Remove many output_attentions and other traced outputs on 100+ models

[molbap]

What does this PR do?

In model additions, we often see old standards not using check_model_inputs, can_return_tuple and it's often a first review comment/something that can slip through. Doing a wide scan to try to remove all occurrences systematically.

Background

Every model used to manually resolve output_attentions, output_hidden_states, and return_dict in each forward, then collect intermediate outputs in a loop, then convert to tuple at the end. That's ~30 lines of boilerplate per model, reimplemented everywhere with subtle inconsistencies.

Two decorators now handle this:

• @capture_outputs goes on the base model forward (the one with the layer loop). It reads output_attentions/output_hidden_states from kwargs or config, installs hooks on modules listed in _can_record_outputs, collects intermediate outputs automatically, injects them into the ModelOutput, and handles return_dict. The model just needs to declare which module classes produce which outputs (e.g. _can_record_outputs = {"hidden_states": DecoderLayer, "attentions": Attention}).

• @can_return_tuple goes on wrapper forwards (ForCausalLM, ForSequenceClassification, VLM wrappers) that only need return_dict conversion. Wrapper models should not use @capture_outputs to avoid nested hook chains.

What changes per model

• output_attentions, output_hidden_states, return_dict dropped from forward signatures, replaced by **kwargs: Unpack[TransformersKwargs]
• Explicit parameter resolution lines removed
• Manual all_hidden_states += (hidden_states,) collection loops removed
• Decoder layers return a single tensor instead of a tuple
• Attention modules always return (attn_output, attn_weights) — the if not output_attentions: attn_weights = None guard is removed since hooks capture directly from the module output

[HuggingFaceDocBuilderDev]

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.

[github-actions]

Workflow Run ⚙️

This comment contains run-slow, running the specified jobs:

models: ["models/roberta", "models/roberta_prelayernorm", "models/roc_bert", "models/sam", "models/speech_to_text", "models/splinter", "models/stablelm", "models/time_series_transformer", "models/timesfm2_5", "models/timm_wrapper", "models/video_llama_3", "models/video_llava", "models/videomae"]
quantizations: []

[github-actions]

CI Results

Workflow Run ⚙️

Commit Info

Context	Commit	Description
RUN	68de84a6	workflow commit (merge commit)
PR	fdaa667f	branch commit (from PR)
main	535f289d	base commit (on main)

Model CI Report

❌ 5 new failed tests from this PR 😭

• time_series_transformer:
  tests/models/time_series_transformer/test_modeling_time_series_transformer.py::TimeSeriesTransformerModelTest::test_forward_signature (✅ ⟹ ❌)

• video_llama_3:
  tests/models/video_llama_3/test_modeling_video_llama_3.py::VideoLlama3IntegrationTest::test_small_model_integration_test (❌ ⟹ ❌)
  tests/models/video_llama_3/test_modeling_video_llama_3.py::VideoLlama3IntegrationTest::test_small_model_integration_test_batch (❌ ⟹ ❌)
  tests/models/video_llama_3/test_modeling_video_llama_3.py::VideoLlama3IntegrationTest::test_small_model_integration_test_batch_different_resolutions (❌ ⟹ ❌)
  tests/models/video_llama_3/test_modeling_video_llama_3.py::VideoLlama3IntegrationTest::test_small_model_integration_test_batch_wo_image (❌ ⟹ ❌)

[vasqu]

run-slow: vipllava,vit_mae,vit_msn,vitpose_backbone,vivit,vjepa2,voxtral_realtime,xglm,xlm_roberta,xlm_roberta_xl,xlstm,xmod,yolos,zamba,qwen2_5_omni,qwen2_audio,qwen2_vl,qwen3_5

[github-actions]

Workflow Run ⚙️

This comment contains run-slow, running the specified jobs:

models: ["models/qwen2_5_omni", "models/qwen2_audio", "models/qwen2_vl", "models/qwen3_5", "models/vipllava", "models/vit_mae", "models/vit_msn", "models/vitpose_backbone", "models/vivit", "models/vjepa2", "models/voxtral_realtime", "models/xglm", "models/xlm_roberta", "models/xlm_roberta_xl", "models/xlstm", "models/xmod", "models/yolos", "models/zamba"]
quantizations: []

[github-actions]

CI Results

Workflow Run ⚙️

Commit Info

Context	Commit	Description
RUN	68de84a6	workflow commit (merge commit)
PR	fdaa667f	branch commit (from PR)
main	535f289d	base commit (on main)

✅ No failing test specific to this PR 🎉 👏 !

[vasqu]

run-slow: qwen3_5_moe,qwen3_omni_moe,qwen3_vl,qwen3_vl_moe

[github-actions]

Workflow Run ⚙️

This comment contains run-slow, running the specified jobs:

models: ["models/qwen3_5_moe", "models/qwen3_omni_moe", "models/qwen3_vl", "models/qwen3_vl_moe"]
quantizations: []

[github-actions]

[For maintainers] Suggested jobs to run (before merge)

run-slow: aimv2, align, altclip, apertus, aria, audio_spectrogram_transformer, audioflamingo3, autoformer, aya_vision, bamba, bart, beit, bert, bert_generation, big_bird, bigbird_pegasus

[github-actions]

[For maintainers] Suggested jobs to run (before merge)

run-slow: aimv2, align, altclip, apertus, aria, audio_spectrogram_transformer, audioflamingo3, autoformer, aya_vision, bamba, bart, beit, bert, bert_generation, big_bird, bigbird_pegasus

[github-actions]

[For maintainers] Suggested jobs to run (before merge)

run-slow: aimv2, align, altclip, apertus, aria, audio_spectrogram_transformer, audioflamingo3, autoformer, aya_vision, bamba, bart, beit, bert, bert_generation, big_bird, bigbird_pegasus

[github-actions]

[For maintainers] Suggested jobs to run (before merge)

run-slow: aimv2, align, altclip, apertus, aria, audio_spectrogram_transformer, audioflamingo3, autoformer, aya_vision, bamba, bart, beit, bert, bert_generation, big_bird, bigbird_pegasus

[github-actions]

CI Results

Workflow Run ⚙️

Commit Info

Context	Commit	Description
RUN	68de84a6	workflow commit (merge commit)
PR	fdaa667f	branch commit (from PR)
main	535f289d	base commit (on main)

⚠️ Model CI failed to report results

The test failure analysis could not be completed. Please check the workflow run for details.

[github-actions]

[For maintainers] Suggested jobs to run (before merge)

run-slow: aimv2, align, altclip, apertus, aria, audio_spectrogram_transformer, audioflamingo3, autoformer, aya_vision, bamba, bart, beit, bert, bert_generation, big_bird, bigbird_pegasus

[vasqu]

Force merging flaky test, and also important for a model addition (and several other refactors)