ENH [AutoQuantizer]: enhance trainer + not supported quant methods
#28991
Conversation
This file contains 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
|
The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update. |
amyeroberts
approved these changes
Feb 13, 2024
src/transformers/modeling_utils.py
Outdated
Comment on lines
4195
to
4198
| logger.warning( | ||
| "`_is_quantized_training_enabled` is going to be deprecated in a future version. Please use `model.hf_quantizer.is_trainable` instead", | ||
| FutureWarning, | ||
| ) |
There was a problem hiding this comment.
Nice and safe :) Tbh, with private methods we can probably get away with no deprecation warning as it's not public.
There was a problem hiding this comment.
Yeah in my first commits I just removed that private attribute, but after giving it some thoughts I realised maybe better to go for that option just to be on the safe zone
src/transformers/modeling_utils.py
Outdated
| @property | ||
| def _is_quantized_training_enabled(self): | ||
| logger.warning( | ||
| "`_is_quantized_training_enabled` is going to be deprecated in a future version. Please use `model.hf_quantizer.is_trainable` instead", |
There was a problem hiding this comment.
Message should have a specific version listed here for deprecation
There was a problem hiding this comment.
ok perfect, I will set it on the next 2 minor releases
sbucaille
pushed a commit
to sbucaille/transformers
that referenced
this pull request
Feb 14, 2024
…uggingface#28991) * enhance trainer + not support quant methods * remove all old logic * add version
5 tasks
itazap
pushed a commit
that referenced
this pull request
May 14, 2024
…28991) * enhance trainer + not support quant methods * remove all old logic * add version
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?
Currently, if a quantization method do not support PEFT fine-tuning an old error with bistandbytes is raised:
Regardless of the quantization method. In fact, for example if one uses AWQ + Trainer (which is not supported yet, but soon with #28987 / huggingface/peft#1399), they'll get the old error which is very confusing.
Moreover, we should rely on the variable
hf_quantizer.is_trainableinstead of_is_quantized_training_enabledWe should instead be more precise and throw an error that states why this is not supported and how to request a fix.
cc @amyeroberts