You can use WandB to track your training runs to gain useful insights into metrics like the eval loss and VRAM usage.
Each TrainArgs child class, such as GENTrainArgs has three parameters that are used to control the WandB tracking.
|report_to||()||Platforms where metrics will be saved to|
|project_name||“happy-transformer”||Project name within WandB|
|run_name||“test”||run name within WandB|
from happytransformer import GENTrainArgs args = GENTrainArgs(report_to=tuple(["wandb"]), )