loss=0.2338, loss.loc_loss_rt=0.249, loss.loc_elem=[0.005794, 0.004837, 0.03231, 0.01261, 0.02078, 0.02218, 0.02602], loss.cls_pos_rt=0.1032, loss.cls_neg_rt=0.04456, loss.dir_rt=0.2099, rpn_acc=0.9995, pr.prec@10=0.2135, pr.rec@10=0.9515, pr.prec@30=0.7684, pr.rec@30=0.8131, pr.prec@50=0.95, pr.rec@50=0.6396, pr.prec@70=0.9927, pr.rec@70=0.3948, pr.prec@80=0.9985, pr.rec@80=0.2062, pr.prec@90=1.0, pr.rec@90=0.01074, pr.prec@95=1.0, pr.rec@95=1.65e-05, misc.num_vox=175211, misc.num_pos=65, misc.num_neg=42088, misc.num_anchors=42240, misc.lr=0.002184, misc.mem_usage=55.5\n", + "runtime.step=9500, runtime.steptime=0.2653, runtime.voxel_gene_time=0.000989, runtime.prep_time=0.03051, loss.cls_loss=0.128, loss.cls_loss_rt=0.1305, loss.loc_loss=0.2342, loss.loc_loss_rt=0.2397, loss.loc_elem=[0.006274, 0.004325, 0.02728, 0.01445, 0.02462, 0.01689, 0.02601], loss.cls_pos_rt=0.095, loss.cls_neg_rt=0.03548, loss.dir_rt=0.1603, rpn_acc=0.9995, pr.prec@10=0.2127, pr.rec@10=0.9508, pr.prec@30=0.7678, pr.rec@30=0.813, pr.prec@50=0.9498, pr.rec@50=0.6383, pr.prec@70=0.9924, pr.rec@70=0.3948, pr.prec@80=0.9986, pr.rec@80=0.2094, pr.prec@90=1.0, pr.rec@90=0.0127, pr.prec@95=1.0, pr.rec@95=7.098e-05, misc.num_vox=176806, misc.num_pos=56, misc.num_neg=42105, misc.num_anchors=42240, misc.lr=0.002161, misc.mem_usage=55.5\n", + "runtime.step=9550, runtime.steptime=0.264, runtime.voxel_gene_time=0.0008738, runtime.prep_time=0.02815, loss.cls_loss=0.127, loss.cls_loss_rt=0.1434, loss.loc_loss=0.2319, loss.loc_loss_rt=0.2351, loss.loc_elem=[0.006306, 0.003715, 0.01896, 0.01525, 0.02416, 0.02365, 0.0255], loss.cls_pos_rt=0.1029, loss.cls_neg_rt=0.04048, loss.dir_rt=0.1574, rpn_acc=0.9995, pr.prec@10=0.2157, pr.rec@10=0.9514, pr.prec@30=0.7693, pr.rec@30=0.8142, pr.prec@50=0.9499, pr.rec@50=0.6408, pr.prec@70=0.9929, pr.rec@70=0.3997, pr.prec@80=0.9987, pr.rec@80=0.2137, pr.prec@90=1.0, pr.rec@90=0.01429, pr.prec@95=1.0, pr.rec@95=6.879e-05, misc.num_vox=178202, misc.num_pos=52, misc.num_neg=42101, misc.num_anchors=42240, misc.lr=0.002138, misc.mem_usage=55.5\n", + "reset Car\n", + "reset Car\n", + "reset Car\n", + "WORKER 0 seed: 1592960079\n", + "WORKER 1 seed: 1592960080\n", + "WORKER 2 seed: 1592960081\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "runtime.step=9600, runtime.steptime=0.4271, runtime.voxel_gene_time=0.001092, runtime.prep_time=0.02954, loss.cls_loss=0.1288, loss.cls_loss_rt=0.1332, loss.loc_loss=0.236, loss.loc_loss_rt=0.2371, loss.loc_elem=[0.004849, 0.006008, 0.0194, 0.02071, 0.02536, 0.02098, 0.02125], loss.cls_pos_rt=0.1059, loss.cls_neg_rt=0.02733, loss.dir_rt=0.1669, rpn_acc=0.9995, pr.prec@10=0.2075, pr.rec@10=0.9513, pr.prec@30=0.7713, pr.rec@30=0.8096, pr.prec@50=0.9523, pr.rec@50=0.6351, pr.prec@70=0.9912, pr.rec@70=0.4041, pr.prec@80=0.9991, pr.rec@80=0.2337, pr.prec@90=1.0, pr.rec@90=0.03008, pr.prec@95=1.0, pr.rec@95=0.000634, misc.num_vox=177791, misc.num_pos=76, misc.num_neg=42067, misc.num_anchors=42240, misc.lr=0.002115, misc.mem_usage=55.4\n", + "runtime.step=9650, runtime.steptime=0.2687, runtime.voxel_gene_time=0.001084, runtime.prep_time=0.03362, loss.cls_loss=0.1273, loss.cls_loss_rt=0.1219, loss.loc_loss=0.2324, loss.loc_loss_rt=0.2112, loss.loc_elem=[0.005388, 0.005046, 0.016, 0.01466, 0.02353, 0.01755, 0.02344], loss.cls_pos_rt=0.08996, loss.cls_neg_rt=0.03193, loss.dir_rt=0.1195, rpn_acc=0.9995, pr.prec@10=0.2118, pr.rec@10=0.9516, pr.prec@30=0.7736, pr.rec@30=0.8136, pr.prec@50=0.9503, pr.rec@50=0.6389, pr.prec@70=0.9922, pr.rec@70=0.402, pr.prec@80=0.9992, pr.rec@80=0.2283, pr.prec@90=1.0, pr.rec@90=0.02405, pr.prec@95=1.0, pr.rec@95=0.0002299, misc.num_vox=177109, misc.num_pos=56, misc.num_neg=42087, misc.num_anchors=42240, misc.lr=0.002092, misc.mem_usage=55.4\n", + "runtime.step=9700, runtime.steptime=0.266, runtime.voxel_gene_time=0.001525, runtime.prep_time=0.03928, loss.cls_loss=0.1266, loss.cls_loss_rt=0.1203, loss.loc_loss=0.2314, loss.loc_loss_rt=0.2438, loss.loc_elem=[0.004653, 0.004378, 0.02216, 0.01517, 0.02697, 0.02315, 0.02543], loss.cls_pos_rt=0.08522, loss.cls_neg_rt=0.03507, loss.dir_rt=0.1866, rpn_acc=0.9995, pr.prec@10=0.2165, pr.rec@10=0.9512, pr.prec@30=0.7728, pr.rec@30=0.8169, pr.prec@50=0.9482, pr.rec@50=0.6433, pr.prec@70=0.9918, pr.rec@70=0.4074, pr.prec@80=0.9987, pr.rec@80=0.2321, pr.prec@90=1.0, pr.rec@90=0.02354, pr.prec@95=1.0, pr.rec@95=0.000159, misc.num_vox=174954, misc.num_pos=50, misc.num_neg=42108, misc.num_anchors=42240, misc.lr=0.002069, misc.mem_usage=55.4\n", + "runtime.step=9750, runtime.steptime=0.2684, runtime.voxel_gene_time=0.00085, runtime.prep_time=0.02671, loss.cls_loss=0.1256, loss.cls_loss_rt=0.1329, loss.loc_loss=0.2304, loss.loc_loss_rt=0.2357, loss.loc_elem=[0.005105, 0.005342, 0.02111, 0.01183, 0.02536, 0.01727, 0.03184], loss.cls_pos_rt=0.09971, loss.cls_neg_rt=0.03318, loss.dir_rt=0.1902, rpn_acc=0.9995, pr.prec@10=0.2174, pr.rec@10=0.9517, pr.prec@30=0.7728, pr.rec@30=0.8172, pr.prec@50=0.9485, pr.rec@50=0.6441, pr.prec@70=0.9924, pr.rec@70=0.4073, pr.prec@80=0.9988, pr.rec@80=0.2313, pr.prec@90=1.0, pr.rec@90=0.02439, pr.prec@95=1.0, pr.rec@95=0.0002425, misc.num_vox=177861, misc.num_pos=72, misc.num_neg=42070, misc.num_anchors=42240, misc.lr=0.002045, misc.mem_usage=55.5\n", + "runtime.step=9800, runtime.steptime=0.2698, runtime.voxel_gene_time=0.001009, runtime.prep_time=0.02357, loss.cls_loss=0.1252, loss.cls_loss_rt=0.1118, loss.loc_loss=0.2294, loss.loc_loss_rt=0.2099, loss.loc_elem=[0.005037, 0.003941, 0.01958, 0.01313, 0.0234, 0.02281, 0.01702], loss.cls_pos_rt=0.07676, loss.cls_neg_rt=0.03503, loss.dir_rt=0.1064, rpn_acc=0.9995, pr.prec@10=0.218, pr.rec@10=0.9518, pr.prec@30=0.7719, pr.rec@30=0.8171, pr.prec@50=0.9488, pr.rec@50=0.6441, pr.prec@70=0.9924, pr.rec@70=0.4076, pr.prec@80=0.9988, pr.rec@80=0.2312, pr.prec@90=1.0, pr.rec@90=0.02492, pr.prec@95=1.0, pr.rec@95=0.0002274, misc.num_vox=179373, misc.num_pos=56, misc.num_neg=42098, misc.num_anchors=42240, misc.lr=0.002021, misc.mem_usage=55.5\n", + "runtime.step=9850, runtime.steptime=0.2677, runtime.voxel_gene_time=0.0009673, runtime.prep_time=0.03248, loss.cls_loss=0.1248, loss.cls_loss_rt=0.1094, loss.loc_loss=0.2274, loss.loc_loss_rt=0.2331, loss.loc_elem=[0.004643, 0.004728, 0.0256, 0.01288, 0.02303, 0.02171, 0.02395], loss.cls_pos_rt=0.07447, loss.cls_neg_rt=0.03494, loss.dir_rt=0.1653, rpn_acc=0.9995, pr.prec@10=0.2184, pr.rec@10=0.9523, pr.prec@30=0.7713, pr.rec@30=0.8177, pr.prec@50=0.9484, pr.rec@50=0.6442, pr.prec@70=0.9925, pr.rec@70=0.4062, pr.prec@80=0.9988, pr.rec@80=0.2292, pr.prec@90=1.0, pr.rec@90=0.02455, pr.prec@95=1.0, pr.rec@95=0.0001855, misc.num_vox=173458, misc.num_pos=35, misc.num_neg=42150, misc.num_anchors=42240, misc.lr=0.001998, misc.mem_usage=55.5\n", + "reset Car\n", + "reset Car\n", + "reset Car\n", + "WORKER 0 seed: 1592960170\n", + "WORKER 1 seed: 1592960171\n", + "WORKER 2 seed: 1592960172\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "runtime.step=9900, runtime.steptime=0.4314, runtime.voxel_gene_time=0.001427, runtime.prep_time=0.03037, loss.cls_loss=0.1217, loss.cls_loss_rt=0.1034, loss.loc_loss=0.2297, loss.loc_loss_rt=0.2109, loss.loc_elem=[0.004461, 0.003769, 0.01986, 0.01262, 0.02469, 0.01873, 0.02134], loss.cls_pos_rt=0.07205, loss.cls_neg_rt=0.03137, loss.dir_rt=0.1371, rpn_acc=0.9995, pr.prec@10=0.2264, pr.rec@10=0.9559, pr.prec@30=0.7754, pr.rec@30=0.8196, pr.prec@50=0.9571, pr.rec@50=0.6503, pr.prec@70=0.9942, pr.rec@70=0.4045, pr.prec@80=0.9995, pr.rec@80=0.2403, pr.prec@90=1.0, pr.rec@90=0.02656, pr.prec@95=1.0, pr.rec@95=0.0004965, misc.num_vox=176343, misc.num_pos=51, misc.num_neg=42111, misc.num_anchors=42240, misc.lr=0.001974, misc.mem_usage=55.3\n", + "runtime.step=9950, runtime.steptime=0.2676, runtime.voxel_gene_time=0.0009539, runtime.prep_time=0.02823, loss.cls_loss=0.1242, loss.cls_loss_rt=0.1316, loss.loc_loss=0.2281, loss.loc_loss_rt=0.222, loss.loc_elem=[0.004867, 0.004329, 0.01863, 0.01284, 0.02276, 0.01593, 0.03165], loss.cls_pos_rt=0.09542, loss.cls_neg_rt=0.03614, loss.dir_rt=0.1776, rpn_acc=0.9995, pr.prec@10=0.2167, pr.rec@10=0.9523, pr.prec@30=0.7716, pr.rec@30=0.8144, pr.prec@50=0.9529, pr.rec@50=0.6436, pr.prec@70=0.9943, pr.rec@70=0.4115, pr.prec@80=0.9993, pr.rec@80=0.2433, pr.prec@90=1.0, pr.rec@90=0.02988, pr.prec@95=1.0, pr.rec@95=0.0002152, misc.num_vox=178316, misc.num_pos=59, misc.num_neg=42101, misc.num_anchors=42240, misc.lr=0.00195, misc.mem_usage=55.3\n", + "runtime.step=10000, runtime.steptime=0.2681, runtime.voxel_gene_time=0.001075, runtime.prep_time=0.03269, loss.cls_loss=0.1245, loss.cls_loss_rt=0.1275, loss.loc_loss=0.2275, loss.loc_loss_rt=0.245, loss.loc_elem=[0.006554, 0.004614, 0.02343, 0.01525, 0.02747, 0.02414, 0.02102], loss.cls_pos_rt=0.09349, loss.cls_neg_rt=0.03396, loss.dir_rt=0.1292, rpn_acc=0.9995, pr.prec@10=0.2189, pr.rec@10=0.9519, pr.prec@30=0.7704, pr.rec@30=0.818, pr.prec@50=0.9501, pr.rec@50=0.6473, pr.prec@70=0.9933, pr.rec@70=0.4161, pr.prec@80=0.999, pr.rec@80=0.2471, pr.prec@90=1.0, pr.rec@90=0.03452, pr.prec@95=1.0, pr.rec@95=0.0003285, misc.num_vox=175575, misc.num_pos=57, misc.num_neg=42105, misc.num_anchors=42240, misc.lr=0.001925, misc.mem_usage=55.3\n", + "runtime.step=10050, runtime.steptime=0.2646, runtime.voxel_gene_time=0.001467, runtime.prep_time=0.03534, loss.cls_loss=0.1236, loss.cls_loss_rt=0.1415, loss.loc_loss=0.2262, loss.loc_loss_rt=0.2402, loss.loc_elem=[0.00664, 0.005, 0.02237, 0.01181, 0.02029, 0.01934, 0.03466], loss.cls_pos_rt=0.1056, loss.cls_neg_rt=0.03588, loss.dir_rt=0.1312, rpn_acc=0.9995, pr.prec@10=0.2194, pr.rec@10=0.9523, pr.prec@30=0.7743, pr.rec@30=0.8195, pr.prec@50=0.9509, pr.rec@50=0.6491, pr.prec@70=0.9929, pr.rec@70=0.4159, pr.prec@80=0.9987, pr.rec@80=0.2443, pr.prec@90=1.0, pr.rec@90=0.03179, pr.prec@95=1.0, pr.rec@95=0.0002376, misc.num_vox=176829, misc.num_pos=40, misc.num_neg=42136, misc.num_anchors=42240, misc.lr=0.001901, misc.mem_usage=55.4\n", + "runtime.step=10100, runtime.steptime=0.2691, runtime.voxel_gene_time=0.0009904, runtime.prep_time=0.02516, loss.cls_loss=0.1233, loss.cls_loss_rt=0.1255, loss.loc_loss=0.2255, loss.loc_loss_rt=0.2137, loss.loc_elem=[0.005276, 0.004102, 0.01908, 0.01462, 0.02336, 0.01979, 0.02064], loss.cls_pos_rt=0.08523, loss.cls_neg_rt=0.04025, loss.dir_rt=0.1669, rpn_acc=0.9995, pr.prec@10=0.2211, pr.rec@10=0.9526, pr.prec@30=0.7733, pr.rec@30=0.82, pr.prec@50=0.9502, pr.rec@50=0.6502, pr.prec@70=0.9932, pr.rec@70=0.4153, pr.prec@80=0.9987, pr.rec@80=0.2432, pr.prec@90=1.0, pr.rec@90=0.03027, pr.prec@95=1.0, pr.rec@95=0.0002021, misc.num_vox=176745, misc.num_pos=54, misc.num_neg=42111, misc.num_anchors=42240, misc.lr=0.001876, misc.mem_usage=55.4\n", + "runtime.step=10150, runtime.steptime=0.2706, runtime.voxel_gene_time=0.001012, runtime.prep_time=0.03108, loss.cls_loss=0.1233, loss.cls_loss_rt=0.1072, loss.loc_loss=0.2251, loss.loc_loss_rt=0.2114, loss.loc_elem=[0.004304, 0.004145, 0.02118, 0.0127, 0.0231, 0.02275, 0.0175], loss.cls_pos_rt=0.07131, loss.cls_neg_rt=0.03592, loss.dir_rt=0.1246, rpn_acc=0.9995, pr.prec@10=0.2221, pr.rec@10=0.9524, pr.prec@30=0.7737, pr.rec@30=0.8208, pr.prec@50=0.9505, pr.rec@50=0.6504, pr.prec@70=0.9928, pr.rec@70=0.4138, pr.prec@80=0.9987, pr.rec@80=0.2398, pr.prec@90=1.0, pr.rec@90=0.02885, pr.prec@95=1.0, pr.rec@95=0.0001632, misc.num_vox=177504, misc.num_pos=51, misc.num_neg=42104, misc.num_anchors=42240, misc.lr=0.001852, misc.mem_usage=55.4\n", + "reset Car\n", + "reset Car\n", + "reset Car\n", + "WORKER 0 seed: 1592960261\n", + "WORKER 1 seed: 1592960262\n", + "WORKER 2 seed: 1592960263\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "runtime.step=10200, runtime.steptime=0.4366, runtime.voxel_gene_time=0.03331, runtime.prep_time=7.283, loss.cls_loss=0.1384, loss.cls_loss_rt=0.123, loss.loc_loss=0.2224, loss.loc_loss_rt=0.2073, loss.loc_elem=[0.005118, 0.002904, 0.0216, 0.01381, 0.02034, 0.0192, 0.02067], loss.cls_pos_rt=0.08467, loss.cls_neg_rt=0.03831, loss.dir_rt=0.122, rpn_acc=0.9994, pr.prec@10=0.2118, pr.rec@10=0.9421, pr.prec@30=0.7524, pr.rec@30=0.7872, pr.prec@50=0.9465, pr.rec@50=0.6115, pr.prec@70=0.9938, pr.rec@70=0.3909, pr.prec@80=1.0, pr.rec@80=0.2278, pr.prec@90=1.0, pr.rec@90=0.02824, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=175688, misc.num_pos=45, misc.num_neg=42125, misc.num_anchors=42240, misc.lr=0.001827, misc.mem_usage=55.0\n", + "runtime.step=10250, runtime.steptime=0.2694, runtime.voxel_gene_time=0.0008969, runtime.prep_time=0.02942, loss.cls_loss=0.1247, loss.cls_loss_rt=0.1323, loss.loc_loss=0.2282, loss.loc_loss_rt=0.1919, loss.loc_elem=[0.005512, 0.004596, 0.01874, 0.01275, 0.01967, 0.01791, 