diff --git a/analysis/current/analysis-report-hybrid.json b/analysis/current/analysis-report-hybrid.json index 59efcc3070f06b519a78d0ffffea51ecf1a90625..94ce5ee822ed2dd88041faad6b2a22d3bb3cafc7 100644 --- a/analysis/current/analysis-report-hybrid.json +++ b/analysis/current/analysis-report-hybrid.json @@ -1,8 +1,8 @@ { "schema_version": "1.0", "repo": "huggingface/transformers", - "snapshot_id": "20260506T120019Z", - "generated_at": "2026-05-06T12:08:31Z", + "snapshot_id": "20260506T180045Z", + "generated_at": "2026-05-06T18:15:23Z", "evidence_quality": "full", "llm_enrichment": true, "meta_bugs": [ diff --git a/analysis/current/analysis-report-hybrid.llm-reviews.json b/analysis/current/analysis-report-hybrid.llm-reviews.json index 6b094d5e2c8d03e9f03d99a72f42725937955979..f79fecff12cfa0fdf7e433cc4edcb3b7cd8dee0c 100644 --- a/analysis/current/analysis-report-hybrid.llm-reviews.json +++ b/analysis/current/analysis-report-hybrid.llm-reviews.json @@ -1,8 +1,8 @@ { "schema_version": "1.0", "repo": "huggingface/transformers", - "snapshot_id": "20260506T120019Z", - "generated_at": "2026-05-06T12:08:31Z", + "snapshot_id": "20260506T180045Z", + "generated_at": "2026-05-06T18:15:23Z", "model": "gpt-5.4-mini?service_tier=flex", "reviews": [ { @@ -26,12 +26,12 @@ "issue:13244|issue:16618", "issue:15980|issue:16618" ], - "prepared_review_unit_hash": "ac92f70755fb4554ee819df382a68030fa39e5126100746d019060de84808fd7", + "prepared_review_unit_hash": "285d43af9708d76b766e40051da8365659a39c2e42962ae4c4220e916293123f", "component_budget": { "node_count": 3, "item_count": 3, "soft_pair_count": 2, - "serialized_chars": 2479, + "serialized_chars": 2478, "estimated_input_tokens": 620, "estimated_eval_tokens": 1496 }, @@ -39,7 +39,7 @@ "node_count": 3, "item_count": 3, "soft_pair_count": 2, - "serialized_chars": 2479, + "serialized_chars": 2478, "estimated_input_tokens": 620, "estimated_eval_tokens": 1496 }, @@ -48,24 +48,24 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The cluster is not a true duplicate set. One issue is about Tapas tokenization, one is a private-model auth/error-message problem, and one is a generic model-loading failure. The only somewhat adjacent pair is 15980 and 16618, but they do not clearly describe the same bug.", - "confidence": 0.81, - "canonical_issue_reason": "issue #16618 is the broadest and most representative model-loading complaint in the set, while #15980 is specifically about private-model authentication/error text and #13244 is unrelated tokenization behavior.", + "summary": "This cluster looks heterogeneous: one issue is about Tapas tokenization differences, one about private-model auth/error messaging, and one about failing to load bert-base-uncased. The only plausible anchor is the generic model-loading report, but the soft links do not justify deduping the issues together.", + "confidence": 0.92, + "canonical_issue_reason": "issue:16618 is the most central and broadly phrased report in the cluster, and it is the only node softly linked to both other issues, so it is the best anchor even though the topics are likely distinct.", "canonical_pr_reason": null, - "best_issue_reason": "#16618 is the best single issue to represent the cluster only because it is the most general model-loading report; however, it should not be treated as a strong duplicate target for the other two.", + "best_issue_reason": "issue:16618 is the best overall issue candidate because it is the most general failure description and sits at the center of the soft-similarity links.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:13244", "right": "issue:16618", "accept": false, - "reason": "Tapas tokenization behavior is unrelated to a failure to load bert-base-uncased; different subsystem and different bug." + "reason": "Tokenization mismatch in Tapas and inability to load bert-base-uncased are different problem types with no clear shared code-path or fix." }, { "left": "issue:15980", "right": "issue:16618", "accept": false, - "reason": "Both mention loading models, but #15980 is specifically about private-model download/login handling and #16618 is a generic load failure for a public model; not the same concrete bug." + "reason": "Private-model login/error-message handling and a generic model-loading failure are not the same underlying bug; the overlap is too broad to merge." } ] }, @@ -186,8 +186,8 @@ "issue:44560|issue:45685", "issue:43039|issue:43828", "issue:30277|issue:36331", - "issue:43039|issue:44805", "issue:34928|issue:36010", + "issue:43039|issue:44805", "issue:41762|issue:43596", "issue:43901|issue:44512", "issue:39290|issue:41720", @@ -208,364 +208,352 @@ "issue:44841|issue:45084", "issue:41093|issue:41762" ], - "prepared_review_unit_hash": "defa533fd181f7631e8881425979bb98b83800cc6641287fde432b7e2ea4e9ff", + "prepared_review_unit_hash": "05f15a0f5f953554c2204d90c90d285906aade22f4def3724a0442b01d248647", "component_budget": { "node_count": 387, "item_count": 387, "soft_pair_count": 