Fix gemma3 export from a vlm checkpoint #90
Merged
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Outdated suggestions cannot be applied.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.
@thewh1teagle this PR allows to correctly export with text-generation(-with-past) task even when loading from a vlm checkpoint.
it in the last PR there was only a causal lm checkpoint in the onnxruntime tests, and a vlm checkpoint in the exporters tests (which I wasn't aware of its failure)
@echarlaix apparently with gemma3, there's no way to make the model output the past_key_values (even when all configs and subconfigs have their use_cache attribute set to True) because of this line
I have made changes so that we can pass use_cache as an argument instead, this seems to be more reliable and removes the need for patching a couple models where a specific use_cache had to be set (seq2seq models and trocr).