Added parallel device usage for GPT-J#22713
Merged
Merged
Conversation
|
The documentation is not available anymore as the PR was closed or merged. |
novice03
pushed a commit
to novice03/transformers
that referenced
this pull request
Jun 23, 2023
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
What does this PR do?
This PR is within the issue 22561, and is related to issue 22535 which concerns model parallelism. Specifically, this PR fixes the issue in GPT-J where tensors might accidentally be moved to different devices, causing a mismatch. The implemented fix ensures that all tensors are on the same device, preventing potential errors.
Test case:
`
#Setting up the tokenizer and model
tokenizer = GPT2Tokenizer.from_pretrained("EleutherAI/gpt-j-6B")
model = GPTJForSequenceClassification.from_pretrained("EleutherAI/gpt-j-6B")
#Now move the model to the GPU
model.to("cuda:0")
#setting up the text
text = "this is an example of text for device mismatch for GPT-J"
inputs = tokenizer(text,return_tensors = "pt")
#I've already move the model to Cuda:0
for k,v in inputs.items():
inputs[k] = v.to('cuda:0')
labels = torch.tensor([1]).to('cpu')
#Forward pass
outputs = model(**inputs,labels = labels)`
I recreated the issue by running the code without the fix, which resulted in the following error: "RuntimeError: Expected all tensors to be on the same device, ...". After implementing the fix, the error disappeared, and the model now keeps all tensors on the same device, as expected.
Fixes # 22561
Motivation and Context
I worked on helping with the code to make all transformers compatible with model parallelism, specifically GPT-J.
Who can review?
@sgugger