Add support to export ColPali Model to ONNX #2074
Open
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
|
@fxmarty, @echarlaix, @JingyaHuang, @michaelbenayoun Are you open to merging this? |
|
Apologies for the delay @akshayballal95, could you add a test with a tiny random model like https://huggingface.co/hf-internal-testing/tiny-random-PaliGemmaForConditionalGeneration, can be added here https://github.com/huggingface/optimum/blob/main/tests/exporters/exporters_utils.py#L37 |
echarlaix
reviewed
Dec 9, 2024
Comment on lines
+512
to
+527
| class ColPaliModelPatcher(ModelPatcher): | ||
| def __init__( | ||
| self, | ||
| config: "OnnxConfig", | ||
| model: Union["PreTrainedModel", "TFPreTrainedModel"], | ||
| model_kwargs: Optional[Dict[str, Any]] = None, | ||
| ): | ||
| super().__init__(config, model, model_kwargs) | ||
|
|
||
| def patched_forward(input_ids=None, pixel_values=None, attention_mask=None, **kwargs): | ||
| outputs = self.orig_forward( | ||
| input_ids=input_ids, pixel_values=pixel_values, attention_mask=attention_mask, **kwargs | ||
| ) | ||
| return outputs | ||
|
|
||
| self.patched_forward = patched_forward |
Comment on lines
+1175
to
+1189
|
|
||
| class ColPaliModelPatcher(ModelPatcher): | ||
| def __init__( | ||
| self, | ||
| config: "OnnxConfig", | ||
| model: Union["PreTrainedModel", "TFPreTrainedModel"], | ||
| model_kwargs: Optional[Dict[str, Any]] = None, | ||
| ): | ||
| super().__init__(config, model, model_kwargs) | ||
| def patched_forward(input_ids=None, pixel_values=None, attention_mask=None, **kwargs): | ||
| outputs = self.orig_forward( | ||
| input_ids=input_ids, pixel_values=pixel_values, attention_mask=attention_mask, **kwargs | ||
| ) | ||
| return outputs | ||
| self.patched_forward = patched_forward |
There was a problem hiding this comment.
already added above
Suggested change
| class ColPaliModelPatcher(ModelPatcher): | |
| def __init__( | |
| self, | |
| config: "OnnxConfig", | |
| model: Union["PreTrainedModel", "TFPreTrainedModel"], | |
| model_kwargs: Optional[Dict[str, Any]] = None, | |
| ): | |
| super().__init__(config, model, model_kwargs) | |
| def patched_forward(input_ids=None, pixel_values=None, attention_mask=None, **kwargs): | |
| outputs = self.orig_forward( | |
| input_ids=input_ids, pixel_values=pixel_values, attention_mask=attention_mask, **kwargs | |
| ) | |
| return outputs | |
| self.patched_forward = patched_forward |
| ) | ||
| dummy_inputs["input_ids"] = generator.concat_inputs([prefix_tensor, dummy_inputs["input_ids"]], dim=1) | ||
| dummy_inputs["attention_mask"] = generator.random_mask_tensor( | ||
| shape=[generator.batch_size, generator.sequence_length + 1024], |
There was a problem hiding this comment.
where does the value 1024 comes from? shouldn't it depend from the models config ?
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 adds support for exporting the ColPali merged model to ONNX format. The model is based on the "pali gemma" model type, and thus, I have added it under the "feature-extraction" task. Do suggest if there is a better way to integrate this. If this looks fine with a few modifications, I can add support for the Paligemma text-generation task as well.
Before submitting
Who can review?
@fxmarty, @echarlaix, @JingyaHuang, @michaelbenayoun