0.01679], loss.cls_pos_rt=0.09201, loss.cls_neg_rt=0.04024, loss.dir_rt=0.1739, rpn_acc=0.9995, pr.prec@10=0.2191, pr.rec@10=0.9504, pr.prec@30=0.7734, pr.rec@30=0.8176, pr.prec@50=0.9512, pr.rec@50=0.647, pr.prec@70=0.9956, pr.rec@70=0.4118, pr.prec@80=0.9993, pr.rec@80=0.2355, pr.prec@90=1.0, pr.rec@90=0.02862, pr.prec@95=1.0, pr.rec@95=0.0001674, misc.num_vox=174003, misc.num_pos=69, misc.num_neg=42065, misc.num_anchors=42240, misc.lr=0.001802, misc.mem_usage=55.5\n", + "runtime.step=10300, runtime.steptime=0.2684, runtime.voxel_gene_time=0.001074, runtime.prep_time=0.03134, loss.cls_loss=0.1232, loss.cls_loss_rt=0.124, loss.loc_loss=0.2235, loss.loc_loss_rt=0.2177, loss.loc_elem=[0.005506, 0.004406, 0.01721, 0.01321, 0.02976, 0.01828, 0.02049], loss.cls_pos_rt=0.08441, loss.cls_neg_rt=0.0396, loss.dir_rt=0.1364, rpn_acc=0.9995, pr.prec@10=0.224, pr.rec@10=0.9508, pr.prec@30=0.7771, pr.rec@30=0.82, pr.prec@50=0.952, pr.rec@50=0.6505, pr.prec@70=0.9945, pr.rec@70=0.4211, pr.prec@80=0.9986, pr.rec@80=0.249, pr.prec@90=1.0, pr.rec@90=0.03442, pr.prec@95=1.0, pr.rec@95=0.0002573, misc.num_vox=179424, misc.num_pos=52, misc.num_neg=42100, misc.num_anchors=42240, misc.lr=0.001778, misc.mem_usage=55.5\n", + "runtime.step=10350, runtime.steptime=0.2694, runtime.voxel_gene_time=0.0009372, runtime.prep_time=0.02757, loss.cls_loss=0.1225, loss.cls_loss_rt=0.1097, loss.loc_loss=0.2207, loss.loc_loss_rt=0.1947, loss.loc_elem=[0.004826, 0.004181, 0.01313, 0.01309, 0.02501, 0.01606, 0.02106], loss.cls_pos_rt=0.08451, loss.cls_neg_rt=0.0252, loss.dir_rt=0.1383, rpn_acc=0.9995, pr.prec@10=0.2226, pr.rec@10=0.9518, pr.prec@30=0.7772, pr.rec@30=0.8208, pr.prec@50=0.9516, pr.rec@50=0.6512, pr.prec@70=0.9944, pr.rec@70=0.4215, pr.prec@80=0.9987, pr.rec@80=0.2504, pr.prec@90=0.9997, pr.rec@90=0.03496, pr.prec@95=1.0, pr.rec@95=0.0002218, misc.num_vox=177622, misc.num_pos=57, misc.num_neg=42096, misc.num_anchors=42240, misc.lr=0.001753, misc.mem_usage=55.6\n", + "runtime.step=10400, runtime.steptime=0.269, runtime.voxel_gene_time=0.001348, runtime.prep_time=0.03075, loss.cls_loss=0.1217, loss.cls_loss_rt=0.1237, loss.loc_loss=0.2201, loss.loc_loss_rt=0.2366, loss.loc_elem=[0.005473, 0.003816, 0.01926, 0.01704, 0.0244, 0.02174, 0.02656], loss.cls_pos_rt=0.09, loss.cls_neg_rt=0.03369, loss.dir_rt=0.1288, rpn_acc=0.9995, pr.prec@10=0.2245, pr.rec@10=0.9522, pr.prec@30=0.7769, pr.rec@30=0.8232, pr.prec@50=0.952, pr.rec@50=0.6534, pr.prec@70=0.9944, pr.rec@70=0.4246, pr.prec@80=0.9988, pr.rec@80=0.2526, pr.prec@90=0.9998, pr.rec@90=0.03593, pr.prec@95=1.0, pr.rec@95=0.00024, misc.num_vox=175701, misc.num_pos=64, misc.num_neg=42070, misc.num_anchors=42240, misc.lr=0.001728, misc.mem_usage=55.5\n", + "runtime.step=10450, runtime.steptime=0.2666, runtime.voxel_gene_time=0.001005, runtime.prep_time=0.02932, loss.cls_loss=0.121, loss.cls_loss_rt=0.1374, loss.loc_loss=0.2193, loss.loc_loss_rt=0.2269, loss.loc_elem=[0.004734, 0.004277, 0.0243, 0.008519, 0.02048, 0.01858, 0.03254], loss.cls_pos_rt=0.1058, loss.cls_neg_rt=0.03155, loss.dir_rt=0.1399, rpn_acc=0.9995, pr.prec@10=0.2259, pr.rec@10=0.9528, pr.prec@30=0.7773, pr.rec@30=0.8242, pr.prec@50=0.9522, pr.rec@50=0.6539, pr.prec@70=0.9944, pr.rec@70=0.4243, pr.prec@80=0.9989, pr.rec@80=0.2529, pr.prec@90=0.9998, pr.rec@90=0.03691, pr.prec@95=1.0, pr.rec@95=0.0002453, misc.num_vox=175444, misc.num_pos=44, misc.num_neg=42123, misc.num_anchors=42240, misc.lr=0.001703, misc.mem_usage=55.5\n", + "reset Car\n", + "reset Car\n", + "reset Car\n", + "runtime.step=10500, runtime.steptime=0.2685, runtime.voxel_gene_time=0.0009439, runtime.prep_time=0.02861, loss.cls_loss=0.1202, loss.cls_loss_rt=0.09087, loss.loc_loss=0.2183, loss.loc_loss_rt=0.1751, loss.loc_elem=[0.003362, 0.003386, 0.01251, 0.01398, 0.0206, 0.01807, 0.01563], loss.cls_pos_rt=0.06673, loss.cls_neg_rt=0.02414, loss.dir_rt=0.1251, rpn_acc=0.9995, pr.prec@10=0.227, pr.rec@10=0.9537, pr.prec@30=0.7785, pr.rec@30=0.8256, pr.prec@50=0.9518, pr.rec@50=0.656, pr.prec@70=0.994, pr.rec@70=0.4249, pr.prec@80=0.9988, pr.rec@80=0.2525, pr.prec@90=0.9999, pr.rec@90=0.03621, pr.prec@95=1.0, pr.rec@95=0.0003119, misc.num_vox=176573, misc.num_pos=67, misc.num_neg=42069, misc.num_anchors=42240, misc.lr=0.001677, misc.mem_usage=55.6\n", + "WORKER 0 seed: 1592960352\n", + "WORKER 1 seed: 1592960353\n", + "WORKER 2 seed: 1592960354\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "runtime.step=10550, runtime.steptime=0.4293, runtime.voxel_gene_time=0.0009444, runtime.prep_time=0.02782, loss.cls_loss=0.1253, loss.cls_loss_rt=0.1605, loss.loc_loss=0.2235, loss.loc_loss_rt=0.229, loss.loc_elem=[0.005465, 0.004904, 0.02453, 0.01301, 0.0223, 0.02009, 0.02419], loss.cls_pos_rt=0.1021, loss.cls_neg_rt=0.05839, loss.dir_rt=0.1729, rpn_acc=0.9995, pr.prec@10=0.2197, pr.rec@10=0.9513, pr.prec@30=0.7716, pr.rec@30=0.8209, pr.prec@50=0.9481, pr.rec@50=0.6449, pr.prec@70=0.9926, pr.rec@70=0.4111, pr.prec@80=0.9992, pr.rec@80=0.2453, pr.prec@90=1.0, pr.rec@90=0.03286, pr.prec@95=1.0, pr.rec@95=0.0001675, misc.num_vox=177654, misc.num_pos=68, misc.num_neg=42055, misc.num_anchors=42240, misc.lr=0.001652, misc.mem_usage=55.3\n", + "runtime.step=10600, runtime.steptime=0.2668, runtime.voxel_gene_time=0.00109, runtime.prep_time=0.03076, loss.cls_loss=0.1216, loss.cls_loss_rt=0.13, loss.loc_loss=0.2197, loss.loc_loss_rt=0.2314, loss.loc_elem=[0.0061, 0.004188, 0.02143, 0.0103, 0.02187, 0.02521, 0.02662], loss.cls_pos_rt=0.08975, loss.cls_neg_rt=0.04022, loss.dir_rt=0.1666, rpn_acc=0.9995, pr.prec@10=0.2243, pr.rec@10=0.9541, pr.prec@30=0.7744, pr.rec@30=0.8236, pr.prec@50=0.9497, pr.rec@50=0.6505, pr.prec@70=0.9934, pr.rec@70=0.4159, pr.prec@80=0.9987, pr.rec@80=0.248, pr.prec@90=1.0, pr.rec@90=0.03268, pr.prec@95=1.0, pr.rec@95=7.834e-05, misc.num_vox=176959, misc.num_pos=73, misc.num_neg=42060, misc.num_anchors=42240, misc.lr=0.001627, misc.mem_usage=55.3\n", + "runtime.step=10650, runtime.steptime=0.2651, runtime.voxel_gene_time=0.001031, runtime.prep_time=0.02856, loss.cls_loss=0.1211, loss.cls_loss_rt=0.1137, loss.loc_loss=0.2188, loss.loc_loss_rt=0.2302, loss.loc_elem=[0.004889, 0.004706, 0.02068, 0.01501, 0.02444, 0.01977, 0.02561], loss.cls_pos_rt=0.08491, loss.cls_neg_rt=0.02876, loss.dir_rt=0.1473, rpn_acc=0.9995, pr.prec@10=0.2247, pr.rec@10=0.9541, pr.prec@30=0.7751, pr.rec@30=0.8239, pr.prec@50=0.95, pr.rec@50=0.6519, pr.prec@70=0.994, pr.rec@70=0.4193, pr.prec@80=0.9989, pr.rec@80=0.2511, pr.prec@90=1.0, pr.rec@90=0.03509, pr.prec@95=1.0, pr.rec@95=0.0002461, misc.num_vox=179274, misc.num_pos=69, misc.num_neg=42080, misc.num_anchors=42240, misc.lr=0.001602, misc.mem_usage=55.3\n", + "runtime.step=10700, runtime.steptime=0.2691, runtime.voxel_gene_time=0.001413, runtime.prep_time=0.03949, loss.cls_loss=0.1197, loss.cls_loss_rt=0.1215, loss.loc_loss=0.2156, loss.loc_loss_rt=0.2345, loss.loc_elem=[0.004273, 0.004323, 0.02074, 0.01656, 0.02424, 0.02258, 0.02453], loss.cls_pos_rt=0.08305, loss.cls_neg_rt=0.03842, loss.dir_rt=0.1145, rpn_acc=0.9995, pr.prec@10=0.2274, pr.rec@10=0.9551, pr.prec@30=0.7766, pr.rec@30=0.8261, pr.prec@50=0.95, pr.rec@50=0.6545, pr.prec@70=0.994, pr.rec@70=0.4235, pr.prec@80=0.9987, pr.rec@80=0.257, pr.prec@90=1.0, pr.rec@90=0.03836, pr.prec@95=1.0, pr.rec@95=0.0003125, misc.num_vox=173187, misc.num_pos=42, misc.num_neg=42143, misc.num_anchors=42240, misc.lr=0.001576, misc.mem_usage=55.3\n", + "runtime.step=10750, runtime.steptime=0.2704, runtime.voxel_gene_time=0.001043, runtime.prep_time=0.03494, loss.cls_loss=0.12, loss.cls_loss_rt=0.136, loss.loc_loss=0.2165, loss.loc_loss_rt=0.2159, loss.loc_elem=[0.005539, 0.003598, 0.01584, 0.01445, 0.02428, 0.02041, 0.02384], loss.cls_pos_rt=0.09968, loss.cls_neg_rt=0.0363, loss.dir_rt=0.1528, rpn_acc=0.9995, pr.prec@10=0.2275, pr.rec@10=0.955, pr.prec@30=0.7766, pr.rec@30=0.8255, pr.prec@50=0.9501, pr.rec@50=0.6538, pr.prec@70=0.9936, pr.rec@70=0.4225, pr.prec@80=0.9985, pr.rec@80=0.2557, pr.prec@90=1.0, pr.rec@90=0.0391, pr.prec@95=1.0, pr.rec@95=0.0003336, misc.num_vox=179807, misc.num_pos=66, misc.num_neg=42069, misc.num_anchors=42240, misc.lr=0.001551, misc.mem_usage=55.3\n", + "runtime.step=10800, runtime.steptime=0.2682, runtime.voxel_gene_time=0.001052, runtime.prep_time=0.02504, loss.cls_loss=0.1195, loss.cls_loss_rt=0.1374, loss.loc_loss=0.2163, loss.loc_loss_rt=0.2581, loss.loc_elem=[0.006482, 0.005818, 0.02205, 0.01216, 0.02751, 0.02146, 0.03358], loss.cls_pos_rt=0.09789, loss.cls_neg_rt=0.03948, loss.dir_rt=0.1667, rpn_acc=0.9995, pr.prec@10=0.2277, pr.rec@10=0.9553, pr.prec@30=0.7772, pr.rec@30=0.8266, pr.prec@50=0.9505, pr.rec@50=0.6568, pr.prec@70=0.9935, pr.rec@70=0.4266, pr.prec@80=0.9986, pr.rec@80=0.2589, pr.prec@90=1.0, pr.rec@90=0.03963, pr.prec@95=1.0, pr.rec@95=0.0003519, misc.num_vox=180000, misc.num_pos=61, misc.num_neg=42094, misc.num_anchors=42240, misc.lr=0.001526, misc.mem_usage=55.4\n", + "reset Car\n", + "reset Car\n", + "reset Car\n", + "WORKER 0 seed: 1592960443\n", + "WORKER 1 seed: 1592960444\n", + "WORKER 2 seed: 1592960445\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "runtime.step=10850, runtime.steptime=0.4392, runtime.voxel_gene_time=0.001569, runtime.prep_time=0.03767, loss.cls_loss=0.1202, loss.cls_loss_rt=0.1232, loss.loc_loss=0.2168, loss.loc_loss_rt=0.2061, loss.loc_elem=[0.003682, 0.004412, 0.01815, 0.01234, 0.0213, 0.0212, 0.02195], loss.cls_pos_rt=0.09285, loss.cls_neg_rt=0.03035, loss.dir_rt=0.1423, rpn_acc=0.9995, pr.prec@10=0.2342, pr.rec@10=0.9542, pr.prec@30=0.7823, pr.rec@30=0.8266, pr.prec@50=0.9541, pr.rec@50=0.6587, pr.prec@70=0.9918, pr.rec@70=0.4212, pr.prec@80=0.9972, pr.rec@80=0.2474, pr.prec@90=1.0, pr.rec@90=0.03366, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=179835, misc.num_pos=63, misc.num_neg=42095, misc.num_anchors=42240, misc.lr=0.001501, misc.mem_usage=55.2\n", + "#################################\n", + "# EVAL\n", + "#################################\n", + "Generate output labels...\n", + "[100.0%][===================>][12.24it/s][00:26>00:00] \n", + "generate label finished(141.38/s). start eval:\n", + "Evaluation official\n", + "Car AP(Average Precision)@0.70, 0.70, 0.70:\n", + "bbox AP:90.45, 88.17, 87.14\n", + "bev AP:89.63, 85.77, 79.33\n", + "3d AP:85.36, 74.50, 68.12\n", + "aos AP:0.53, 1.81, 2.94\n", + "Car AP(Average Precision)@0.70, 0.50, 0.50:\n", + "bbox AP:90.45, 88.17, 87.14\n", + "bev AP:90.59, 89.31, 88.80\n", + "3d AP:90.59, 89.21, 88.56\n", + "aos AP:0.53, 1.81, 2.94\n", + "\n", + "Evaluation coco\n", + "Car coco AP@0.50:0.05:0.95:\n", + "bbox AP:70.89, 66.60, 64.98\n", + "bev AP:68.58, 64.46, 63.04\n", + "3d AP:57.89, 53.34, 51.30\n", + "aos AP:0.40, 1.28, 2.13\n", + "\n", + "eval.kitti.official.Car.bbox@0.70=[90.45, 88.17, 87.14], eval.kitti.official.Car.bev@0.70=[89.63, 85.77, 79.33], eval.kitti.official.Car.3d@0.70=[85.36, 74.5, 68.12], eval.kitti.official.Car.aos=[0.5274, 1.815, 2.94], eval.kitti.official.Car.bev@0.50=[90.59, 89.31, 88.8], eval.kitti.official.Car.3d@0.50=[90.59, 89.21, 88.56], eval.kitti.coco.Car.bbox=[70.89, 66.6, 64.98], eval.kitti.coco.Car.bev=[68.58, 64.46, 63.04], eval.kitti.coco.Car.3d=[57.89, 53.34, 51.3], eval.kitti.coco.Car.aos=[0.3953, 1.283, 2.129]\n", + "runtime.step=10900, runtime.steptime=1.039, runtime.voxel_gene_time=0.0006454, runtime.prep_time=0.02223, loss.cls_loss=0.1185, loss.cls_loss_rt=0.1435, loss.loc_loss=0.2135, loss.loc_loss_rt=0.2487, loss.loc_elem=[0.004908, 0.004343, 0.02589, 0.014, 0.02266, 0.02179, 0.03073], loss.cls_pos_rt=0.1077, loss.cls_neg_rt=0.03578, loss.dir_rt=0.1313, rpn_acc=0.9995, pr.prec@10=0.2325, pr.rec@10=0.9561, pr.prec@30=0.7803, pr.rec@30=0.8304, pr.prec@50=0.9514, pr.rec@50=0.6605, pr.prec@70=0.9933, pr.rec@70=0.4283, pr.prec@80=0.9975, pr.rec@80=0.2543, pr.prec@90=1.0, pr.rec@90=0.03426, pr.prec@95=1.0, pr.rec@95=0.0002266, misc.num_vox=170772, misc.num_pos=55, misc.num_neg=42090, misc.num_anchors=42240, misc.lr=0.001475, misc.mem_usage=55.3\n", + "runtime.step=10950, runtime.steptime=0.2649, runtime.voxel_gene_time=0.0008364, runtime.prep_time=0.02125, loss.cls_loss=0.119, loss.cls_loss_rt=0.1192, loss.loc_loss=0.2136, loss.loc_loss_rt=0.222, loss.loc_elem=[0.004613, 0.004241, 0.0203, 0.01317, 0.01959, 0.01753, 0.03154], loss.cls_pos_rt=0.09223, loss.cls_neg_rt=0.02699, loss.dir_rt=0.1681, rpn_acc=0.9995, pr.prec@10=0.2326, pr.rec@10=0.9541, pr.prec@30=0.7824, pr.rec@30=0.8297, pr.prec@50=0.9523, pr.rec@50=0.6622, pr.prec@70=0.9936, pr.rec@70=0.4321, pr.prec@80=0.9982, pr.rec@80=0.2604, pr.prec@90=1.0, pr.rec@90=0.0367, pr.prec@95=1.0, pr.rec@95=0.0002523, misc.num_vox=175157, misc.num_pos=37, misc.num_neg=42151, misc.num_anchors=42240, misc.lr=0.00145, misc.mem_usage=55.4\n", + "runtime.step=11000, runtime.steptime=0.2652, runtime.voxel_gene_time=0.0008261, runtime.prep_time=0.02408, loss.cls_loss=0.1178, loss.cls_loss_rt=0.1354, loss.loc_loss=0.2114, loss.loc_loss_rt=0.2282, loss.loc_elem=[0.003794, 0.004472, 0.0249, 0.01556, 0.01982, 0.01826, 