Basic command to run the code using a YAML config file:
python3 bin/run.py config/test_uresnet.cfg
For more details and to use the API, you can look at this Jupyter notebook.
For your inspiration, the following are available in the config
folder:
test_chain_ppn.cfg
UResNet + PPN as two separate modules, can load separate weights for UResNet only.test_chain.cfg
Tests the chain UResNet + PPN + DBSCAN for clustering purposes.test_uresnet_ppn.cfg
UResNet + PPN as a monolithic model.test_uresnet.cfg
UResNet alone.
Typically in a configuration file you want to edit:
batch_size
(in 2 places)weight_prefix
log_dir
iterations
model_path
train
gpus
If you want more, you can use analysis_keys
, analysis
(scripts) and outputs
(formatters)
to store events in CSV format and run your custom analysis scripts (see folder analysis
).