Fix OOMs happening in case of accelerate >= 0.16.0#310
Merged
Conversation
mryab
approved these changes
Apr 25, 2023
mryab
left a comment
mryab
left a comment
Member
There was a problem hiding this comment.
Thanks for a quick investigation!
borzunov
commented
Apr 25, 2023
| torch>=1.12 | ||
| bitsandbytes==0.38.0.post2 | ||
| accelerate>=0.15.0,<1.0.0 | ||
| accelerate>=0.16.0,<1.0.0 |
Collaborator
Author
There was a problem hiding this comment.
set_module_tensor_to_device's dtype arg didn't exist before 0.16.0
borzunov
changed the title
borzunov
changed the title
borzunov
changed the title
fthrvi
pushed a commit
to fthrvi/nakshatra
that referenced
this pull request
Jun 1, 2026
…op#310) - After bigscience-workshop#285, `load_pretrained_block()` uses `accelerate.utils.set_module_tensor_to_device()` - In accelerate>=0.16.0, it saves the tensor in the dtype previously used by the model instead of dtype of the weights (huggingface/accelerate#920) - Because of that, blocks and attention caches used float32, which caused OOMs - This PR makes `load_pretrained_block()` respect `torch_dtype` (default: `"auto"`, which means reading `torch_dtype` from `config.json`)
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
2 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
load_pretrained_block()usesaccelerate.utils.set_module_tensor_to_device()load_checkpointhuggingface/accelerate#920)load_pretrained_block()respecttorch_dtype(default:"auto", which means readingtorch_dtypefromconfig.json)