0.02727], loss.cls_pos_rt=0.09376, loss.cls_neg_rt=0.04163, loss.dir_rt=0.1471, rpn_acc=0.9995, pr.prec@10=0.2337, pr.rec@10=0.9554, pr.prec@30=0.7838, pr.rec@30=0.8309, pr.prec@50=0.9515, pr.rec@50=0.6634, pr.prec@70=0.9936, pr.rec@70=0.4337, pr.prec@80=0.9983, pr.rec@80=0.263, pr.prec@90=1.0, pr.rec@90=0.03794, pr.prec@95=1.0, pr.rec@95=0.0002719, misc.num_vox=173760, misc.num_pos=45, misc.num_neg=42124, misc.num_anchors=42240, misc.lr=0.001425, misc.mem_usage=55.4\n", + "runtime.step=11050, runtime.steptime=0.2641, runtime.voxel_gene_time=0.0009577, runtime.prep_time=0.02877, loss.cls_loss=0.1181, loss.cls_loss_rt=0.1074, loss.loc_loss=0.2119, loss.loc_loss_rt=0.2284, loss.loc_elem=[0.005159, 0.004345, 0.01881, 0.01541, 0.02762, 0.01701, 0.02588], loss.cls_pos_rt=0.06658, loss.cls_neg_rt=0.04085, loss.dir_rt=0.1049, rpn_acc=0.9995, pr.prec@10=0.2326, pr.rec@10=0.9555, pr.prec@30=0.7819, pr.rec@30=0.831, pr.prec@50=0.9512, pr.rec@50=0.6616, pr.prec@70=0.9931, pr.rec@70=0.4304, pr.prec@80=0.9982, pr.rec@80=0.2586, pr.prec@90=1.0, pr.rec@90=0.03724, pr.prec@95=1.0, pr.rec@95=0.0003271, misc.num_vox=179803, misc.num_pos=67, misc.num_neg=42084, misc.num_anchors=42240, misc.lr=0.001399, misc.mem_usage=55.4\n", + "runtime.step=11100, runtime.steptime=0.2638, runtime.voxel_gene_time=0.0009871, runtime.prep_time=0.02853, loss.cls_loss=0.1178, loss.cls_loss_rt=0.1175, loss.loc_loss=0.2109, loss.loc_loss_rt=0.1986, loss.loc_elem=[0.004486, 0.004019, 0.01605, 0.01445, 0.02207, 0.0163, 0.0219], loss.cls_pos_rt=0.08356, loss.cls_neg_rt=0.03391, loss.dir_rt=0.1422, rpn_acc=0.9995, pr.prec@10=0.2336, pr.rec@10=0.956, pr.prec@30=0.7817, pr.rec@30=0.8304, pr.prec@50=0.9522, pr.rec@50=0.6629, pr.prec@70=0.9933, pr.rec@70=0.4333, pr.prec@80=0.9983, pr.rec@80=0.262, pr.prec@90=1.0, pr.rec@90=0.03861, pr.prec@95=1.0, pr.rec@95=0.0003885, misc.num_vox=179629, misc.num_pos=52, misc.num_neg=42109, misc.num_anchors=42240, misc.lr=0.001374, misc.mem_usage=55.4\n", + "reset Car\n", + "reset Car\n", + "reset Car\n", + "WORKER 0 seed: 1592960572\n", + "WORKER 1 seed: 1592960573\n", + "WORKER 2 seed: 1592960574\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "runtime.step=11150, runtime.steptime=0.4297, runtime.voxel_gene_time=0.0009785, runtime.prep_time=0.03116, loss.cls_loss=0.1187, loss.cls_loss_rt=0.1241, loss.loc_loss=0.2113, loss.loc_loss_rt=0.2354, loss.loc_elem=[0.005249, 0.004083, 0.0248, 0.01481, 0.02319, 0.01632, 0.02923], loss.cls_pos_rt=0.1008, loss.cls_neg_rt=0.02326, loss.dir_rt=0.1935, rpn_acc=0.9995, pr.prec@10=0.2213, pr.rec@10=0.9569, pr.prec@30=0.7747, pr.rec@30=0.827, pr.prec@50=0.9547, pr.rec@50=0.6543, pr.prec@70=0.9946, pr.rec@70=0.4307, pr.prec@80=0.9979, pr.rec@80=0.2672, pr.prec@90=1.0, pr.rec@90=0.05152, pr.prec@95=1.0, pr.rec@95=0.0006809, misc.num_vox=179962, misc.num_pos=70, misc.num_neg=42070, misc.num_anchors=42240, misc.lr=0.001349, misc.mem_usage=55.3\n", + "runtime.step=11200, runtime.steptime=0.2701, runtime.voxel_gene_time=0.001134, runtime.prep_time=0.0327, loss.cls_loss=0.12, loss.cls_loss_rt=0.101, loss.loc_loss=0.2128, loss.loc_loss_rt=0.2207, loss.loc_elem=[0.003858, 0.004348, 0.03002, 0.01114, 0.02198, 0.0156, 0.02339], loss.cls_pos_rt=0.08264, loss.cls_neg_rt=0.0184, loss.dir_rt=0.1326, rpn_acc=0.9995, pr.prec@10=0.2262, pr.rec@10=0.9549, pr.prec@30=0.7779, pr.rec@30=0.8243, pr.prec@50=0.9534, pr.rec@50=0.6562, pr.prec@70=0.994, pr.rec@70=0.4304, pr.prec@80=0.9987, pr.rec@80=0.2621, pr.prec@90=1.0, pr.rec@90=0.04517, pr.prec@95=1.0, pr.rec@95=0.0005435, misc.num_vox=174796, misc.num_pos=70, misc.num_neg=42077, misc.num_anchors=42240, misc.lr=0.001324, misc.mem_usage=55.5\n", + "runtime.step=11250, runtime.steptime=0.2665, runtime.voxel_gene_time=0.001058, runtime.prep_time=0.03387, loss.cls_loss=0.1173, loss.cls_loss_rt=0.09809, loss.loc_loss=0.2107, loss.loc_loss_rt=0.1877, loss.loc_elem=[0.004948, 0.003361, 0.01441, 0.01219, 0.02225, 0.01728, 0.01941], loss.cls_pos_rt=0.05902, loss.cls_neg_rt=0.03907, loss.dir_rt=0.08186, rpn_acc=0.9995, pr.prec@10=0.231, pr.rec@10=0.9563, pr.prec@30=0.7805, pr.rec@30=0.8289, pr.prec@50=0.9533, pr.rec@50=0.6616, pr.prec@70=0.9944, pr.rec@70=0.4358, pr.prec@80=0.9987, pr.rec@80=0.2687, pr.prec@90=1.0, pr.rec@90=0.04977, pr.prec@95=1.0, pr.rec@95=0.0006226, misc.num_vox=179385, misc.num_pos=52, misc.num_neg=42102, misc.num_anchors=42240, misc.lr=0.001298, misc.mem_usage=55.4\n", + "runtime.step=11300, runtime.steptime=0.269, runtime.voxel_gene_time=0.001022, runtime.prep_time=0.03519, loss.cls_loss=0.1165, loss.cls_loss_rt=0.1036, loss.loc_loss=0.2102, loss.loc_loss_rt=0.2011, loss.loc_elem=[0.004217, 0.00307, 0.01715, 0.01375, 0.02332, 0.02134, 0.01772], loss.cls_pos_rt=0.07166, loss.cls_neg_rt=0.03198, loss.dir_rt=0.1562, rpn_acc=0.9995, pr.prec@10=0.2332, pr.rec@10=0.9568, pr.prec@30=0.7812, pr.rec@30=0.8293, pr.prec@50=0.9534, pr.rec@50=0.6629, pr.prec@70=0.9942, pr.rec@70=0.4393, pr.prec@80=0.9989, pr.rec@80=0.273, pr.prec@90=1.0, pr.rec@90=0.05128, pr.prec@95=1.0, pr.rec@95=0.0005388, misc.num_vox=175112, misc.num_pos=42, misc.num_neg=42137, misc.num_anchors=42240, misc.lr=0.001273, misc.mem_usage=55.4\n", + "runtime.step=11350, runtime.steptime=0.269, runtime.voxel_gene_time=0.001068, runtime.prep_time=0.0337, loss.cls_loss=0.1164, loss.cls_loss_rt=0.131, loss.loc_loss=0.2115, loss.loc_loss_rt=0.2267, loss.loc_elem=[0.006951, 0.004027, 0.02562, 0.01293, 0.02416, 0.01919, 0.02046], loss.cls_pos_rt=0.08718, loss.cls_neg_rt=0.04383, loss.dir_rt=0.1712, rpn_acc=0.9995, pr.prec@10=0.2323, pr.rec@10=0.9571, pr.prec@30=0.7805, pr.rec@30=0.8304, pr.prec@50=0.9529, pr.rec@50=0.6641, pr.prec@70=0.994, pr.rec@70=0.4386, pr.prec@80=0.9986, pr.rec@80=0.273, pr.prec@90=1.0, pr.rec@90=0.0518, pr.prec@95=1.0, pr.rec@95=0.0006574, misc.num_vox=175118, misc.num_pos=52, misc.num_neg=42106, misc.num_anchors=42240, misc.lr=0.001248, misc.mem_usage=55.3\n", + "runtime.step=11400, runtime.steptime=0.2674, runtime.voxel_gene_time=0.0008543, runtime.prep_time=0.02067, loss.cls_loss=0.1156, loss.cls_loss_rt=0.09385, loss.loc_loss=0.2105, loss.loc_loss_rt=0.1909, loss.loc_elem=[0.004152, 0.004737, 0.01626, 0.01321, 0.01883, 0.01923, 0.01902], loss.cls_pos_rt=0.06366, loss.cls_neg_rt=0.03019, loss.dir_rt=0.15, rpn_acc=0.9995, pr.prec@10=0.2344, pr.rec@10=0.9574, pr.prec@30=0.7817, pr.rec@30=0.8321, pr.prec@50=0.9537, pr.rec@50=0.6661, pr.prec@70=0.9939, pr.rec@70=0.4408, pr.prec@80=0.9986, pr.rec@80=0.2736, pr.prec@90=1.0, pr.rec@90=0.0511, pr.prec@95=1.0, pr.rec@95=0.0006862, misc.num_vox=178433, misc.num_pos=48, misc.num_neg=42106, misc.num_anchors=42240, misc.lr=0.001223, misc.mem_usage=55.4\n", + "reset Car\n", + "reset Car\n", + "reset Car\n", + "WORKER 0 seed: 1592960663\n", + "WORKER 1 seed: 1592960664\n", + "WORKER 2 seed: 1592960665\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "runtime.step=11450, runtime.steptime=0.4323, runtime.voxel_gene_time=0.0009754, runtime.prep_time=0.02107, loss.cls_loss=0.1191, loss.cls_loss_rt=0.1221, loss.loc_loss=0.2033, loss.loc_loss_rt=0.1827, loss.loc_elem=[0.004363, 0.003784, 0.0132, 0.01327, 0.02034, 0.01774, 0.01864], loss.cls_pos_rt=0.08501, loss.cls_neg_rt=0.0371, loss.dir_rt=0.1592, rpn_acc=0.9995, pr.prec@10=0.2231, pr.rec@10=0.9567, pr.prec@30=0.785, pr.rec@30=0.8316, pr.prec@50=0.9566, pr.rec@50=0.6632, pr.prec@70=0.9914, pr.rec@70=0.4382, pr.prec@80=0.9969, pr.rec@80=0.2788, pr.prec@90=1.0, pr.rec@90=0.05268, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=178712, misc.num_pos=57, misc.num_neg=42093, misc.num_anchors=42240, misc.lr=0.001199, misc.mem_usage=55.1\n", + "runtime.step=11500, runtime.steptime=0.2706, runtime.voxel_gene_time=0.001555, runtime.prep_time=0.03719, loss.cls_loss=0.1212, loss.cls_loss_rt=0.1125, loss.loc_loss=0.2107, loss.loc_loss_rt=0.1952, loss.loc_elem=[0.004961, 0.003184, 0.01551, 0.01376, 0.02623, 0.01417, 0.01979], loss.cls_pos_rt=0.06755, loss.cls_neg_rt=0.04496, loss.dir_rt=0.1169, rpn_acc=0.9995, pr.prec@10=0.2236, pr.rec@10=0.9527, pr.prec@30=0.7814, pr.rec@30=0.8251, pr.prec@50=0.9542, pr.rec@50=0.6601, pr.prec@70=0.9923, pr.rec@70=0.439, pr.prec@80=0.998, pr.rec@80=0.2785, pr.prec@90=1.0, pr.rec@90=0.05839, pr.prec@95=1.0, pr.rec@95=0.00103, misc.num_vox=178817, misc.num_pos=52, misc.num_neg=42113, misc.num_anchors=42240, misc.lr=0.001174, misc.mem_usage=55.3\n", + "runtime.step=11550, runtime.steptime=0.2677, runtime.voxel_gene_time=0.001019, runtime.prep_time=0.03401, loss.cls_loss=0.1176, loss.cls_loss_rt=0.1123, loss.loc_loss=0.2085, loss.loc_loss_rt=0.1921, loss.loc_elem=[0.005146, 0.004118, 0.02087, 0.01137, 0.01817, 0.01864, 0.01775], loss.cls_pos_rt=0.07129, loss.cls_neg_rt=0.04102, loss.dir_rt=0.1302, rpn_acc=0.9995, pr.prec@10=0.2312, pr.rec@10=0.9547, pr.prec@30=0.7858, pr.rec@30=0.8298, pr.prec@50=0.9542, pr.rec@50=0.6664, pr.prec@70=0.9932, pr.rec@70=0.4439, pr.prec@80=0.9987, pr.rec@80=0.2804, pr.prec@90=1.0, pr.rec@90=0.05345, pr.prec@95=1.0, pr.rec@95=0.0007322, misc.num_vox=176580, misc.num_pos=44, misc.num_neg=42125, misc.num_anchors=42240, misc.lr=0.001149, misc.mem_usage=55.3\n", + "runtime.step=11600, runtime.steptime=0.2684, runtime.voxel_gene_time=0.001574, runtime.prep_time=0.04195, loss.cls_loss=0.1172, loss.cls_loss_rt=0.1258, loss.loc_loss=0.2072, loss.loc_loss_rt=0.2106, loss.loc_elem=[0.005293, 0.003833, 0.02482, 0.01176, 0.01942, 0.01539, 0.02476], loss.cls_pos_rt=0.09394, loss.cls_neg_rt=0.03188, loss.dir_rt=0.1371, rpn_acc=0.9995, pr.prec@10=0.2339, pr.rec@10=0.9546, pr.prec@30=0.7858, pr.rec@30=0.8299, pr.prec@50=0.9536, pr.rec@50=0.6669, pr.prec@70=0.9932, pr.rec@70=0.4447, pr.prec@80=0.9987, pr.rec@80=0.2822, pr.prec@90=1.0, pr.rec@90=0.05463, pr.prec@95=1.0, pr.rec@95=0.0009068, misc.num_vox=176764, misc.num_pos=70, misc.num_neg=42063, misc.num_anchors=42240, misc.lr=0.001125, misc.mem_usage=55.3\n", + "runtime.step=11650, runtime.steptime=0.2676, runtime.voxel_gene_time=0.001466, runtime.prep_time=0.03676, loss.cls_loss=0.117, loss.cls_loss_rt=0.1098, loss.loc_loss=0.2076, loss.loc_loss_rt=0.2411, loss.loc_elem=[0.003824, 0.004276, 0.03157, 0.01617, 0.02535, 0.01695, 0.02242], loss.cls_pos_rt=0.07135, loss.cls_neg_rt=0.03842, loss.dir_rt=0.1712, rpn_acc=0.9995, pr.prec@10=0.2333, pr.rec@10=0.955, pr.prec@30=0.7838, pr.rec@30=0.8301, pr.prec@50=0.954, pr.rec@50=0.6665, pr.prec@70=0.9934, pr.rec@70=0.4435, pr.prec@80=0.9989, pr.rec@80=0.2811, pr.prec@90=1.0, pr.rec@90=0.0538, pr.prec@95=1.0, pr.rec@95=0.0008188, misc.num_vox=177848, misc.num_pos=68, misc.num_neg=42070, misc.num_anchors=42240, misc.lr=0.0011, misc.mem_usage=55.3\n", + "runtime.step=11700, runtime.steptime=0.2679, runtime.voxel_gene_time=0.0009532, runtime.prep_time=0.03147, loss.cls_loss=0.1163, loss.cls_loss_rt=0.0972, loss.loc_loss=0.2073, loss.loc_loss_rt=0.186, loss.loc_elem=[0.003613, 0.004225, 0.01922, 0.009697, 0.02246, 0.01901, 0.01478], loss.cls_pos_rt=0.06796, loss.cls_neg_rt=0.02924, loss.dir_rt=0.1119, rpn_acc=0.9995, pr.prec@10=0.236, pr.rec@10=0.9556, pr.prec@30=0.784, pr.rec@30=0.8318, pr.prec@50=0.9539, pr.rec@50=0.6679, pr.prec@70=0.9934, pr.rec@70=0.4435, pr.prec@80=0.999, pr.rec@80=0.2804, pr.prec@90=1.0, pr.rec@90=0.05298, pr.prec@95=1.0, pr.rec@95=0.0008754, misc.num_vox=174967, misc.num_pos=53, misc.num_neg=42104, misc.num_anchors=42240, misc.lr=0.001076, misc.mem_usage=55.3\n", + "reset Car\n", + "reset Car\n", + "reset Car\n", + "WORKER 0 seed: 1592960754\n", + "WORKER 1 seed: 1592960755\n", + "WORKER 2 seed: 1592960756\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "runtime.step=11750, runtime.steptime=0.4303, runtime.voxel_gene_time=0.001335, runtime.prep_time=0.02447, loss.cls_loss=0.1302, loss.cls_loss_rt=0.1154, loss.loc_loss=0.2226, loss.loc_loss_rt=0.2034, loss.loc_elem=[0.004518, 0.003508, 0.02078, 0.01074, 0.0202, 0.02289, 0.01905], loss.cls_pos_rt=0.08274, loss.cls_neg_rt=0.03265, loss.dir_rt=0.1342, rpn_acc=0.9995, pr.prec@10=0.2254, pr.rec@10=0.9431, pr.prec@30=0.788, pr.rec@30=0.8122, pr.prec@50=0.9535, pr.rec@50=0.6347, pr.prec@70=0.9946, pr.rec@70=0.4067, pr.prec@80=0.9971, pr.rec@80=0.2532, pr.prec@90=1.0, pr.rec@90=0.04597, pr.prec@95=0.0, pr.rec@95=0.0, misc.num_vox=178269, misc.num_pos=67, misc.num_neg=42076, misc.num_anchors=42240, misc.lr=0.001051, misc.mem_usage=55.0\n", + "runtime.step=11800, runtime.steptime=0.2666, runtime.voxel_gene_time=0.001035, runtime.prep_time=0.03033, loss.cls_loss=0.1161, loss.cls_loss_rt=0.1103, loss.loc_loss=0.2098, loss.loc_loss_rt=0.202, loss.loc_elem=[0.004276, 0.005253, 0.019, 0.01287, 0.01901, 0.0202, 0.0204], loss.cls_pos_rt=0.08193, loss.cls_neg_rt=0.02835, loss.dir_rt=0.1169, rpn_acc=0.9995, pr.prec@10=0.2339, pr.rec@10=0.9533, pr.prec@30=0.7876, pr.rec@30=0.83, pr.prec@50=0.9556, pr.rec@50=0.6678, pr.prec@70=0.9953, pr.rec@70=0.4414, pr.prec@80=0.9991, pr.rec@80=0.2797, pr.prec@90=1.0, pr.rec@90=0.05819, pr.prec@95=1.0, pr.rec@95=0.001203, misc.num_vox=180000, misc.num_pos=51, misc.num_neg=42114, misc.num_anchors=42240, misc.lr=0.001027, misc.mem_usage=55.3\n", + "runtime.step=11850, runtime.steptime=0.2677, runtime.voxel_gene_time=0.001131, runtime.prep_time=0.03171, loss.cls_loss=0.1154, loss.cls_loss_rt=0.1059, loss.loc_loss=0.2075, loss.loc_loss_rt=0.1827, loss.loc_elem=[0.005188, 