Could be as simple as adding a new field tracker_kwargs that is added to init_kwargs when calling Accelerator.init_trackers.
|
init_kwargs={ |
|
"wandb": {"name": self.run_name}, |
|
"mlflow": {"run_name": self.run_name} |
|
}, |
This solution does not work if we want to support multiple trackers at once in the future.
Could be as simple as adding a new field
tracker_kwargsthat is added toinit_kwargswhen callingAccelerator.init_trackers.frogbox/frogbox/pipelines/pipeline.py
Lines 97 to 100 in edb9d71
This solution does not work if we want to support multiple trackers at once in the future.