Why Bad Behavior

[Netceor]

I can run the program successfully, using the your parameters but it shows bad. Collision-rate is 1 and no success.
It can not perform as well as it did in the link https://bark-simulator.github.io/tutorials/bark_ml_getting_started/.
Do I need to retrain? I trained yesterday, but the results still don't seem good

:~/Project/bark-ml$ bazel run //examples:tfa_gnn
INFO: Analyzed target //examples:tfa_gnn (0 packages loaded, 0 targets configured).
INFO: Found 1 target...
Target //examples:tfa_gnn up-to-date:
bazel-bin/examples/tfa_gnn
INFO: Elapsed time: 0.202s, Critical Path: 0.00s
INFO: 1 process: 1 internal.
INFO: Build completed successfully, 1 total action
INFO: Running command line: external/bazel_tools/tools/test/test-setup.sh examplINFO: Build completed successfully, 1 total action
exec ${PAGER:-/usr/bin/less} "$0" || exit 1
Executing tests from //examples:tfa_gnn
2022-07-21 09:06:25.154909: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:975] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2022-07-21 09:06:25.175083: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:975] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2022-07-21 09:06:25.175223: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:975] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
I0721 09:06:25.176883 139873437524800 xodr_parser.py:317] Transforming PlanView with given offset {'x': 0.0, 'y': -0.4, 'z': 0.0, 'hdg': 0.0}
I0721 09:06:25.178073 139873437524800 xodr_parser.py:317] Transforming PlanView with given offset {'x': 0.0, 'y': -0.4, 'z': 0.0, 'hdg': 0.0}
I0721 09:06:26.265485 139873437524800 graph_observer.py:77] GraphObserver configured with node attributes: ['x', 'y', 'theta', 'vel', 'goal_x', 'goal_y', 'goal_dx', 'goal_dy', 'goal_theta', 'goal_d', 'goal_vel']
I0721 09:06:26.265587 139873437524800 graph_observer.py:92] GraphObserver configured with edge attributes: ['dx', 'dy', 'dvel', 'dtheta']
2022-07-21 09:06:26.268973: I tensorflow/core/platform/cpu_feature_guard.cc:193] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations: AVX2 FMA
To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.
2022-07-21 09:06:26.269864: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:975] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2022-07-21 09:06:26.270013: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:975] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2022-07-21 09:06:26.270111: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:975] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2022-07-21 09:06:26.627559: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:975] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2022-07-21 09:06:26.627707: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:975] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2022-07-21 09:06:26.627814: I tensorflow/stream_executor/cuda/cuda_gpu_executor.cc:975] successful NUMA node read from SysFS had negative value (-1), but there must be at least one NUMA node, so returning NUMA node zero
2022-07-21 09:06:26.627906: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1532] Created device /job:localhost/replica:0/task:0/device:GPU:0 with 20880 MB memory: -> device: 0, name: NVIDIA GeForce RTX 3090, pci bus id: 0000:08:00.0, compute capability: 8.6
2022-07-21 09:06:27.487744: I tensorflow/stream_executor/cuda/cuda_blas.cc:1786] TensorFloat-32 will be used for the matrix multiplication. This will only be logged once.
/home/myx/anaconda3/envs/bark-ml/lib/python3.7/site-packages/gym/spaces/box.py:84: UserWarning: WARN: Box bound precision lowered by casting to float32
logger.warn(f"Box bound precision lowered by casting to {self.dtype}")
I0721 09:06:27.772876 139873437524800 common.py:1007] No checkpoint available at
I0721 09:06:27.773449 139873437524800 common.py:1007] No checkpoint available at best_checkpoint/
WARNING:tensorflow:From /home/myx/anaconda3/envs/bark-ml/lib/python3.7/site-packages/tensorflow/python/autograph/impl/api.py:377: ReplayBuffer.get_next (from tf_agents.replay_buffers.replay_buffer) is deprecated and will be removed in a future version.
Instructions for updating:
Use as_dataset(..., single_deterministic_pass=False) instead. W0721 09:06:27.844532 139873437524800 api.py:459] From /home/myx/anaconda3/envs/bark-ml/lib/python3.7/site-packages/tensorflow/python/autograph/impl/api.py:377: ReplayBuffer.get_next (from tf_agents.replay_buffers.replay_buffer) is deprecated and will be removed in a future version. Instructions for updating: Use as_dataset(..., single_deterministic_pass=False) instead.
I0721 09:06:28.160851 139873437524800 tfa_runner.py:150] Simulating episode 0.
I0721 09:06:29.264624 139873437524800 tfa_runner.py:150] Simulating episode 1.
I0721 09:06:29.730968 139873437524800 tfa_runner.py:150] Simulating episode 2.
I0721 09:06:30.606807 139873437524800 tfa_runner.py:150] Simulating episode 3.
I0721 09:06:31.119165 139873437524800 tfa_runner.py:150] Simulating episode 4.
I0721 09:06:31.633056 139873437524800 tfa_runner.py:150] Simulating episode 5.
I0721 09:06:32.500190 139873437524800 tfa_runner.py:150] Simulating episode 6.
I0721 09:06:34.434643 139873437524800 tfa_runner.py:150] Simulating episode 7.
I0721 09:06:34.834882 139873437524800 tfa_runner.py:150] Simulating episode 8.
I0721 09:06:35.307642 139873437524800 tfa_runner.py:150] Simulating episode 9.
The agent achieved an average reward of -0.185, collision-rate of 1.00000, took on average 12.300 steps, and reached a success-rate of 0.000 (evaluated over 10 episodes).`

[CesarLiu]

Hi,
this question is not that clear. There is no pre-trained model in the repo. If you just run this example using "bazel run //examples:tfa_gnn", the default mode is "visualize". It is quite reasonable, that the network achieves zero success without training.
Please add more details.
Best