0.004489, 0.01614, 0.01224, 0.02282, 0.01862, 0.01185], loss.cls_pos_rt=0.06713, loss.cls_neg_rt=0.03875, loss.dir_rt=0.122, rpn_acc=0.9995, pr.prec@10=0.2354, pr.rec@10=0.9557, pr.prec@30=0.7867, pr.rec@30=0.8326, pr.prec@50=0.9528, pr.rec@50=0.669, pr.prec@70=0.9936, pr.rec@70=0.4459, pr.prec@80=0.999, pr.rec@80=0.2828, pr.prec@90=1.0, pr.rec@90=0.05646, pr.prec@95=1.0, pr.rec@95=0.0008206, misc.num_vox=178657, misc.num_pos=46, misc.num_neg=42121, misc.num_anchors=42240, misc.lr=0.001003, misc.mem_usage=55.3\n", + "runtime.step=11900, runtime.steptime=0.2656, runtime.voxel_gene_time=0.001021, runtime.prep_time=0.03057, loss.cls_loss=0.1138, loss.cls_loss_rt=0.1068, loss.loc_loss=0.2042, loss.loc_loss_rt=0.1951, loss.loc_elem=[0.004164, 0.003974, 0.01956, 0.01293, 0.02151, 0.0166, 0.0188], loss.cls_pos_rt=0.07924, loss.cls_neg_rt=0.02759, loss.dir_rt=0.1301, rpn_acc=0.9995, pr.prec@10=0.2389, pr.rec@10=0.9565, pr.prec@30=0.7884, pr.rec@30=0.8344, pr.prec@50=0.9536, pr.rec@50=0.6718, pr.prec@70=0.9937, pr.rec@70=0.4489, pr.prec@80=0.9989, pr.rec@80=0.2863, pr.prec@90=1.0, pr.rec@90=0.05971, pr.prec@95=1.0, pr.rec@95=0.001058, misc.num_vox=177209, misc.num_pos=45, misc.num_neg=42119, misc.num_anchors=42240, misc.lr=0.0009795, misc.mem_usage=55.3\n", + "runtime.step=11950, runtime.steptime=0.2675, runtime.voxel_gene_time=0.001001, runtime.prep_time=0.02541, loss.cls_loss=0.1138, loss.cls_loss_rt=0.1141, loss.loc_loss=0.2037, loss.loc_loss_rt=0.2353, loss.loc_elem=[0.003856, 0.004331, 0.02521, 0.01312, 0.02459, 0.0206, 0.02594], loss.cls_pos_rt=0.07563, loss.cls_neg_rt=0.03849, loss.dir_rt=0.1142, rpn_acc=0.9995, pr.prec@10=0.2404, pr.rec@10=0.9568, pr.prec@30=0.7873, pr.rec@30=0.8354, pr.prec@50=0.9533, pr.rec@50=0.671, pr.prec@70=0.9937, pr.rec@70=0.4492, pr.prec@80=0.999, pr.rec@80=0.2888, pr.prec@90=1.0, pr.rec@90=0.06302, pr.prec@95=1.0, pr.rec@95=0.001295, misc.num_vox=178933, misc.num_pos=46, misc.num_neg=42122, misc.num_anchors=42240, misc.lr=0.0009559, misc.mem_usage=55.3\n", + "runtime.step=12000, runtime.steptime=0.269, runtime.voxel_gene_time=0.000946, runtime.prep_time=0.02719, loss.cls_loss=0.114, loss.cls_loss_rt=0.09803, loss.loc_loss=0.203, loss.loc_loss_rt=0.1771, loss.loc_elem=[0.003566, 0.003319, 0.01577, 0.009028, 0.01917, 0.02403, 0.01367], loss.cls_pos_rt=0.06534, loss.cls_neg_rt=0.03269, loss.dir_rt=0.1217, rpn_acc=0.9995, pr.prec@10=0.2402, pr.rec@10=0.957, pr.prec@30=0.7856, pr.rec@30=0.8346, pr.prec@50=0.953, pr.rec@50=0.6695, pr.prec@70=0.9935, pr.rec@70=0.4478, pr.prec@80=0.9989, pr.rec@80=0.2869, pr.prec@90=1.0, pr.rec@90=0.06111, pr.prec@95=1.0, pr.rec@95=0.001314, misc.num_vox=176266, misc.num_pos=55, misc.num_neg=42103, misc.num_anchors=42240, misc.lr=0.0009323, misc.mem_usage=55.3\n", + "reset Car\n", + "reset Car\n", + "reset Car\n", + "runtime.step=12050, runtime.steptime=0.2683, runtime.voxel_gene_time=0.001242, runtime.prep_time=0.03968, loss.cls_loss=0.1138, loss.cls_loss_rt=0.1347, loss.loc_loss=0.2026, loss.loc_loss_rt=0.199, loss.loc_elem=[0.005984, 0.003752, 0.0158, 0.01621, 0.02354, 0.01619, 0.01803], loss.cls_pos_rt=0.09792, loss.cls_neg_rt=0.03681, loss.dir_rt=0.1617, rpn_acc=0.9995, pr.prec@10=0.241, pr.rec@10=0.9575, pr.prec@30=0.7863, pr.rec@30=0.8347, pr.prec@50=0.9532, pr.rec@50=0.6705, pr.prec@70=0.9935, pr.rec@70=0.4478, pr.prec@80=0.9989, pr.rec@80=0.2866, pr.prec@90=1.0, pr.rec@90=0.06035, pr.prec@95=1.0, pr.rec@95=0.001187, misc.num_vox=175385, misc.num_pos=44, misc.num_neg=42116, misc.num_anchors=42240, misc.lr=0.000909, misc.mem_usage=55.4\n", + "WORKER 0 seed: 1592960845\n", + "WORKER 1 seed: 1592960846\n", + "WORKER 2 seed: 1592960847\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "runtime.step=12100, runtime.steptime=0.4308, runtime.voxel_gene_time=0.0009906, runtime.prep_time=0.02439, loss.cls_loss=0.1154, loss.cls_loss_rt=0.1031, loss.loc_loss=0.207, loss.loc_loss_rt=0.1773, loss.loc_elem=[0.003627, 0.003439, 0.01646, 0.01161, 0.01913, 0.01431, 0.02005], loss.cls_pos_rt=0.0718, loss.cls_neg_rt=0.03135, loss.dir_rt=0.08553, rpn_acc=0.9995, pr.prec@10=0.2314, pr.rec@10=0.9566, pr.prec@30=0.7798, pr.rec@30=0.8309, pr.prec@50=0.9552, pr.rec@50=0.6709, pr.prec@70=0.9951, pr.rec@70=0.4429, pr.prec@80=0.9996, pr.rec@80=0.2806, pr.prec@90=1.0, pr.rec@90=0.05822, pr.prec@95=1.0, pr.rec@95=0.001148, misc.num_vox=177406, misc.num_pos=58, misc.num_neg=42097, misc.num_anchors=42240, misc.lr=0.0008858, misc.mem_usage=55.3\n", + "runtime.step=12150, runtime.steptime=0.2672, runtime.voxel_gene_time=0.0009568, runtime.prep_time=0.02336, loss.cls_loss=0.1143, loss.cls_loss_rt=0.1215, loss.loc_loss=0.2047, loss.loc_loss_rt=0.2333, loss.loc_elem=[0.0049, 0.004219, 0.03319, 0.01155, 0.0235, 0.01684, 0.02242], loss.cls_pos_rt=0.08767, loss.cls_neg_rt=0.03382, loss.dir_rt=0.1476, rpn_acc=0.9995, pr.prec@10=0.235, pr.rec@10=0.9569, pr.prec@30=0.7845, pr.rec@30=0.8335, pr.prec@50=0.9556, pr.rec@50=0.6708, pr.prec@70=0.9952, pr.rec@70=0.4467, pr.prec@80=0.9995, pr.rec@80=0.2864, pr.prec@90=1.0, pr.rec@90=0.05968, pr.prec@95=1.0, pr.rec@95=0.001383, misc.num_vox=178566, misc.num_pos=67, misc.num_neg=42081, misc.num_anchors=42240, misc.lr=0.0008627, misc.mem_usage=55.3\n", + "runtime.step=12200, runtime.steptime=0.2672, runtime.voxel_gene_time=0.00151, runtime.prep_time=0.04257, loss.cls_loss=0.1131, loss.cls_loss_rt=0.1091, loss.loc_loss=0.2035, loss.loc_loss_rt=0.2141, loss.loc_elem=[0.005193, 0.003745, 0.02679, 0.01412, 0.02231, 0.01641, 0.0185], loss.cls_pos_rt=0.07749, loss.cls_neg_rt=0.03159, loss.dir_rt=0.1271, rpn_acc=0.9995, pr.prec@10=0.2384, pr.rec@10=0.9579, pr.prec@30=0.7876, pr.rec@30=0.8366, pr.prec@50=0.9558, pr.rec@50=0.6739, pr.prec@70=0.9942, pr.rec@70=0.4506, pr.prec@80=0.9996, pr.rec@80=0.2926, pr.prec@90=1.0, pr.rec@90=0.06501, pr.prec@95=1.0, pr.rec@95=0.00142, misc.num_vox=178725, misc.num_pos=68, misc.num_neg=42073, misc.num_anchors=42240, misc.lr=0.0008399, misc.mem_usage=55.3\n", + "runtime.step=12250, runtime.steptime=0.2703, runtime.voxel_gene_time=0.001149, runtime.prep_time=0.02593, loss.cls_loss=0.1124, loss.cls_loss_rt=0.1073, loss.loc_loss=0.2026, loss.loc_loss_rt=0.2006, loss.loc_elem=[0.005402, 0.003558, 0.01297, 0.01607, 0.02019, 0.02175, 0.02038], loss.cls_pos_rt=0.07512, loss.cls_neg_rt=0.03219, loss.dir_rt=0.1182, rpn_acc=0.9995, pr.prec@10=0.241, pr.rec@10=0.9577, pr.prec@30=0.7886, pr.rec@30=0.8375, pr.prec@50=0.9561, pr.rec@50=0.6767, pr.prec@70=0.9944, pr.rec@70=0.4578, pr.prec@80=0.9994, pr.rec@80=0.2999, pr.prec@90=1.0, pr.rec@90=0.07144, pr.prec@95=1.0, pr.rec@95=0.00201, misc.num_vox=180000, misc.num_pos=57, misc.num_neg=42106, misc.num_anchors=42240, misc.lr=0.0008172, misc.mem_usage=55.6\n", + "runtime.step=12300, runtime.steptime=0.2666, runtime.voxel_gene_time=0.0009298, runtime.prep_time=0.02714, loss.cls_loss=0.1128, loss.cls_loss_rt=0.1041, loss.loc_loss=0.2026, loss.loc_loss_rt=0.1834, loss.loc_elem=[0.003789, 0.004716, 0.01839, 0.01308, 0.01936, 0.0178, 0.01456], loss.cls_pos_rt=0.0802, loss.cls_neg_rt=0.02385, loss.dir_rt=0.1794, rpn_acc=0.9995, pr.prec@10=0.2402, pr.rec@10=0.9578, pr.prec@30=0.7877, pr.rec@30=0.8368, pr.prec@50=0.9556, pr.rec@50=0.675, pr.prec@70=0.9943, pr.rec@70=0.4543, pr.prec@80=0.9992, pr.rec@80=0.2957, pr.prec@90=1.0, pr.rec@90=0.06963, pr.prec@95=1.0, pr.rec@95=0.001949, misc.num_vox=178861, misc.num_pos=69, misc.num_neg=42070, misc.num_anchors=42240, misc.lr=0.0007948, misc.mem_usage=55.5\n", + "reset Car\n", + "reset Car\n", + "reset Car\n", + "runtime.step=12350, runtime.steptime=0.2682, runtime.voxel_gene_time=0.0009956, runtime.prep_time=0.03634, loss.cls_loss=0.1122, loss.cls_loss_rt=0.08953, loss.loc_loss=0.2015, loss.loc_loss_rt=0.1696, loss.loc_elem=[0.004991, 0.003437, 0.01398, 0.01099, 0.01947, 0.01432, 0.01761], loss.cls_pos_rt=0.05692, loss.cls_neg_rt=0.03262, loss.dir_rt=0.1037, rpn_acc=0.9995, pr.prec@10=0.2417, pr.rec@10=0.9585, pr.prec@30=0.7874, pr.rec@30=0.8378, pr.prec@50=0.9552, pr.rec@50=0.6769, pr.prec@70=0.9942, pr.rec@70=0.4558, pr.prec@80=0.9991, pr.rec@80=0.2961, pr.prec@90=1.0, pr.rec@90=0.07019, pr.prec@95=1.0, pr.rec@95=0.002009, misc.num_vox=172758, misc.num_pos=56, misc.num_neg=42101, misc.num_anchors=42240, misc.lr=0.0007725, misc.mem_usage=55.4\n", + "WORKER 0 seed: 1592960936\n", + "WORKER 1 seed: 1592960937\n", + "WORKER 2 seed: 1592960938\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "runtime.step=12400, runtime.steptime=0.4269, runtime.voxel_gene_time=0.001088, runtime.prep_time=0.03729, loss.cls_loss=0.1162, loss.cls_loss_rt=0.1252, loss.loc_loss=0.2052, loss.loc_loss_rt=0.2283, loss.loc_elem=[0.005275, 0.004312, 0.02283, 0.01668, 0.02195, 0.01805, 0.02506], loss.cls_pos_rt=0.09415, loss.cls_neg_rt=0.03108, loss.dir_rt=0.1638, rpn_acc=0.9995, pr.prec@10=0.2293, pr.rec@10=0.9545, pr.prec@30=0.7839, pr.rec@30=0.8289, pr.prec@50=0.9555, pr.rec@50=0.6685, pr.prec@70=0.9934, pr.rec@70=0.452, pr.prec@80=0.999, pr.rec@80=0.2999, pr.prec@90=1.0, pr.rec@90=0.0797, pr.prec@95=1.0, pr.rec@95=0.002733, misc.num_vox=175320, misc.num_pos=48, misc.num_neg=42122, misc.num_anchors=42240, misc.lr=0.0007505, misc.mem_usage=55.3\n", + "#################################\n", + "# EVAL\n", + "#################################\n", + "Generate output labels...\n", + "[100.0%][===================>][12.30it/s][00:26>00:00] \n", + "generate label finished(142.03/s). start eval:\n", + "Evaluation official\n", + "Car AP(Average Precision)@0.70, 0.70, 0.70:\n", + "bbox AP:90.65, 88.69, 87.53\n", + "bev AP:89.99, 86.54, 79.74\n", + "3d AP:87.00, 76.88, 75.07\n", + "aos AP:0.58, 1.53, 2.31\n", + "Car AP(Average Precision)@0.70, 0.50, 0.50:\n", + "bbox AP:90.65, 88.69, 87.53\n", + "bev AP:90.73, 89.59, 88.97\n", + "3d AP:90.73, 89.49, 88.76\n", + "aos AP:0.58, 1.53, 2.31\n", + "\n", + "Evaluation coco\n", + "Car coco AP@0.50:0.05:0.95:\n", + "bbox AP:71.70, 68.48, 66.12\n", + "bev AP:69.89, 66.34, 64.51\n", + "3d AP:60.48, 56.29, 54.11\n", + "aos AP:0.42, 1.13, 1.69\n", + "\n", + "eval.kitti.official.Car.bbox@0.70=[90.65, 88.69, 87.53], eval.kitti.official.Car.bev@0.70=[89.99, 86.54, 79.74], eval.kitti.official.Car.3d@0.70=[87.0, 76.88, 75.07], eval.kitti.official.Car.aos=[0.5832, 1.53, 2.306], eval.kitti.official.Car.bev@0.50=[90.73, 89.59, 88.97], eval.kitti.official.Car.3d@0.50=[90.73, 89.49, 88.76], eval.kitti.coco.Car.bbox=[71.7, 68.48, 66.12], eval.kitti.coco.Car.bev=[69.89, 66.34, 64.51], eval.kitti.coco.Car.3d=[60.48, 56.29, 54.11], eval.kitti.coco.Car.aos=[0.423, 1.129, 1.692]\n", + "runtime.step=12450, runtime.steptime=1.042, runtime.voxel_gene_time=0.0008917, runtime.prep_time=0.02648, loss.cls_loss=0.1156, loss.cls_loss_rt=0.09811, loss.loc_loss=0.2041, loss.loc_loss_rt=0.1731, loss.loc_elem=[0.004371, 0.003438, 0.01333, 0.0121, 0.02107, 0.01948, 0.01278], loss.cls_pos_rt=0.07316, loss.cls_neg_rt=0.02496, loss.dir_rt=0.0873, rpn_acc=0.9995, pr.prec@10=0.2375, pr.rec@10=0.9553, pr.prec@30=0.7864, pr.rec@30=0.8312, pr.prec@50=0.9553, pr.rec@50=0.6707, pr.prec@70=0.9938, pr.rec@70=0.4527, pr.prec@80=0.9989, pr.rec@80=0.2981, pr.prec@90=1.0, pr.rec@90=0.07558, pr.prec@95=1.0, pr.rec@95=0.002372, misc.num_vox=179784, misc.num_pos=47, misc.num_neg=42122, misc.num_anchors=42240, misc.lr=0.0007286, misc.mem_usage=55.4\n", + "runtime.step=12500, runtime.steptime=0.2639, runtime.voxel_gene_time=0.001356, runtime.prep_time=0.0364, loss.cls_loss=0.1131, loss.cls_loss_rt=0.1117, loss.loc_loss=0.2025, loss.loc_loss_rt=0.175, loss.loc_elem=[0.004517, 0.003009, 0.01586, 0.01133, 0.01928, 0.01586, 0.01766], loss.cls_pos_rt=0.08346, loss.cls_neg_rt=0.02824, loss.dir_rt=0.1291, rpn_acc=0.9995, pr.prec@10=0.2426, pr.rec@10=0.957, pr.prec@30=0.7903, pr.rec@30=0.8357, pr.prec@50=0.9558, pr.rec@50=0.6758, pr.prec@70=0.9938, pr.rec@70=0.4591, pr.prec@80=0.999, pr.rec@80=0.3036, pr.prec@90=1.0, pr.rec@90=0.07904, pr.prec@95=1.0, pr.rec@95=0.002684, misc.num_vox=173719, misc.num_pos=62, misc.num_neg=42092, misc.num_anchors=42240, misc.lr=0.000707, misc.mem_usage=55.4\n", + "runtime.step=12550, runtime.steptime=0.2654, runtime.voxel_gene_time=0.0009787, runtime.prep_time=0.02986, loss.cls_loss=0.1124, loss.cls_loss_rt=0.1057, loss.loc_loss=0.2012, loss.loc_loss_rt=0.1942, loss.loc_elem=[0.003363, 0.003214, 0.02011, 0.01228, 0.01854, 0.02093, 0.01867], loss.cls_pos_rt=0.07634, loss.cls_neg_rt=0.02931, loss.dir_rt=0.1451, rpn_acc=0.9995, pr.prec@10=0.2449, pr.rec@10=0.9572, pr.prec@30=0.7908, pr.rec@30=0.8384, pr.prec@50=0.9558, pr.rec@50=0.6774, pr.prec@70=0.9939, pr.rec@70=0.461, pr.prec@80=0.9989, pr.rec@80=0.3066, pr.prec@90=1.0, pr.rec@90=0.08094, pr.prec@95=1.0, pr.rec@95=0.00269, misc.num_vox=175817, misc.num_pos=66, misc.num_neg=42078, misc.num_anchors=42240, misc.lr=0.0006856, misc.mem_usage=55.4\n", + "runtime.step=12600, runtime.steptime=0.266, runtime.voxel_gene_time=0.001094, runtime.prep_time=0.03217, loss.cls_loss=0.1121, loss.cls_loss_rt=0.1184, loss.loc_loss=0.2006, loss.loc_loss_rt=0.2131, loss.loc_elem=[0.004668, 