aws-deepracer-community / deepracer-for-cloud

Creates an AWS DeepRacing training environment which can be deployed in the cloud, or locally on Ubuntu Linux, Windows or Mac.
MIT No Attribution
325 stars 176 forks source link

feat: Add feature to store custom_files and config in a subdir for each experiment #165

Open mattcamp opened 2 months ago

mattcamp commented 2 months ago

The feature enables each training session to have it's config and custom_files stored within a subdir under experiments/.

This simplifies being able to locate the config and files that were used for a particular training session.

An additional flag is added to system.env to control the experiment name, which when set will cause the rest of the config to be loaded from $DR_DIR/experiments/<experiment_name>/run.env and $DR_DIR/experiments/<experiment_name>/custom_files/.

larsll commented 2 months ago

Question; does things like dr-increment-training and the -1 flag to dr-upload-model work?

mattcamp commented 2 months ago

Question; does things like dr-increment-training and the -1 flag to dr-upload-model work?

Yes, they work fine, although the increment-training will just update the current run.env inside the same experiment folder.

I plan to add another command to increment the whole experiment folder which will also update the DR_EXPERIMENT_NAME and increment model prefix + set pretrained prefix.