Repository navigation
Add SigLIP to NormalizedConfigManager - #2516
Open
kratos0718 wants to merge 1 commit into
Open
kratos0718 wants to merge 1 commit into
kratos0718 wants to merge 1 commit into
Conversation
…ionConfig fix SigLIP has an ONNX export config but was missing from NormalizedConfigManager. Same dual text/vision encoder shape as GroupViT/OwlViT/OwlV2/Pix2Struct, registered via Pix2StructNormalizedTextConfig. Carries the same NormalizedTextAndVisionConfig.__getattr__ fix as huggingface#2513, huggingface#2514 and huggingface#2515 for the same reason: SigLIP's text_config only has num_hidden_layers, not a num_layers alias.
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
Part of #351. SigLIP already has an ONNX export config but was missing from
NormalizedConfigManager."siglip": Pix2StructNormalizedTextConfigto_conf(alphabetically, betweensegformerandspeech_to_text)siglipfrom theTODO: missing normalized configscommentSame dual text/vision encoder shape as GroupViT (#2513), OwlViT (#2514) and OwlV2 (#2515), all also open — carries the same
NormalizedTextAndVisionConfig.__getattr__fix, for the same reason. See #2513 for the full explanation. That's the whole batch of models with this composite shape now covered.Test plan
test_siglip_normalizes_expected_attributes, instantiating a realSiglipConfig()and assertinghidden_size/num_attention_heads/num_layers/image_size/num_channelsresolve correctlypython -m unittest) — all 3 tests pass