0.004819, 0.02145, 0.01042, 0.02508, 0.01769, 0.02245], loss.cls_pos_rt=0.08211, loss.cls_neg_rt=0.03632, loss.dir_rt=0.1739, rpn_acc=0.9995, pr.prec@10=0.2457, pr.rec@10=0.958, pr.prec@30=0.7899, pr.rec@30=0.8387, pr.prec@50=0.9549, pr.rec@50=0.6767, pr.prec@70=0.9937, pr.rec@70=0.4583, pr.prec@80=0.9988, pr.rec@80=0.3027, pr.prec@90=1.0, pr.rec@90=0.07845, pr.prec@95=1.0, pr.rec@95=0.002715, misc.num_vox=175589, misc.num_pos=66, misc.num_neg=42062, misc.num_anchors=42240, misc.lr=0.0006645, misc.mem_usage=55.4\n", + "runtime.step=12650, runtime.steptime=0.265, runtime.voxel_gene_time=0.0009418, runtime.prep_time=0.02266, loss.cls_loss=0.1115, loss.cls_loss_rt=0.111, loss.loc_loss=0.1994, loss.loc_loss_rt=0.2072, loss.loc_elem=[0.005509, 0.004098, 0.01562, 0.01306, 0.02396, 0.01779, 0.02356], loss.cls_pos_rt=0.07526, loss.cls_neg_rt=0.03574, loss.dir_rt=0.1188, rpn_acc=0.9995, pr.prec@10=0.2464, pr.rec@10=0.9585, pr.prec@30=0.79, pr.rec@30=0.8403, pr.prec@50=0.9551, pr.rec@50=0.6784, pr.prec@70=0.9938, pr.rec@70=0.459, pr.prec@80=0.9989, pr.rec@80=0.3018, pr.prec@90=1.0, pr.rec@90=0.07626, pr.prec@95=1.0, pr.rec@95=0.00249, misc.num_vox=176405, misc.num_pos=64, misc.num_neg=42078, misc.num_anchors=42240, misc.lr=0.0006435, misc.mem_usage=55.5\n", + "reset Car\n", + "reset Car\n", + "reset Car\n", + "WORKER 0 seed: 1592961065\n", + "WORKER 1 seed: 1592961066\n", + "WORKER 2 seed: 1592961067\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "runtime.step=12700, runtime.steptime=0.431, runtime.voxel_gene_time=0.0008676, runtime.prep_time=0.0236, loss.cls_loss=0.1125, loss.cls_loss_rt=0.1106, loss.loc_loss=0.2074, loss.loc_loss_rt=0.1839, loss.loc_elem=[0.004797, 0.003819, 0.01664, 0.01388, 0.01857, 0.0194, 0.01485], loss.cls_pos_rt=0.06824, loss.cls_neg_rt=0.04232, loss.dir_rt=0.1396, rpn_acc=0.9995, pr.prec@10=0.2384, pr.rec@10=0.9555, pr.prec@30=0.7949, pr.rec@30=0.8352, pr.prec@50=0.9574, pr.rec@50=0.6813, pr.prec@70=0.9955, pr.rec@70=0.4654, pr.prec@80=0.9989, pr.rec@80=0.3121, pr.prec@90=1.0, pr.rec@90=0.08612, pr.prec@95=1.0, pr.rec@95=0.002875, misc.num_vox=178011, misc.num_pos=41, misc.num_neg=42130, misc.num_anchors=42240, misc.lr=0.0006229, misc.mem_usage=55.0\n", + "runtime.step=12750, runtime.steptime=0.2659, runtime.voxel_gene_time=0.00139, runtime.prep_time=0.03964, loss.cls_loss=0.1126, loss.cls_loss_rt=0.1197, loss.loc_loss=0.2, loss.loc_loss_rt=0.2203, loss.loc_elem=[0.004947, 0.005507, 0.01966, 0.01245, 0.02715, 0.01352, 0.02693], loss.cls_pos_rt=0.08259, loss.cls_neg_rt=0.03709, loss.dir_rt=0.1426, rpn_acc=0.9995, pr.prec@10=0.2413, pr.rec@10=0.9554, pr.prec@30=0.7913, pr.rec@30=0.8367, pr.prec@50=0.9552, pr.rec@50=0.6797, pr.prec@70=0.9953, pr.rec@70=0.4628, pr.prec@80=0.9991, pr.rec@80=0.3073, pr.prec@90=1.0, pr.rec@90=0.08134, pr.prec@95=1.0, pr.rec@95=0.002862, misc.num_vox=175335, misc.num_pos=70, misc.num_neg=42069, misc.num_anchors=42240, misc.lr=0.0006024, misc.mem_usage=55.4\n", + "runtime.step=12800, runtime.steptime=0.2692, runtime.voxel_gene_time=0.0009921, runtime.prep_time=0.03009, loss.cls_loss=0.1106, loss.cls_loss_rt=0.1123, loss.loc_loss=0.1985, loss.loc_loss_rt=0.1908, loss.loc_elem=[0.004144, 0.004545, 0.01875, 0.0129, 0.01586, 0.01647, 0.02271], loss.cls_pos_rt=0.08239, loss.cls_neg_rt=0.02994, loss.dir_rt=0.152, rpn_acc=0.9995, pr.prec@10=0.248, pr.rec@10=0.9577, pr.prec@30=0.7936, pr.rec@30=0.8399, pr.prec@50=0.9552, pr.rec@50=0.6832, pr.prec@70=0.9946, pr.rec@70=0.4678, pr.prec@80=0.999, pr.rec@80=0.3129, pr.prec@90=1.0, pr.rec@90=0.08647, pr.prec@95=1.0, pr.rec@95=0.003305, misc.num_vox=172074, misc.num_pos=65, misc.num_neg=42064, misc.num_anchors=42240, misc.lr=0.0005823, misc.mem_usage=55.4\n", + "runtime.step=12850, runtime.steptime=0.271, runtime.voxel_gene_time=0.001023, runtime.prep_time=0.03045, loss.cls_loss=0.109, loss.cls_loss_rt=0.08743, loss.loc_loss=0.1965, loss.loc_loss_rt=0.1564, loss.loc_elem=[0.005326, 0.003701, 0.01203, 0.01098, 0.01674, 0.01612, 0.0133], loss.cls_pos_rt=0.05559, loss.cls_neg_rt=0.03185, loss.dir_rt=0.1398, rpn_acc=0.9995, pr.prec@10=0.2489, pr.rec@10=0.9595, pr.prec@30=0.7948, pr.rec@30=0.843, pr.prec@50=0.9561, pr.rec@50=0.6853, pr.prec@70=0.9942, pr.rec@70=0.4705, pr.prec@80=0.9989, pr.rec@80=0.3143, pr.prec@90=1.0, pr.rec@90=0.08728, pr.prec@95=1.0, pr.rec@95=0.003444, misc.num_vox=176562, misc.num_pos=61, misc.num_neg=42103, misc.num_anchors=42240, misc.lr=0.0005624, misc.mem_usage=55.3\n", + "runtime.step=12900, runtime.steptime=0.2668, runtime.voxel_gene_time=0.001595, runtime.prep_time=0.03621, loss.cls_loss=0.1097, loss.cls_loss_rt=0.1103, loss.loc_loss=0.1977, loss.loc_loss_rt=0.1838, loss.loc_elem=[0.004391, 0.004042, 0.02005, 0.01132, 0.02003, 0.01473, 0.01736], loss.cls_pos_rt=0.0729, loss.cls_neg_rt=0.03738, loss.dir_rt=0.191, rpn_acc=0.9995, pr.prec@10=0.2486, pr.rec@10=0.9586, pr.prec@30=0.7928, pr.rec@30=0.8421, pr.prec@50=0.9563, pr.rec@50=0.683, pr.prec@70=0.9944, pr.rec@70=0.4679, pr.prec@80=0.9988, pr.rec@80=0.3125, pr.prec@90=1.0, pr.rec@90=0.08855, pr.prec@95=1.0, pr.rec@95=0.003696, misc.num_vox=179006, misc.num_pos=54, misc.num_neg=42111, misc.num_anchors=42240, misc.lr=0.0005427, misc.mem_usage=55.4\n", + "runtime.step=12950, runtime.steptime=0.2694, runtime.voxel_gene_time=0.001003, runtime.prep_time=0.02899, loss.cls_loss=0.109, loss.cls_loss_rt=0.1039, loss.loc_loss=0.1956, loss.loc_loss_rt=0.1943, loss.loc_elem=[0.004044, 0.003551, 0.0174, 0.01249, 0.02598, 0.01837, 0.01532], loss.cls_pos_rt=0.06753, loss.cls_neg_rt=0.03641, loss.dir_rt=0.1576, rpn_acc=0.9995, pr.prec@10=0.25, pr.rec@10=0.9592, pr.prec@30=0.7933, pr.rec@30=0.8432, pr.prec@50=0.9562, pr.rec@50=0.6854, pr.prec@70=0.9944, pr.rec@70=0.4715, pr.prec@80=0.9987, pr.rec@80=0.3169, pr.prec@90=1.0, pr.rec@90=0.09124, pr.prec@95=1.0, pr.rec@95=0.003953, misc.num_vox=178666, misc.num_pos=77, misc.num_neg=42044, misc.num_anchors=42240, misc.lr=0.0005234, misc.mem_usage=55.4\n", + "reset Car\n", + "reset Car\n", + "reset Car\n", + "WORKER 0 seed: 1592961156\n", + "WORKER 1 seed: 1592961157\n", + "WORKER 2 seed: 1592961158\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "runtime.step=13000, runtime.steptime=0.429, runtime.voxel_gene_time=0.0008907, runtime.prep_time=0.0265, loss.cls_loss=0.1116, loss.cls_loss_rt=0.1081, loss.loc_loss=0.1975, loss.loc_loss_rt=0.2013, loss.loc_elem=[0.003534, 0.004121, 0.01879, 0.01222, 0.02131, 0.02228, 0.01839], loss.cls_pos_rt=0.07514, loss.cls_neg_rt=0.03297, loss.dir_rt=0.1034, rpn_acc=0.9995, pr.prec@10=0.2376, pr.rec@10=0.9567, pr.prec@30=0.7884, pr.rec@30=0.835, pr.prec@50=0.9641, pr.rec@50=0.6712, pr.prec@70=0.9965, pr.rec@70=0.453, pr.prec@80=0.9996, pr.rec@80=0.3027, pr.prec@90=1.0, pr.rec@90=0.08268, pr.prec@95=1.0, pr.rec@95=0.002534, misc.num_vox=176114, misc.num_pos=64, misc.num_neg=42078, misc.num_anchors=42240, misc.lr=0.0005043, misc.mem_usage=55.3\n", + "runtime.step=13050, runtime.steptime=0.2656, runtime.voxel_gene_time=0.001432, runtime.prep_time=0.03827, loss.cls_loss=0.1136, loss.cls_loss_rt=0.08469, loss.loc_loss=0.1985, loss.loc_loss_rt=0.1786, loss.loc_elem=[0.003897, 0.003462, 0.01342, 0.01126, 0.02238, 0.01747, 0.01742], loss.cls_pos_rt=0.06258, loss.cls_neg_rt=0.02211, loss.dir_rt=0.1184, rpn_acc=0.9995, pr.prec@10=0.2359, pr.rec@10=0.9566, pr.prec@30=0.7873, pr.rec@30=0.8352, pr.prec@50=0.9569, pr.rec@50=0.6728, pr.prec@70=0.9936, pr.rec@70=0.4601, pr.prec@80=0.9985, pr.rec@80=0.3095, pr.prec@90=0.9997, pr.rec@90=0.0814, pr.prec@95=1.0, pr.rec@95=0.003028, misc.num_vox=174041, misc.num_pos=47, misc.num_neg=42130, misc.num_anchors=42240, misc.lr=0.0004855, misc.mem_usage=55.3\n", + "runtime.step=13100, runtime.steptime=0.2668, runtime.voxel_gene_time=0.0009656, runtime.prep_time=0.02646, loss.cls_loss=0.1104, loss.cls_loss_rt=0.1092, loss.loc_loss=0.1955, loss.loc_loss_rt=0.2042, loss.loc_elem=[0.004167, 0.004277, 0.01709, 0.01501, 0.02693, 0.01554, 0.0191], loss.cls_pos_rt=0.07637, loss.cls_neg_rt=0.03279, loss.dir_rt=0.1363, rpn_acc=0.9995, pr.prec@10=0.2454, pr.rec@10=0.9577, pr.prec@30=0.793, pr.rec@30=0.8417, pr.prec@50=0.9567, pr.rec@50=0.6818, pr.prec@70=0.9937, pr.rec@70=0.4681, pr.prec@80=0.9988, pr.rec@80=0.3144, pr.prec@90=0.9999, pr.rec@90=0.08155, pr.prec@95=1.0, pr.rec@95=0.002687, misc.num_vox=174797, misc.num_pos=48, misc.num_neg=42126, misc.num_anchors=42240, misc.lr=0.0004669, misc.mem_usage=55.3\n", + "runtime.step=13150, runtime.steptime=0.2639, runtime.voxel_gene_time=0.0008497, runtime.prep_time=0.02711, loss.cls_loss=0.1097, loss.cls_loss_rt=0.09395, loss.loc_loss=0.1941, loss.loc_loss_rt=0.1786, loss.loc_elem=[0.004033, 0.004046, 0.01697, 0.01165, 0.01955, 0.01723, 0.01582], loss.cls_pos_rt=0.06895, loss.cls_neg_rt=0.025, loss.dir_rt=0.1749, rpn_acc=0.9995, pr.prec@10=0.247, pr.rec@10=0.9581, pr.prec@30=0.7944, pr.rec@30=0.8426, pr.prec@50=0.9568, pr.rec@50=0.6837, pr.prec@70=0.994, pr.rec@70=0.4694, pr.prec@80=0.999, pr.rec@80=0.315, pr.prec@90=0.9999, pr.rec@90=0.08401, pr.prec@95=1.0, pr.rec@95=0.003058, misc.num_vox=179191, misc.num_pos=67, misc.num_neg=42065, misc.num_anchors=42240, misc.lr=0.0004487, misc.mem_usage=55.4\n", + "runtime.step=13200, runtime.steptime=0.2648, runtime.voxel_gene_time=0.00111, runtime.prep_time=0.03332, loss.cls_loss=0.1104, loss.cls_loss_rt=0.1224, loss.loc_loss=0.1942, loss.loc_loss_rt=0.1991, loss.loc_elem=[0.004949, 0.003605, 0.02096, 0.01271, 0.01976, 0.01634, 0.02121], loss.cls_pos_rt=0.08762, loss.cls_neg_rt=0.03479, loss.dir_rt=0.129, rpn_acc=0.9995, pr.prec@10=0.2458, pr.rec@10=0.9583, pr.prec@30=0.7928, pr.rec@30=0.8408, pr.prec@50=0.9562, pr.rec@50=0.6807, pr.prec@70=0.9938, pr.rec@70=0.4673, pr.prec@80=0.999, pr.rec@80=0.313, pr.prec@90=0.9999, pr.rec@90=0.08587, pr.prec@95=1.0, pr.rec@95=0.003068, misc.num_vox=177192, misc.num_pos=74, misc.num_neg=42060, misc.num_anchors=42240, misc.lr=0.0004308, misc.mem_usage=55.4\n", + "runtime.step=13250, runtime.steptime=0.2668, runtime.voxel_gene_time=0.0009782, runtime.prep_time=0.02902, loss.cls_loss=0.1099, loss.cls_loss_rt=0.1134, loss.loc_loss=0.1937, loss.loc_loss_rt=0.2045, loss.loc_elem=[0.005159, 0.003255, 0.01903, 0.01312, 0.02012, 0.01897, 0.02258], loss.cls_pos_rt=0.0807, loss.cls_neg_rt=0.03269, loss.dir_rt=0.1044, rpn_acc=0.9995, pr.prec@10=0.2463, pr.rec@10=0.9588, pr.prec@30=0.7925, pr.rec@30=0.8422, pr.prec@50=0.9562, pr.rec@50=0.683, pr.prec@70=0.9938, pr.rec@70=0.4683, pr.prec@80=0.999, pr.rec@80=0.3142, pr.prec@90=0.9999, pr.rec@90=0.08732, pr.prec@95=1.0, pr.rec@95=0.003291, misc.num_vox=175064, misc.num_pos=66, misc.num_neg=42075, misc.num_anchors=42240, misc.lr=0.0004132, misc.mem_usage=55.4\n", + "reset Car\n", + "reset Car\n", + "reset Car\n", + "WORKER 0 seed: 1592961246\n", + "WORKER 1 seed: 1592961247\n", + "WORKER 2 seed: 1592961248\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "runtime.step=13300, runtime.steptime=0.4283, runtime.voxel_gene_time=0.001228, runtime.prep_time=0.02442, loss.cls_loss=0.1156, loss.cls_loss_rt=0.146, loss.loc_loss=0.2034, loss.loc_loss_rt=0.2169, loss.loc_elem=[0.004945, 0.004853, 0.02085, 0.009144, 0.02069, 0.01949, 0.0285], loss.cls_pos_rt=0.1122, loss.cls_neg_rt=0.03386, loss.dir_rt=0.1573, rpn_acc=0.9995, pr.prec@10=0.2498, pr.rec@10=0.9523, pr.prec@30=0.7929, pr.rec@30=0.8314, pr.prec@50=0.9587, pr.rec@50=0.6673, pr.prec@70=0.995, pr.rec@70=0.4582, pr.prec@80=0.9996, pr.rec@80=0.3118, pr.prec@90=1.0, pr.rec@90=0.09587, pr.prec@95=1.0, pr.rec@95=0.003196, misc.num_vox=175958, misc.num_pos=41, misc.num_neg=42143, misc.num_anchors=42240, misc.lr=0.0003959, misc.mem_usage=55.1\n", + "runtime.step=13350, runtime.steptime=0.2675, runtime.voxel_gene_time=0.0009885, runtime.prep_time=0.03191, loss.cls_loss=0.1141, loss.cls_loss_rt=0.1266, loss.loc_loss=0.1976, loss.loc_loss_rt=0.2022, loss.loc_elem=[0.005272, 0.004763, 0.01709, 0.008723, 0.02575, 0.01574, 0.02373], loss.cls_pos_rt=0.0998, loss.cls_neg_rt=0.02676, loss.dir_rt=0.1804, rpn_acc=0.9995, pr.prec@10=0.2401, pr.rec@10=0.956, pr.prec@30=0.7888, pr.rec@30=0.8349, pr.prec@50=0.9534, pr.rec@50=0.6734, pr.prec@70=0.9942, pr.rec@70=0.4554, pr.prec@80=0.9991, pr.rec@80=0.3056, pr.prec@90=1.0, pr.rec@90=0.08481, pr.prec@95=1.0, pr.rec@95=0.003165, misc.num_vox=171966, misc.num_pos=70, misc.num_neg=42085, misc.num_anchors=42240, misc.lr=0.0003789, misc.mem_usage=55.3\n", + "runtime.step=13400, runtime.steptime=0.2699, runtime.voxel_gene_time=0.0009985, runtime.prep_time=0.03046, loss.cls_loss=0.1111, loss.cls_loss_rt=0.1188, loss.loc_loss=0.1952, loss.loc_loss_rt=0.1968, loss.loc_elem=[0.005076, 