MisconfigurationException: Trainer(strategy='ddp')
is not compatible with an interactive environment.
#413
Labels
bug
Something isn't working
Describe the bug
Default distribution strategy of
Trainer
inModel
isddp
which is incompatible with notebook environments. We can set the distribution strategy using configs forModel.fit()
, but not for Model.Predict(). Therefore, fresh models that have state dicts loaded into them won't run in notebooks as they default toddp.
To Reproduce
Steps to reproduce the behavior:
Follow steps in #403 but in notebook.
Expected behavior
We should have control over this argument for
Model.predict
too.Full traceback
Please include the full error message to allow for debugging
The text was updated successfully, but these errors were encountered: