This PR improves the CLI interface by making the default values for the CLI arguments to be retrieved from the config file and adding the following CLI arguments:
batch_size: Batch size for training
val_path: Path to val.txt
log_path: Path to store tensorbordX log files
checkpoint_path: Path to store checkpoints
config: Path to a custom yaml config file that overrides cfg.py
The benefit of this PR is that training from CLI is simpler and more versatile. The optional possibility to pass in a custom yaml config file from the command line also makes it great for remote deployment.
I hope you see the benefit and believe it is an improvement. Please contact me with any questions and feel free to edit the code.
This PR improves the CLI interface by making the default values for the CLI arguments to be retrieved from the config file and adding the following CLI arguments:
The benefit of this PR is that training from CLI is simpler and more versatile. The optional possibility to pass in a custom yaml config file from the command line also makes it great for remote deployment.
I hope you see the benefit and believe it is an improvement. Please contact me with any questions and feel free to edit the code.