0.004142, 0.01831, 0.01229, 0.01828, 0.01877, 0.02151], loss.cls_pos_rt=0.09009, loss.cls_neg_rt=0.02871, loss.dir_rt=0.1595, rpn_acc=0.9995, pr.prec@10=0.2453, pr.rec@10=0.9579, pr.prec@30=0.79, pr.rec@30=0.8398, pr.prec@50=0.9554, pr.rec@50=0.6807, pr.prec@70=0.9941, pr.rec@70=0.4647, pr.prec@80=0.999, pr.rec@80=0.3137, pr.prec@90=1.0, pr.rec@90=0.09024, pr.prec@95=1.0, pr.rec@95=0.003764, misc.num_vox=178515, misc.num_pos=46, misc.num_neg=42127, misc.num_anchors=42240, misc.lr=0.0003622, misc.mem_usage=55.3\n", + "runtime.step=13450, runtime.steptime=0.2672, runtime.voxel_gene_time=0.001378, runtime.prep_time=0.03409, loss.cls_loss=0.1091, loss.cls_loss_rt=0.1204, loss.loc_loss=0.1936, loss.loc_loss_rt=0.2437, loss.loc_elem=[0.005791, 0.005029, 0.02347, 0.01182, 0.02378, 0.02004, 0.03195], loss.cls_pos_rt=0.09109, loss.cls_neg_rt=0.02926, loss.dir_rt=0.1272, rpn_acc=0.9995, pr.prec@10=0.2465, pr.rec@10=0.9598, pr.prec@30=0.793, pr.rec@30=0.842, pr.prec@50=0.9564, pr.rec@50=0.6833, pr.prec@70=0.9943, pr.rec@70=0.4696, pr.prec@80=0.9991, pr.rec@80=0.3188, pr.prec@90=1.0, pr.rec@90=0.09336, pr.prec@95=1.0, pr.rec@95=0.004313, misc.num_vox=179187, misc.num_pos=60, misc.num_neg=42084, misc.num_anchors=42240, misc.lr=0.0003459, misc.mem_usage=55.4\n", + "runtime.step=13500, runtime.steptime=0.2689, runtime.voxel_gene_time=0.00106, runtime.prep_time=0.03156, loss.cls_loss=0.1081, loss.cls_loss_rt=0.1242, loss.loc_loss=0.1934, loss.loc_loss_rt=0.2211, loss.loc_elem=[0.004731, 0.00471, 0.02068, 0.01153, 0.02315, 0.02113, 0.02459], loss.cls_pos_rt=0.09067, loss.cls_neg_rt=0.03353, loss.dir_rt=0.1762, rpn_acc=0.9995, pr.prec@10=0.2494, pr.rec@10=0.9602, pr.prec@30=0.7942, pr.rec@30=0.8441, pr.prec@50=0.9567, pr.rec@50=0.6852, pr.prec@70=0.9945, pr.rec@70=0.4722, pr.prec@80=0.999, pr.rec@80=0.3215, pr.prec@90=1.0, pr.rec@90=0.09649, pr.prec@95=1.0, pr.rec@95=0.004969, misc.num_vox=178301, misc.num_pos=68, misc.num_neg=42064, misc.num_anchors=42240, misc.lr=0.0003299, misc.mem_usage=55.4\n", + "runtime.step=13550, runtime.steptime=0.2687, runtime.voxel_gene_time=0.001036, runtime.prep_time=0.02998, loss.cls_loss=0.1084, loss.cls_loss_rt=0.1163, loss.loc_loss=0.1924, loss.loc_loss_rt=0.2078, loss.loc_elem=[0.003604, 0.004259, 0.02037, 0.01049, 0.02264, 0.02589, 0.01667], loss.cls_pos_rt=0.08677, loss.cls_neg_rt=0.02949, loss.dir_rt=0.08519, rpn_acc=0.9995, pr.prec@10=0.2506, pr.rec@10=0.9598, pr.prec@30=0.7946, pr.rec@30=0.8443, pr.prec@50=0.9562, pr.rec@50=0.6864, pr.prec@70=0.9941, pr.rec@70=0.4728, pr.prec@80=0.999, pr.rec@80=0.3216, pr.prec@90=1.0, pr.rec@90=0.09724, pr.prec@95=1.0, pr.rec@95=0.005088, misc.num_vox=178378, misc.num_pos=58, misc.num_neg=42074, misc.num_anchors=42240, misc.lr=0.0003142, misc.mem_usage=55.4\n", + "reset Car\n", + "reset Car\n", + "reset Car\n", + "WORKER 0 seed: 1592961337\n", + "WORKER 1 seed: 1592961338\n", + "WORKER 2 seed: 1592961339\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "runtime.step=13600, runtime.steptime=0.4321, runtime.voxel_gene_time=0.001863, runtime.prep_time=0.02867, loss.cls_loss=0.135, loss.cls_loss_rt=0.1485, loss.loc_loss=0.2262, loss.loc_loss_rt=0.2745, loss.loc_elem=[0.005455, 0.004576, 0.04179, 0.01711, 0.02347, 0.01617, 0.02867], loss.cls_pos_rt=0.1086, loss.cls_neg_rt=0.03998, loss.dir_rt=0.214, rpn_acc=0.9995, pr.prec@10=0.2193, pr.rec@10=0.9444, pr.prec@30=0.7599, pr.rec@30=0.8027, pr.prec@50=0.9339, pr.rec@50=0.6389, pr.prec@70=0.9906, pr.rec@70=0.4308, pr.prec@80=1.0, pr.rec@80=0.2914, pr.prec@90=1.0, pr.rec@90=0.09492, pr.prec@95=1.0, pr.rec@95=0.008528, misc.num_vox=177004, misc.num_pos=54, misc.num_neg=42102, misc.num_anchors=42240, misc.lr=0.0002988, misc.mem_usage=55.2\n", + "runtime.step=13650, runtime.steptime=0.2644, runtime.voxel_gene_time=0.0008929, runtime.prep_time=0.02603, loss.cls_loss=0.1104, loss.cls_loss_rt=0.08125, loss.loc_loss=0.194, loss.loc_loss_rt=0.1706, loss.loc_elem=[0.003804, 0.003246, 0.01504, 0.0115, 0.01692, 0.01514, 0.01962], loss.cls_pos_rt=0.05852, loss.cls_neg_rt=0.02273, loss.dir_rt=0.1233, rpn_acc=0.9995, pr.prec@10=0.2419, pr.rec@10=0.958, pr.prec@30=0.7946, pr.rec@30=0.8412, pr.prec@50=0.954, pr.rec@50=0.6856, pr.prec@70=0.9935, pr.rec@70=0.4713, pr.prec@80=0.9988, pr.rec@80=0.3168, pr.prec@90=1.0, pr.rec@90=0.08958, pr.prec@95=1.0, pr.rec@95=0.004454, misc.num_vox=178623, misc.num_pos=58, misc.num_neg=42106, misc.num_anchors=42240, misc.lr=0.0002838, misc.mem_usage=55.3\n", + "runtime.step=13700, runtime.steptime=0.2662, runtime.voxel_gene_time=0.0009339, runtime.prep_time=0.02758, loss.cls_loss=0.1089, loss.cls_loss_rt=0.1054, loss.loc_loss=0.1935, loss.loc_loss_rt=0.1835, loss.loc_elem=[0.005153, 0.003047, 0.01467, 0.01132, 0.01879, 0.01677, 0.02197], loss.cls_pos_rt=0.07382, loss.cls_neg_rt=0.03162, loss.dir_rt=0.09945, rpn_acc=0.9995, pr.prec@10=0.2457, pr.rec@10=0.96, pr.prec@30=0.7945, pr.rec@30=0.8429, pr.prec@50=0.9545, pr.rec@50=0.683, pr.prec@70=0.9939, pr.rec@70=0.4674, pr.prec@80=0.9985, pr.rec@80=0.3145, pr.prec@90=1.0, pr.rec@90=0.0911, pr.prec@95=1.0, pr.rec@95=0.004832, misc.num_vox=176397, misc.num_pos=40, misc.num_neg=42113, misc.num_anchors=42240, misc.lr=0.0002692, misc.mem_usage=55.3\n", + "runtime.step=13750, runtime.steptime=0.2648, runtime.voxel_gene_time=0.0009553, runtime.prep_time=0.02291, loss.cls_loss=0.1075, loss.cls_loss_rt=0.09439, loss.loc_loss=0.1914, loss.loc_loss_rt=0.2219, loss.loc_elem=[0.003748, 0.003666, 0.03069, 0.009701, 0.02212, 0.01862, 0.02241], loss.cls_pos_rt=0.06492, loss.cls_neg_rt=0.02947, loss.dir_rt=0.1413, rpn_acc=0.9995, pr.prec@10=0.2501, pr.rec@10=0.9607, pr.prec@30=0.7959, pr.rec@30=0.8446, pr.prec@50=0.9557, pr.rec@50=0.6869, pr.prec@70=0.9942, pr.rec@70=0.4725, pr.prec@80=0.9987, pr.rec@80=0.3194, pr.prec@90=1.0, pr.rec@90=0.09612, pr.prec@95=1.0, pr.rec@95=0.005505, misc.num_vox=177990, misc.num_pos=67, misc.num_neg=42070, misc.num_anchors=42240, misc.lr=0.0002549, misc.mem_usage=55.4\n", + "runtime.step=13800, runtime.steptime=0.2664, runtime.voxel_gene_time=0.00126, runtime.prep_time=0.03385, loss.cls_loss=0.1076, loss.cls_loss_rt=0.121, loss.loc_loss=0.1914, loss.loc_loss_rt=0.2279, loss.loc_elem=[0.004806, 0.004439, 0.02267, 0.01353, 0.02354, 0.01948, 0.02549], loss.cls_pos_rt=0.08106, loss.cls_neg_rt=0.03997, loss.dir_rt=0.1422, rpn_acc=0.9995, pr.prec@10=0.2506, pr.rec@10=0.9606, pr.prec@30=0.796, pr.rec@30=0.8449, pr.prec@50=0.956, pr.rec@50=0.6874, pr.prec@70=0.9943, pr.rec@70=0.4737, pr.prec@80=0.9987, pr.rec@80=0.3216, pr.prec@90=1.0, pr.rec@90=0.09973, pr.prec@95=1.0, pr.rec@95=0.005933, misc.num_vox=178538, misc.num_pos=53, misc.num_neg=42095, misc.num_anchors=42240, misc.lr=0.0002409, misc.mem_usage=55.4\n", + "runtime.step=13850, runtime.steptime=0.266, runtime.voxel_gene_time=0.001063, runtime.prep_time=0.03293, loss.cls_loss=0.1077, loss.cls_loss_rt=0.1522, loss.loc_loss=0.1908, loss.loc_loss_rt=0.2277, loss.loc_elem=[0.007607, 0.003398, 0.0308, 0.01337, 0.01852, 0.0152, 0.02494], loss.cls_pos_rt=0.1153, loss.cls_neg_rt=0.03685, loss.dir_rt=0.109, rpn_acc=0.9995, pr.prec@10=0.2513, pr.rec@10=0.9605, pr.prec@30=0.7953, pr.rec@30=0.8447, pr.prec@50=0.9563, pr.rec@50=0.6868, pr.prec@70=0.9941, pr.rec@70=0.4724, pr.prec@80=0.9987, pr.rec@80=0.3206, pr.prec@90=1.0, pr.rec@90=0.0991, pr.prec@95=1.0, pr.rec@95=0.005788, misc.num_vox=177118, misc.num_pos=57, misc.num_neg=42077, misc.num_anchors=42240, misc.lr=0.0002273, misc.mem_usage=55.3\n", + "reset Car\n", + "reset Car\n", + "reset Car\n", + "runtime.step=13900, runtime.steptime=0.2676, runtime.voxel_gene_time=0.001281, runtime.prep_time=0.03002, loss.cls_loss=0.1076, loss.cls_loss_rt=0.1063, loss.loc_loss=0.1904, loss.loc_loss_rt=0.1907, loss.loc_elem=[0.003849, 0.003966, 0.01925, 0.01222, 0.02081, 0.01953, 0.01571], loss.cls_pos_rt=0.07746, loss.cls_neg_rt=0.0288, loss.dir_rt=0.1188, rpn_acc=0.9995, pr.prec@10=0.2516, pr.rec@10=0.9604, pr.prec@30=0.7959, pr.rec@30=0.845, pr.prec@50=0.9568, pr.rec@50=0.6881, pr.prec@70=0.9942, pr.rec@70=0.4749, pr.prec@80=0.9988, pr.rec@80=0.3227, pr.prec@90=1.0, pr.rec@90=0.1, pr.prec@95=1.0, pr.rec@95=0.00576, misc.num_vox=176482, misc.num_pos=49, misc.num_neg=42128, misc.num_anchors=42240, misc.lr=0.0002141, misc.mem_usage=55.4\n", + "WORKER 0 seed: 1592961427\n", + "WORKER 1 seed: 1592961428\n", + "WORKER 2 seed: 1592961429\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "runtime.step=13950, runtime.steptime=0.4303, runtime.voxel_gene_time=0.0009952, runtime.prep_time=0.03602, loss.cls_loss=0.1105, loss.cls_loss_rt=0.1164, loss.loc_loss=0.2006, loss.loc_loss_rt=0.2092, loss.loc_elem=[0.006033, 0.003878, 0.01655, 0.01171, 0.02752, 0.02081, 0.0181], loss.cls_pos_rt=0.08529, loss.cls_neg_rt=0.03108, loss.dir_rt=0.1387, rpn_acc=0.9995, pr.prec@10=0.2481, pr.rec@10=0.9566, pr.prec@30=0.797, pr.rec@30=0.8423, pr.prec@50=0.9578, pr.rec@50=0.6874, pr.prec@70=0.9944, pr.rec@70=0.4757, pr.prec@80=0.9988, pr.rec@80=0.3245, pr.prec@90=1.0, pr.rec@90=0.1037, pr.prec@95=1.0, pr.rec@95=0.006936, misc.num_vox=179026, misc.num_pos=54, misc.num_neg=42104, misc.num_anchors=42240, misc.lr=0.0002012, misc.mem_usage=55.3\n", + "#################################\n", + "# EVAL\n", + "#################################\n", + "Generate output labels...\n", + "[100.0%][===================>][12.29it/s][00:26>00:00] \n", + "generate label finished(141.00/s). start eval:\n", + "Evaluation official\n", + "Car AP(Average Precision)@0.70, 0.70, 0.70:\n", + "bbox AP:90.64, 89.00, 87.52\n", + "bev AP:89.88, 86.22, 79.69\n", + "3d AP:86.61, 76.95, 74.98\n", + "aos AP:0.63, 1.58, 2.35\n", + "Car AP(Average Precision)@0.70, 0.50, 0.50:\n", + "bbox AP:90.64, 89.00, 87.52\n", + "bev AP:90.70, 89.62, 88.99\n", + "3d AP:90.70, 89.53, 88.74\n", + "aos AP:0.63, 1.58, 2.35\n", + "\n", + "Evaluation coco\n", + "Car coco AP@0.50:0.05:0.95:\n", + "bbox AP:72.52, 68.62, 66.50\n", + "bev AP:70.47, 66.61, 64.76\n", + "3d AP:60.15, 55.85, 53.67\n", + "aos AP:0.46, 1.14, 1.72\n", + "\n", + "eval.kitti.official.Car.bbox@0.70=[90.64, 89.0, 87.52], eval.kitti.official.Car.bev@0.70=[89.88, 86.22, 79.69], eval.kitti.official.Car.3d@0.70=[86.61, 76.95, 74.98], eval.kitti.official.Car.aos=[0.6284, 1.582, 2.352], eval.kitti.official.Car.bev@0.50=[90.7, 89.62, 88.99], eval.kitti.official.Car.3d@0.50=[90.7, 89.53, 88.74], eval.kitti.coco.Car.bbox=[72.52, 68.62, 66.5], eval.kitti.coco.Car.bev=[70.47, 66.61, 64.76], eval.kitti.coco.Car.3d=[60.15, 55.85, 53.67], eval.kitti.coco.Car.aos=[0.4564, 1.137, 1.721]\n", + "runtime.step=14000, runtime.steptime=1.054, runtime.voxel_gene_time=0.001119, runtime.prep_time=0.02934, loss.cls_loss=0.1064, loss.cls_loss_rt=0.1101, loss.loc_loss=0.1919, loss.loc_loss_rt=0.1988, loss.loc_elem=[0.004784, 0.004644, 0.02099, 0.01067, 0.01846, 0.01632, 0.02352], loss.cls_pos_rt=0.08303, loss.cls_neg_rt=0.02705, loss.dir_rt=0.1411, rpn_acc=0.9996, pr.prec@10=0.255, pr.rec@10=0.9592, pr.prec@30=0.8018, pr.rec@30=0.8475, pr.prec@50=0.9591, pr.rec@50=0.6942, pr.prec@70=0.9946, pr.rec@70=0.4817, pr.prec@80=0.9985, pr.rec@80=0.3289, pr.prec@90=1.0, pr.rec@90=0.1043, pr.prec@95=1.0, pr.rec@95=0.006871, misc.num_vox=170108, misc.num_pos=50, misc.num_neg=42118, misc.num_anchors=42240, misc.lr=0.0001888, misc.mem_usage=55.6\n", + "runtime.step=14050, runtime.steptime=0.2703, runtime.voxel_gene_time=0.001378, runtime.prep_time=0.03177, loss.cls_loss=0.106, loss.cls_loss_rt=0.1167, loss.loc_loss=0.1912, loss.loc_loss_rt=0.2495, loss.loc_elem=[0.005285, 0.004829, 0.03284, 0.01277, 0.02383, 0.01707, 0.02812], loss.cls_pos_rt=0.08814, loss.cls_neg_rt=0.02855, loss.dir_rt=0.1881, rpn_acc=0.9996, pr.prec@10=0.2575, pr.rec@10=0.9594, pr.prec@30=0.8014, pr.rec@30=0.8486, pr.prec@50=0.9585, pr.rec@50=0.6952, pr.prec@70=0.9948, pr.rec@70=0.4858, pr.prec@80=0.9987, pr.rec@80=0.3336, pr.prec@90=1.0, pr.rec@90=0.1093, pr.prec@95=1.0, pr.rec@95=0.00742, misc.num_vox=172971, misc.num_pos=42, misc.num_neg=42129, misc.num_anchors=42240, misc.lr=0.0001766, misc.mem_usage=55.6\n", + "runtime.step=14100, runtime.steptime=0.2654, runtime.voxel_gene_time=0.0009828, runtime.prep_time=0.02167, loss.cls_loss=0.106, loss.cls_loss_rt=0.1294, loss.loc_loss=0.1898, loss.loc_loss_rt=0.2354, loss.loc_elem=[0.005735, 0.004694, 0.02017, 0.01452, 0.02268, 0.02069, 0.02919], loss.cls_pos_rt=0.08755, loss.cls_neg_rt=0.04189, loss.dir_rt=0.1692, rpn_acc=0.9996, pr.prec@10=0.2571, pr.rec@10=0.9601, pr.prec@30=0.8007, pr.rec@30=0.8481, pr.prec@50=0.9581, pr.rec@50=0.6945, pr.prec@70=0.9946, pr.rec@70=0.4852, pr.prec@80=0.9988, pr.rec@80=0.3345, pr.prec@90=1.0, pr.rec@90=0.1106, pr.prec@95=1.0, pr.rec@95=0.007378, misc.num_vox=175818, misc.num_pos=50, misc.num_neg=42124, misc.num_anchors=42240, misc.lr=0.0001649, misc.mem_usage=55.5\n", + "runtime.step=14150, runtime.steptime=0.2673, runtime.voxel_gene_time=0.001134, runtime.prep_time=0.02485, loss.cls_loss=0.106, loss.cls_loss_rt=0.06897, loss.loc_loss=0.1895, loss.loc_loss_rt=0.1477, loss.loc_elem=[0.003782, 