Add DataParallel and make Block support DataParallel #87
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
There's a lambda function in Block that causes errors when trying to train with DataParallel enabled. I'm not sure if you want to add DataParallel to minGPT because maybe that's not minimal enough. Could reduce this to just make the change to
mingpt/model.py
to eliminate the error if people use DataParallel with this model.Example error:
Removing the lambda function resolves it, although I don't fully understand why.