transformers

d72aef09 - [serge] Fix 12 integration tests for model `edgetam` failing with `import_or_config` (other (12)) (#48322)

[serge] Fix 12 integration tests for model `edgetam` failing with `import_or_config` (other (12)) (#48322) * [serge] Fix 12 integration tests for model `edgetam` failing with `import_or_config` (other (12)) * [EdgeTAM] Move original_sizes pop to EdgeTamModel.get_image_features Pop `original_sizes` at the `get_image_features` entry point (per @zucchini-nlp review) instead of deep inside `EdgeTamVisionModel.forward`. This is the correct interception point: `original_sizes` is a post-processing artifact returned by the processor and should never reach the timm FeatureListNet backbone, which does not accept it. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * Fix modular conversion check: add full type annotations to get_image_features Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * [EdgeTAM] Refresh golden values for test_inference_mask_generation_batched_images_multi_points Stale values for scores2/masks_logits2 were introduced by c694707483 (synthetic test assets PR), which replaced the truck image in prepare_image() but left edgetam untouched (the test was failing with the original_sizes error at the time). Values confirmed stable across two independent runner runs. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * Update expected outputs in EdgeTam batched multi-point integration test scores2 and masks_logits2 values changed with the updated torch/model checkpoint on current runners. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> --------- Co-authored-by: serge[bot] <serge[bot]@users.noreply.github.com> Co-authored-by: ydshieh <ydshieh@users.noreply.github.com> Co-authored-by: Yih-Dar <2521628+ydshieh@users.noreply.github.com>