0.002298, 0.009543, 0.01237, 0.01908, 0.01524, 0.01154], loss.cls_pos_rt=0.04756, loss.cls_neg_rt=0.02141, loss.dir_rt=0.0558, rpn_acc=0.9995, pr.prec@10=0.2556, pr.rec@10=0.9604, pr.prec@30=0.7993, pr.rec@30=0.8485, pr.prec@50=0.9572, pr.rec@50=0.6933, pr.prec@70=0.994, pr.rec@70=0.4837, pr.prec@80=0.9985, pr.rec@80=0.3327, pr.prec@90=1.0, pr.rec@90=0.1109, pr.prec@95=1.0, pr.rec@95=0.007428, misc.num_vox=176210, misc.num_pos=59, misc.num_neg=42100, misc.num_anchors=42240, misc.lr=0.0001535, misc.mem_usage=55.5\n", + "reset Car\n", + "runtime.step=14200, runtime.steptime=0.269, runtime.voxel_gene_time=0.000912, runtime.prep_time=0.02292, loss.cls_loss=0.1057, loss.cls_loss_rt=0.1392, loss.loc_loss=0.189, loss.loc_loss_rt=0.2558, loss.loc_elem=[0.007, 0.004578, 0.02803, 0.01432, 0.02669, 0.01709, 0.03017], loss.cls_pos_rt=0.1039, loss.cls_neg_rt=0.0353, loss.dir_rt=0.171, rpn_acc=0.9995, pr.prec@10=0.2553, pr.rec@10=0.9607, pr.prec@30=0.7996, pr.rec@30=0.848, pr.prec@50=0.9577, pr.rec@50=0.6936, pr.prec@70=0.9941, pr.rec@70=0.4839, pr.prec@80=0.9987, pr.rec@80=0.3328, pr.prec@90=1.0, pr.rec@90=0.1105, pr.prec@95=1.0, pr.rec@95=0.007493, misc.num_vox=179283, misc.num_pos=52, misc.num_neg=42120, misc.num_anchors=42240, misc.lr=0.0001426, misc.mem_usage=55.5\n", + "reset Car\n", + "reset Car\n", + "WORKER 0 seed: 1592961558\n", + "WORKER 1 seed: 1592961559\n", + "WORKER 2 seed: 1592961560\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "runtime.step=14250, runtime.steptime=0.4287, runtime.voxel_gene_time=0.001444, runtime.prep_time=0.03586, loss.cls_loss=0.1093, loss.cls_loss_rt=0.1217, loss.loc_loss=0.1878, loss.loc_loss_rt=0.1949, loss.loc_elem=[0.004992, 0.004548, 0.01732, 0.01002, 0.02112, 0.02077, 0.0187], loss.cls_pos_rt=0.08923, loss.cls_neg_rt=0.03246, loss.dir_rt=0.1152, rpn_acc=0.9995, pr.prec@10=0.2515, pr.rec@10=0.9594, pr.prec@30=0.795, pr.rec@30=0.8397, pr.prec@50=0.959, pr.rec@50=0.6781, pr.prec@70=0.9926, pr.rec@70=0.4632, pr.prec@80=0.9979, pr.rec@80=0.3193, pr.prec@90=1.0, pr.rec@90=0.1067, pr.prec@95=1.0, pr.rec@95=0.007763, misc.num_vox=174543, misc.num_pos=55, misc.num_neg=42104, misc.num_anchors=42240, misc.lr=0.000132, misc.mem_usage=55.4\n", + "runtime.step=14300, runtime.steptime=0.2689, runtime.voxel_gene_time=0.001013, runtime.prep_time=0.03215, loss.cls_loss=0.1071, loss.cls_loss_rt=0.1236, loss.loc_loss=0.1868, loss.loc_loss_rt=0.21, loss.loc_elem=[0.00504, 0.003064, 0.02201, 0.01225, 0.02066, 0.01711, 0.02486], loss.cls_pos_rt=0.09104, loss.cls_neg_rt=0.03254, loss.dir_rt=0.1128, rpn_acc=0.9995, pr.prec@10=0.253, pr.rec@10=0.9604, pr.prec@30=0.7938, pr.rec@30=0.8454, pr.prec@50=0.9565, pr.rec@50=0.6902, pr.prec@70=0.9933, pr.rec@70=0.4785, pr.prec@80=0.9984, pr.rec@80=0.3301, pr.prec@90=1.0, pr.rec@90=0.1107, pr.prec@95=1.0, pr.rec@95=0.007494, misc.num_vox=177282, misc.num_pos=48, misc.num_neg=42108, misc.num_anchors=42240, misc.lr=0.0001218, misc.mem_usage=55.4\n", + "runtime.step=14350, runtime.steptime=0.2672, runtime.voxel_gene_time=0.001061, runtime.prep_time=0.03068, loss.cls_loss=0.1072, loss.cls_loss_rt=0.1153, loss.loc_loss=0.1884, loss.loc_loss_rt=0.1882, loss.loc_elem=[0.005123, 0.004313, 0.01997, 0.00963, 0.01968, 0.01312, 0.02228], loss.cls_pos_rt=0.08566, loss.cls_neg_rt=0.02967, loss.dir_rt=0.1723, rpn_acc=0.9995, pr.prec@10=0.2562, pr.rec@10=0.9595, pr.prec@30=0.7952, pr.rec@30=0.8459, pr.prec@50=0.9568, pr.rec@50=0.6912, pr.prec@70=0.994, pr.rec@70=0.4805, pr.prec@80=0.9987, pr.rec@80=0.3311, pr.prec@90=1.0, pr.rec@90=0.1096, pr.prec@95=1.0, pr.rec@95=0.007034, misc.num_vox=172811, misc.num_pos=66, misc.num_neg=42068, misc.num_anchors=42240, misc.lr=0.000112, misc.mem_usage=55.4\n", + "runtime.step=14400, runtime.steptime=0.2682, runtime.voxel_gene_time=0.001078, runtime.prep_time=0.02739, loss.cls_loss=0.1061, loss.cls_loss_rt=0.1116, loss.loc_loss=0.1862, loss.loc_loss_rt=0.1898, loss.loc_elem=[0.003915, 0.003729, 0.02141, 0.0108, 0.0194, 0.0176, 0.01804], loss.cls_pos_rt=0.08015, loss.cls_neg_rt=0.03141, loss.dir_rt=0.1435, rpn_acc=0.9995, pr.prec@10=0.2565, pr.rec@10=0.9606, pr.prec@30=0.7971, pr.rec@30=0.8475, pr.prec@50=0.9575, pr.rec@50=0.6921, pr.prec@70=0.9943, pr.rec@70=0.4819, pr.prec@80=0.9988, pr.rec@80=0.3322, pr.prec@90=1.0, pr.rec@90=0.11, pr.prec@95=1.0, pr.rec@95=0.007217, misc.num_vox=177552, misc.num_pos=50, misc.num_neg=42110, misc.num_anchors=42240, misc.lr=0.0001026, misc.mem_usage=55.4\n", + "runtime.step=14450, runtime.steptime=0.2704, runtime.voxel_gene_time=0.001299, runtime.prep_time=0.03925, loss.cls_loss=0.1066, loss.cls_loss_rt=0.1608, loss.loc_loss=0.1876, loss.loc_loss_rt=0.2631, loss.loc_elem=[0.005628, 0.004526, 0.03233, 0.01494, 0.02431, 0.01793, 0.0319], loss.cls_pos_rt=0.1201, loss.cls_neg_rt=0.04067, loss.dir_rt=0.1522, rpn_acc=0.9995, pr.prec@10=0.255, pr.rec@10=0.9603, pr.prec@30=0.7964, pr.rec@30=0.8474, pr.prec@50=0.9568, pr.rec@50=0.6925, pr.prec@70=0.9942, pr.rec@70=0.4812, pr.prec@80=0.9987, pr.rec@80=0.3307, pr.prec@90=1.0, pr.rec@90=0.1109, pr.prec@95=1.0, pr.rec@95=0.007451, misc.num_vox=175104, misc.num_pos=52, misc.num_neg=42116, misc.num_anchors=42240, misc.lr=9.358e-05, misc.mem_usage=55.4\n", + "runtime.step=14500, runtime.steptime=0.2685, runtime.voxel_gene_time=0.0009825, runtime.prep_time=0.03, loss.cls_loss=0.1061, loss.cls_loss_rt=0.1263, loss.loc_loss=0.1866, loss.loc_loss_rt=0.1672, loss.loc_elem=[0.004324, 0.004507, 0.01093, 0.01336, 0.021, 0.01595, 0.01355], loss.cls_pos_rt=0.08029, loss.cls_neg_rt=0.046, loss.dir_rt=0.1072, rpn_acc=0.9995, pr.prec@10=0.2564, pr.rec@10=0.9604, pr.prec@30=0.7975, pr.rec@30=0.848, pr.prec@50=0.9571, pr.rec@50=0.6937, pr.prec@70=0.9943, pr.rec@70=0.4839, pr.prec@80=0.9988, pr.rec@80=0.3329, pr.prec@90=1.0, pr.rec@90=0.112, pr.prec@95=1.0, pr.rec@95=0.007715, misc.num_vox=178532, misc.num_pos=69, misc.num_neg=42075, misc.num_anchors=42240, misc.lr=8.497e-05, misc.mem_usage=55.4\n", + "reset Car\n", + "reset Car\n", + "reset Car\n", + "WORKER 0 seed: 1592961649\n", + "WORKER 1 seed: 1592961650\n", + "WORKER 2 seed: 1592961651\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "runtime.step=14550, runtime.steptime=0.4324, runtime.voxel_gene_time=0.001531, runtime.prep_time=0.03618, loss.cls_loss=0.112, loss.cls_loss_rt=0.1395, loss.loc_loss=0.1933, loss.loc_loss_rt=0.2133, loss.loc_elem=[0.005274, 0.00324, 0.02197, 0.01259, 0.02206, 0.02096, 0.02054], loss.cls_pos_rt=0.09182, loss.cls_neg_rt=0.04773, loss.dir_rt=0.1378, rpn_acc=0.9995, pr.prec@10=0.2512, pr.rec@10=0.9542, pr.prec@30=0.7903, pr.rec@30=0.8377, pr.prec@50=0.9547, pr.rec@50=0.6826, pr.prec@70=0.9951, pr.rec@70=0.4729, pr.prec@80=0.9973, pr.rec@80=0.3284, pr.prec@90=1.0, pr.rec@90=0.1147, pr.prec@95=1.0, pr.rec@95=0.009015, misc.num_vox=176599, misc.num_pos=29, misc.num_neg=42161, misc.num_anchors=42240, misc.lr=7.677e-05, misc.mem_usage=55.3\n", + "runtime.step=14600, runtime.steptime=0.2666, runtime.voxel_gene_time=0.0009615, runtime.prep_time=0.02802, loss.cls_loss=0.1082, loss.cls_loss_rt=0.1106, loss.loc_loss=0.19, loss.loc_loss_rt=0.1748, loss.loc_elem=[0.004516, 0.003549, 0.01717, 0.01123, 0.02, 0.01872, 0.01221], loss.cls_pos_rt=0.07346, loss.cls_neg_rt=0.03717, loss.dir_rt=0.1053, rpn_acc=0.9995, pr.prec@10=0.2567, pr.rec@10=0.9586, pr.prec@30=0.7982, pr.rec@30=0.8421, pr.prec@50=0.9574, pr.rec@50=0.6857, pr.prec@70=0.9943, pr.rec@70=0.4782, pr.prec@80=0.9981, pr.rec@80=0.3288, pr.prec@90=1.0, pr.rec@90=0.1115, pr.prec@95=1.0, pr.rec@95=0.008391, misc.num_vox=173328, misc.num_pos=64, misc.num_neg=42075, misc.num_anchors=42240, misc.lr=6.897e-05, misc.mem_usage=55.4\n", + "runtime.step=14650, runtime.steptime=0.2677, runtime.voxel_gene_time=0.0009263, runtime.prep_time=0.03, loss.cls_loss=0.1077, loss.cls_loss_rt=0.1272, loss.loc_loss=0.1906, loss.loc_loss_rt=0.2047, loss.loc_elem=[0.005371, 0.003053, 0.02446, 0.0114, 0.02063, 0.01781, 0.01963], loss.cls_pos_rt=0.09736, loss.cls_neg_rt=0.02985, loss.dir_rt=0.16, rpn_acc=0.9995, pr.prec@10=0.257, pr.rec@10=0.9592, pr.prec@30=0.7971, pr.rec@30=0.8455, pr.prec@50=0.9567, pr.rec@50=0.6891, pr.prec@70=0.9937, pr.rec@70=0.4806, pr.prec@80=0.9982, pr.rec@80=0.3326, pr.prec@90=1.0, pr.rec@90=0.1128, pr.prec@95=1.0, pr.rec@95=0.007979, misc.num_vox=177630, misc.num_pos=51, misc.num_neg=42112, misc.num_anchors=42240, misc.lr=6.158e-05, misc.mem_usage=55.4\n", + "runtime.step=14700, runtime.steptime=0.2694, runtime.voxel_gene_time=0.001068, runtime.prep_time=0.03108, loss.cls_loss=0.1069, loss.cls_loss_rt=0.09142, loss.loc_loss=0.1886, loss.loc_loss_rt=0.1552, loss.loc_elem=[0.003416, 0.002864, 0.0131, 0.01035, 0.01934, 0.01718, 0.01136], loss.cls_pos_rt=0.05774, loss.cls_neg_rt=0.03368, loss.dir_rt=0.1352, rpn_acc=0.9995, pr.prec@10=0.2574, pr.rec@10=0.9601, pr.prec@30=0.7979, pr.rec@30=0.8465, pr.prec@50=0.9571, pr.rec@50=0.6892, pr.prec@70=0.9938, pr.rec@70=0.4803, pr.prec@80=0.9983, pr.rec@80=0.3321, pr.prec@90=1.0, pr.rec@90=0.112, pr.prec@95=1.0, pr.rec@95=0.007737, misc.num_vox=175997, misc.num_pos=66, misc.num_neg=42078, misc.num_anchors=42240, misc.lr=5.461e-05, misc.mem_usage=55.4\n", + "runtime.step=14750, runtime.steptime=0.2689, runtime.voxel_gene_time=0.001047, runtime.prep_time=0.0393, loss.cls_loss=0.1068, loss.cls_loss_rt=0.1212, loss.loc_loss=0.1881, loss.loc_loss_rt=0.2011, loss.loc_elem=[0.004759, 0.004171, 0.01676, 0.01162, 0.02054, 0.01882, 0.02385], loss.cls_pos_rt=0.08758, loss.cls_neg_rt=0.03359, loss.dir_rt=0.08437, rpn_acc=0.9995, pr.prec@10=0.2559, pr.rec@10=0.9603, pr.prec@30=0.7982, pr.rec@30=0.8467, pr.prec@50=0.957, pr.rec@50=0.6889, pr.prec@70=0.9939, pr.rec@70=0.4789, pr.prec@80=0.9983, pr.rec@80=0.3317, pr.prec@90=1.0, pr.rec@90=0.113, pr.prec@95=1.0, pr.rec@95=0.00801, misc.num_vox=169693, misc.num_pos=54, misc.num_neg=42094, misc.num_anchors=42240, misc.lr=4.804e-05, misc.mem_usage=55.4\n", + "runtime.step=14800, runtime.steptime=0.2696, runtime.voxel_gene_time=0.001509, runtime.prep_time=0.03548, loss.cls_loss=0.1058, loss.cls_loss_rt=0.09902, loss.loc_loss=0.1864, loss.loc_loss_rt=0.1795, loss.loc_elem=[0.004422, 0.004225, 0.01814, 0.01084, 0.02085, 0.01204, 0.01923], loss.cls_pos_rt=0.07297, loss.cls_neg_rt=0.02606, loss.dir_rt=0.1211, rpn_acc=0.9995, pr.prec@10=0.2572, pr.rec@10=0.9609, pr.prec@30=0.7998, pr.rec@30=0.8489, pr.prec@50=0.9576, pr.rec@50=0.6921, pr.prec@70=0.9941, pr.rec@70=0.4819, pr.prec@80=0.9983, pr.rec@80=0.3341, pr.prec@90=1.0, pr.rec@90=0.1138, pr.prec@95=1.0, pr.rec@95=0.007959, misc.num_vox=178616, misc.num_pos=58, misc.num_neg=42094, misc.num_anchors=42240, misc.lr=4.189e-05, misc.mem_usage=55.4\n", + "reset Car\n", + "reset Car\n", + "reset Car\n", + "WORKER 0 seed: 1592961740\n", + "WORKER 1 seed: 1592961741\n", + "WORKER 2 seed: 1592961742\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "runtime.step=14850, runtime.steptime=0.4291, runtime.voxel_gene_time=0.0009749, runtime.prep_time=0.02874, loss.cls_loss=0.1126, loss.cls_loss_rt=0.08549, loss.loc_loss=0.1958, loss.loc_loss_rt=0.1849, loss.loc_elem=[0.004329, 0.002781, 0.01502, 0.01176, 0.02069, 0.0193, 0.01856], loss.cls_pos_rt=0.06375, loss.cls_neg_rt=0.02174, loss.dir_rt=0.1393, rpn_acc=0.9995, pr.prec@10=0.2471, pr.rec@10=0.9522, pr.prec@30=0.7936, pr.rec@30=0.8374, pr.prec@50=0.9609, pr.rec@50=0.6834, pr.prec@70=0.995, pr.rec@70=0.483, pr.prec@80=0.9988, pr.rec@80=0.3382, pr.prec@90=1.0, pr.rec@90=0.1193, pr.prec@95=1.0, pr.rec@95=0.01074, misc.num_vox=178186, misc.num_pos=65, misc.num_neg=42081, misc.num_anchors=42240, misc.lr=3.615e-05, misc.mem_usage=55.3\n", + "runtime.step=14900, runtime.steptime=0.2688, runtime.voxel_gene_time=0.001144, runtime.prep_time=0.03078, loss.cls_loss=0.1078, loss.cls_loss_rt=0.09481, loss.loc_loss=0.1886, loss.loc_loss_rt=0.1785, loss.loc_elem=[0.004083, 0.004967, 0.01341, 0.01067, 0.01976, 0.01946, 0.01688], loss.cls_pos_rt=0.06044, loss.cls_neg_rt=0.03437, loss.dir_rt=0.07004, rpn_acc=0.9995, pr.prec@10=0.2522, pr.rec@10=0.9577, pr.prec@30=0.7968, pr.rec@30=0.8452, pr.prec@50=0.9593, pr.rec@50=0.6881, pr.prec@70=0.9948, pr.rec@70=0.4854, pr.prec@80=0.9987, pr.rec@80=0.3431, pr.prec@90=1.0, pr.rec@90=0.1223, pr.prec@95=1.0, pr.rec@95=0.01097, misc.num_vox=179272, misc.num_pos=59, misc.num_neg=42087, misc.num_anchors=42240, misc.lr=3.084e-05, misc.mem_usage=55.3\n", + "runtime.step=14950, runtime.steptime=0.2673, runtime.voxel_gene_time=0.0008748, runtime.prep_time=0.02554, loss.cls_loss=0.1053, loss.cls_loss_rt=0.1116, loss.loc_loss=0.1871, loss.loc_loss_rt=0.2308, loss.loc_elem=[0.005742, 0.003565, 0.01613, 0.01297, 0.02489, 0.01703, 0.0351], loss.cls_pos_rt=0.08553, loss.cls_neg_rt=0.02603, loss.dir_rt=0.1309, rpn_acc=0.9995, pr.prec@10=0.2555, pr.rec@10=0.961, pr.prec@30=0.7956, pr.rec@30=0.8497, pr.prec@50=0.9586, pr.rec@50=0.6927, pr.prec@70=0.9944, pr.rec@70=0.4897, pr.prec@80=0.9987, pr.rec@80=0.3454, pr.prec@90=1.0, pr.rec@90=0.123, pr.prec@95=1.0, pr.rec@95=0.009902, misc.num_vox=174860, misc.num_pos=62, misc.num_neg=42083, misc.num_anchors=42240, misc.lr=2.594e-05, misc.mem_usage=55.3\n", + "runtime.step=15000, runtime.steptime=0.2697, runtime.voxel_gene_time=0.0009065, runtime.prep_time=0.02515, loss.cls_loss=0.1056, loss.cls_loss_rt=0.1126, loss.loc_loss=0.1859, loss.loc_loss_rt=0.1915, loss.loc_elem=[0.004126, 0.003409, 0.02227, 0.01267, 0.02059, 0.01578, 0.01693], loss.cls_pos_rt=0.07817, loss.cls_neg_rt=0.03447, loss.dir_rt=0.2285, rpn_acc=0.9995, pr.prec@10=0.2551, pr.rec@10=0.9609, pr.prec@30=0.7964, pr.rec@30=0.8482, pr.prec@50=0.9593, pr.rec@50=0.6914, pr.prec@70=0.9944, pr.rec@70=0.486, pr.prec@80=0.9987, pr.rec@80=0.3403, pr.prec@90=1.0, pr.rec@90=0.1202, pr.prec@95=1.0, pr.rec@95=0.009446, misc.num_vox=179316, misc.num_pos=61, misc.num_neg=42085, misc.num_anchors=42240, misc.lr=2.146e-05, misc.mem_usage=55.3\n", + "runtime.step=15050, runtime.steptime=0.2701, runtime.voxel_gene_time=0.000999, runtime.prep_time=0.03071, loss.cls_loss=0.1064, loss.cls_loss_rt=0.129, loss.loc_loss=0.1865, loss.loc_loss_rt=0.2343, loss.loc_elem=[0.003866, 0.005727, 0.02604, 0.01427, 0.02119, 0.01863, 0.02742], loss.cls_pos_rt=0.09597, loss.cls_neg_rt=0.033, loss.dir_rt=0.115, rpn_acc=0.9995, pr.prec@10=0.2559, pr.rec@10=0.9602, pr.prec@30=0.7963, pr.rec@30=0.8469, pr.prec@50=0.9579, pr.rec@50=0.6902, pr.prec@70=0.9942, pr.rec@70=0.4848, pr.prec@80=0.9985, pr.rec@80=0.3388, pr.prec@90=1.0, pr.rec@90=0.1195, pr.prec@95=1.0, pr.rec@95=0.009135, misc.num_vox=172294, misc.num_pos=60, misc.num_neg=42082, misc.num_anchors=42240, misc.lr=1.74e-05, misc.mem_usage=55.4\n", + "runtime.step=15100, runtime.steptime=0.2678, runtime.voxel_gene_time=0.001003, runtime.prep_time=0.03536, loss.cls_loss=0.1059, loss.cls_loss_rt=0.109, loss.loc_loss=0.1863, loss.loc_loss_rt=0.1801, loss.loc_elem=[0.004413, 0.002898, 0.01662, 0.01151, 0.01768, 0.01714, 0.01976], loss.cls_pos_rt=0.07959, loss.cls_neg_rt=0.02943, loss.dir_rt=0.1032, rpn_acc=0.9995, pr.prec@10=0.2572, pr.rec@10=0.9609, pr.prec@30=0.7977, pr.rec@30=0.8475, pr.prec@50=0.958, pr.rec@50=0.691, pr.prec@70=0.9941, pr.rec@70=0.4844, pr.prec@80=0.9985, pr.rec@80=0.3383, pr.prec@90=1.0, pr.rec@90=0.1191, pr.prec@95=1.0, pr.rec@95=0.00884, misc.num_vox=174115, misc.num_pos=55, misc.num_neg=42092, misc.num_anchors=42240, misc.lr=1.377e-05, misc.mem_usage=55.4\n", + "reset Car\n", + "reset Car\n", + "reset Car\n", + "WORKER 0 seed: 1592961831\n", + "WORKER 1 seed: 1592961832\n", + "WORKER 2 seed: 1592961833\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "runtime.step=15150, runtime.steptime=0.4326, runtime.voxel_gene_time=0.001524, runtime.prep_time=0.02342, loss.cls_loss=0.1199, loss.cls_loss_rt=0.1088, loss.loc_loss=0.1987, loss.loc_loss_rt=0.2015, loss.loc_elem=[0.004958, 0.003445, 0.02143, 0.0111, 0.02433, 0.01613, 0.01936], loss.cls_pos_rt=0.07962, loss.cls_neg_rt=0.02914, loss.dir_rt=0.1168, rpn_acc=0.9995, pr.prec@10=0.2452, pr.rec@10=0.9536, pr.prec@30=0.789, pr.rec@30=0.831, pr.prec@50=0.947, pr.rec@50=0.6698, pr.prec@70=0.9905, pr.rec@70=0.463, pr.prec@80=0.9965, pr.rec@80=0.3251, pr.prec@90=1.0, pr.rec@90=0.1169, pr.prec@95=1.0, pr.rec@95=0.009533, misc.num_vox=176847, misc.num_pos=68, misc.num_neg=42078, misc.num_anchors=42240, misc.lr=1.056e-05, misc.mem_usage=55.1\n", + "runtime.step=15200, runtime.steptime=0.2678, runtime.voxel_gene_time=0.001011, runtime.prep_time=0.03311, loss.cls_loss=0.1101, loss.cls_loss_rt=0.09845, loss.loc_loss=0.1887, loss.loc_loss_rt=0.2032, loss.loc_elem=[0.004782, 0.003176, 0.0168, 0.01448, 0.02288, 0.01959, 0.01986], loss.cls_pos_rt=0.06913, loss.cls_neg_rt=0.02932, loss.dir_rt=0.145, rpn_acc=0.9995, pr.prec@10=0.2525, pr.rec@10=0.9601, pr.prec@30=0.7922, pr.rec@30=0.8425, pr.prec@50=0.9534, pr.rec@50=0.6868, pr.prec@70=0.9943, pr.rec@70=0.4769, pr.prec@80=0.9987, pr.rec@80=0.333, pr.prec@90=1.0, pr.rec@90=0.121, pr.prec@95=1.0, pr.rec@95=0.01053, misc.num_vox=180000, misc.num_pos=71, misc.num_neg=42045, misc.num_anchors=42240, misc.lr=7.777e-06, misc.mem_usage=55.3\n", + "runtime.step=15250, runtime.steptime=0.2697, runtime.voxel_gene_time=0.0009892, runtime.prep_time=0.023, loss.cls_loss=0.1094, loss.cls_loss_rt=0.1365, loss.loc_loss=0.1899, loss.loc_loss_rt=0.2003, loss.loc_elem=[0.006023, 0.003302, 0.018, 0.01269, 0.01447, 0.01994, 0.02571], loss.cls_pos_rt=0.1023, loss.cls_neg_rt=0.03425, loss.dir_rt=0.1292, rpn_acc=0.9995, pr.prec@10=0.2558, pr.rec@10=0.9592, pr.prec@30=0.7944, pr.rec@30=0.8429, pr.prec@50=0.9559, pr.rec@50=0.6875, pr.prec@70=0.9941, pr.rec@70=0.4801, pr.prec@80=0.9991, pr.rec@80=0.336, pr.prec@90=1.0, pr.rec@90=0.1184, pr.prec@95=1.0, pr.rec@95=0.009362, misc.num_vox=178964, misc.num_pos=61, misc.num_neg=42089, misc.num_anchors=42240, misc.lr=5.419e-06, misc.mem_usage=55.4\n", + "runtime.step=15300, runtime.steptime=0.2669, runtime.voxel_gene_time=0.0009654, runtime.prep_time=0.02924, loss.cls_loss=0.1075, loss.cls_loss_rt=0.08779, loss.loc_loss=0.1882, loss.loc_loss_rt=0.1735, loss.loc_elem=[0.004327, 0.00403, 0.01838, 0.01147, 0.01855, 0.01483, 0.01515], loss.cls_pos_rt=0.06234, loss.cls_neg_rt=0.02545, loss.dir_rt=0.1062, rpn_acc=0.9995, pr.prec@10=0.2562, pr.rec@10=0.9594, pr.prec@30=0.7973, pr.rec@30=0.8467, pr.prec@50=0.9571, pr.rec@50=0.6926, pr.prec@70=0.9941, pr.rec@70=0.4835, pr.prec@80=0.999, pr.rec@80=0.3377, pr.prec@90=1.0, pr.rec@90=0.1182, pr.prec@95=1.0, pr.rec@95=0.009078, misc.num_vox=170847, misc.num_pos=48, misc.num_neg=42127, misc.num_anchors=42240, misc.lr=3.486e-06, misc.mem_usage=55.3\n", + "runtime.step=15350, runtime.steptime=0.2692, runtime.voxel_gene_time=0.001094, runtime.prep_time=0.03437, loss.cls_loss=0.1076, loss.cls_loss_rt=0.09598, loss.loc_loss=0.1883, loss.loc_loss_rt=0.1849, loss.loc_elem=[0.004213, 0.003596, 0.01601, 0.01277, 0.02553, 0.01569, 0.01465], loss.cls_pos_rt=0.07228, loss.cls_neg_rt=0.0237, loss.dir_rt=0.09781, rpn_acc=0.9995, pr.prec@10=0.2564, pr.rec@10=0.9589, pr.prec@30=0.7978, pr.rec@30=0.846, pr.prec@50=0.9578, pr.rec@50=0.6914, pr.prec@70=0.9948, pr.rec@70=0.4826, pr.prec@80=0.9991, pr.rec@80=0.3376, pr.prec@90=1.0, pr.rec@90=0.1182, pr.prec@95=1.0, pr.rec@95=0.009364, misc.num_vox=178448, misc.num_pos=52, misc.num_neg=42121, misc.num_anchors=42240, misc.lr=1.981e-06, misc.mem_usage=55.4\n", + "runtime.step=15400, runtime.steptime=0.2693, runtime.voxel_gene_time=0.0009584, runtime.prep_time=0.03506, loss.cls_loss=0.1077, loss.cls_loss_rt=0.1236, loss.loc_loss=0.188, loss.loc_loss_rt=0.2159, loss.loc_elem=[0.003842, 0.004257, 0.01677, 0.01353, 0.02144, 0.0197, 0.0284], loss.cls_pos_rt=0.09091, loss.cls_neg_rt=0.0327, loss.dir_rt=0.1089, rpn_acc=0.9995, pr.prec@10=0.2566, pr.rec@10=0.959, pr.prec@30=0.7982, pr.rec@30=0.8457, pr.prec@50=0.9579, pr.rec@50=0.6911, pr.prec@70=0.9946, pr.rec@70=0.482, pr.prec@80=0.999, pr.rec@80=0.3362, pr.prec@90=1.0, pr.rec@90=0.1178, pr.prec@95=1.0, pr.rec@95=0.009443, misc.num_vox=170164, misc.num_pos=63, misc.num_neg=42085, misc.num_anchors=42240, misc.lr=9.03e-07, misc.mem_usage=55.4\n", + "reset Car\n", + "reset Car\n", + "reset Car\n", + "runtime.step=15450, runtime.steptime=0.2692, runtime.voxel_gene_time=0.001006, runtime.prep_time=0.03447, loss.cls_loss=0.1067, loss.cls_loss_rt=0.108, loss.loc_loss=0.1876, loss.loc_loss_rt=0.1969, loss.loc_elem=[0.004383, 0.004288, 0.01768, 0.0097, 0.02157, 0.02021, 0.02063], loss.cls_pos_rt=0.07685, loss.cls_neg_rt=0.03114, loss.dir_rt=0.1377, rpn_acc=0.9995, pr.prec@10=0.2568, pr.rec@10=0.9603, pr.prec@30=0.7982, pr.rec@30=0.8472, pr.prec@50=0.958, pr.rec@50=0.6917, pr.prec@70=0.9945, pr.rec@70=0.4831, pr.prec@80=0.999, pr.rec@80=0.337, pr.prec@90=1.0, pr.rec@90=0.1174, pr.prec@95=1.0, pr.rec@95=0.009199, misc.num_vox=180000, misc.num_pos=68, misc.num_neg=42077, misc.num_anchors=42240, misc.lr=2.526e-07, misc.mem_usage=55.4\n", + "WORKER 0 seed: 1592961922\n", + "WORKER 1 seed: 1592961923\n", + "WORKER 2 seed: 1592961924\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/opt/second.pytorch/second/core/preprocess.py:463: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " points[i:i + 1, :3] = points[i:i + 1, :3] @ rot_mat_T[j]\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typing/npydecl.py:952: NumbaPerformanceWarning: '@' is faster on contiguous arrays, called on (array(float32, 2d, A), array(float32, 2d, C))\n", + " warnings.warn(NumbaPerformanceWarning(msg))\n", + "runtime.step=15500, runtime.steptime=0.4305, runtime.voxel_gene_time=0.001017, runtime.prep_time=0.02523, loss.cls_loss=0.1087, loss.cls_loss_rt=0.09563, loss.loc_loss=0.1882, loss.loc_loss_rt=0.1528, loss.loc_elem=[0.003433, 0.003271, 0.01307, 0.01004, 0.01743, 0.01858, 0.0106], loss.cls_pos_rt=0.06248, loss.cls_neg_rt=0.03315, loss.dir_rt=0.1297, rpn_acc=0.9995, pr.prec@10=0.2553, pr.rec@10=0.9574, pr.prec@30=0.7943, pr.rec@30=0.8424, pr.prec@50=0.9553, pr.rec@50=0.6939, pr.prec@70=0.9934, pr.rec@70=0.4873, pr.prec@80=0.9987, pr.rec@80=0.3409, pr.prec@90=0.9998, pr.rec@90=0.1244, pr.prec@95=1.0, pr.rec@95=0.01018, misc.num_vox=175931, misc.num_pos=59, misc.num_neg=42086, misc.num_anchors=42240, misc.lr=3.009e-08, misc.mem_usage=55.3\n", + "#################################\n", + "# EVAL\n", + "#################################\n", + "Generate output labels...\n", + "[100.0%][===================>][12.27it/s][00:26>00:00] \n", + "generate label finished(141.54/s). start eval:\n", + "Evaluation official\n", + "Car AP(Average Precision)@0.70, 0.70, 0.70:\n", + "bbox AP:90.62, 88.96, 87.32\n", + "bev AP:89.85, 86.12, 86.10\n", + "3d AP:86.92, 77.14, 74.97\n", + "aos AP:0.64, 1.53, 2.22\n", + "Car AP(Average Precision)@0.70, 0.50, 0.50:\n", + "bbox AP:90.62, 88.96, 87.32\n", + "bev AP:90.69, 89.61, 88.96\n", + "3d AP:90.69, 89.52, 88.74\n", + "aos AP:0.64, 1.53, 2.22\n", + "\n", + "Evaluation coco\n", + "Car coco AP@0.50:0.05:0.95:\n", + "bbox AP:71.44, 67.96, 65.91\n", + "bev AP:69.96, 66.54, 64.77\n", + "3d AP:59.77, 55.47, 53.13\n", + "aos AP:0.47, 1.09, 1.64\n", + "\n", + "eval.kitti.official.Car.bbox@0.70=[90.62, 88.96, 87.32], eval.kitti.official.Car.bev@0.70=[89.85, 86.12, 86.1], eval.kitti.official.Car.3d@0.70=[86.92, 77.14, 74.97], eval.kitti.official.Car.aos=[0.6422, 1.533, 2.218], eval.kitti.official.Car.bev@0.50=[90.69, 89.61, 88.96], eval.kitti.official.Car.3d@0.50=[90.69, 89.52, 88.74], eval.kitti.coco.Car.bbox=[71.44, 67.96, 65.91], eval.kitti.coco.Car.bev=[69.96, 66.54, 64.77], eval.kitti.coco.Car.3d=[59.77, 55.47, 53.13], eval.kitti.coco.Car.aos=[0.47, 1.092, 1.638]\n" + ] + } + ], "source": [ "!python ${SECOND_API}/pytorch/train.py train --config_path=${MODEL_OUT_PATH}/train.config --model_dir=${MODEL_OUT_PATH}/${MODEL_NAME_FLAT}" ] @@ -287,7 +2452,7 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 21, "metadata": { "scrolled": true }, @@ -305,9 +2470,53 @@ }, { "cell_type": "code", - "execution_count": null, + "execution_count": 22, "metadata": {}, - "outputs": [], + "outputs": [ + { + "name": "stdout", + "output_type": "stream", + "text": [ + "[ 41 1280 1056]\n", + "Restoring parameters from /data/kitti_lidar/trained_model/car_lite/voxelnet-15500.tckpt\n", + "feature_map_size [1, 160, 132]\n", + "remain number of infos: 3769\n", + "Generate output labels...\n", + "[100.0%][===================>][12.30it/s][00:26>00:00] \n", + "generate label finished(141.54/s). start eval:\n", + "/usr/local/lib/python3.6/dist-packages/numba/core/typed_passes.py:314: NumbaPerformanceWarning: \n", + "The keyword argument 'parallel=True' was specified but no transformation for parallel execution was possible.\n", + "\n", + "To find out why, try turning on parallel diagnostics, see http://numba.pydata.org/numba-doc/latest/user/parallel.html#diagnostics for help.\n", + "\n", + "File \"../../opt/second.pytorch/second/utils/eval.py\", line 129:\n", + "@numba.jit(nopython=True, parallel=True)\n", + "def box3d_overlap_kernel(boxes,\n", + "^\n", + "\n", + " state.func_ir.loc))\n", + "Evaluation official\n", + "Car AP(Average Precision)@0.70, 0.70, 0.70:\n", + "bbox AP:90.62, 88.96, 87.32\n", + "bev AP:89.85, 86.12, 86.10\n", + "3d AP:86.92, 77.14, 74.97\n", + "aos AP:0.64, 1.53, 2.22\n", + "Car AP(Average Precision)@0.70, 0.50, 0.50:\n", + "bbox AP:90.62, 88.96, 87.32\n", + "bev AP:90.69, 89.61, 88.96\n", + "3d AP:90.69, 89.52, 88.74\n", + "aos AP:0.64, 1.53, 2.22\n", + "\n", + "Evaluation coco\n", + "Car coco AP@0.50:0.05:0.95:\n", + "bbox AP:71.44, 67.96, 65.91\n", + "bev AP:69.96, 66.54, 64.77\n", + "3d AP:59.77, 55.47, 53.13\n", + "aos AP:0.47, 1.09, 1.64\n", + "\n" + ] + } + ], "source": [ "!python ${SECOND_API}/pytorch/train.py evaluate --config_path=${MODEL_OUT_PATH}/train.config --model_dir=${MODEL_OUT_PATH}/${MODEL_NAME_FLAT}" ]