1455, - "serialized_chars": 749552, - "estimated_input_tokens": 187388, - "estimated_eval_tokens": 375032 + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 68, - "serialized_chars": 47975, - "estimated_input_tokens": 11994, - "estimated_eval_tokens": 24244 + "serialized_chars": 47982, + "estimated_input_tokens": 11996, + "estimated_eval_tokens": 24248 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly heterogeneous: several unrelated model/runtime regressions and a couple of docs issues. The only clear duplicate family is the DeepSpeed ZeRO-3 index-out-of-bounds / empty-tensor load failure around Bert/Gemma3-style model init; most other soft pairs should stay separate.", - "confidence": 0.84, - "canonical_issue_reason": "issue:43596 is the clearest representative of the shared ZeRO-3 empty-index/out-of-bounds load failure, with tight near-duplicates in 43638 and related manifestation in 41762.", + "summary": "This set is mostly a loose mix of unrelated issues. The only clear duplicate pair is the two DeepSpeed ZeRO-3/BERT reports; the rest are at best broad theme matches (docs cleanup, MoE, dtype/config errors) but not the same concrete bug.", + "confidence": 0.85, + "canonical_issue_reason": "issue:43596 is the cleaner canonical report for the ZeRO-3/BERT empty-index failure: it states the general load/init error, while issue:43638 reads like the same bug in a narrower non-pretrained-BERT variant.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43596 best anchors the only strong duplicate family in this set and is broad enough to represent the underlying DeepSpeed ZeRO-3 init bug.", + "best_issue_reason": "issue:43596 is the best global issue anchor because it most clearly captures the shared underlying bug among the only true duplicate pair and is the broader, more reusable description.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44322", "right": "issue:45464", "accept": false, - "reason": "Different Qwen3.5 failures: config attribute error vs streaming/chat API crash." + "reason": "Different Qwen3.5 failures: one is a missing config attribute, the other is a streaming chat/completions runtime issue." }, { "left": "issue:45237", "right": "issue:45685", "accept": false, - "reason": "Different bugs: GPT-OSS/AMD GPU failure vs MPS histogram kernel error." + "reason": "Different model/hardware paths; GPT-OSS on AMD GPUs is unrelated to an MPS histogram kernel limitation." }, { "left": "issue:43716", "right": "issue:45237", "accept": false, - "reason": "Different model-specific dtype/preprocessor bug vs GPT-OSS GPU incompatibility." + "reason": "Different bug classes: image preprocessor dtype mismatch versus a GPT-OSS GPU compatibility failure." }, { "left": "issue:43828", "right": "issue:45237", "accept": false, - "reason": "Different models and code paths; only a broad runtime theme is shared." + "reason": "Autocast dtype mismatch and AMD GPU failure are not the same concrete problem." }, { "left": "issue:45237", "right": "issue:45507", "accept": false, - "reason": "Different failures: GPU incompatibility vs invalid method call in GraniteMoEHybrid." + "reason": "Different models and code paths; this is not one shared underlying defect." }, { "left": "issue:44292", "right": "issue:45237", "accept": false, - "reason": "Different models and symptoms; not the same underlying bug." + "reason": "Qwen-3 NVFP4 runtime failure is unrelated to GPT-OSS AMD GPU support." }, { "left": "issue:30277", "right": "issue:35545", "accept": false, - "reason": "DeepSpeed collective mismatch is unrelated to ONNX export." + "reason": "Collective mismatch under ZeRO-3 and ONNX export failure are distinct issues." }, { "left": "issue:35545", "right": "issue:42915", "accept": false, - "reason": "ModernBERT ONNX export and Qwen3Moe FP8 are unrelated bugs." + "reason": "ModernBERT ONNX export and Qwen3 MoE FP8 loading are unrelated change requests/bugs." }, { "left": "issue:35545", "right": "issue:41720", "accept": false, - "reason": "Export failure vs auto device-map cuda assert are different code paths." + "reason": "ONNX export and CUDA assert with auto device mapping are different failures." }, { "left": "issue:44509", "right": "issue:44512", "accept": false, - "reason": "Both are docs issues, but they reference different removed APIs/commands." + "reason": "Both are v5 docs cleanups, but they refer to different removed documentation surfaces and are not the same bug." }, { "left": "issue:35545", "right": "issue:36010", "accept": false, - "reason": "ONNX export regression is unrelated to the GenerationMixin import error." + "reason": "ModernBERT export error and GenerationMixin import error have no shared concrete code path." }, { "left": "issue:43638", "right": "issue:44805", "accept": false, - "reason": "Different bugs: ZeRO-3 empty-index failure vs mask/tensor shape mismatch." + "reason": "Both are shape/index errors, but the reports point to different symptoms and contexts; not enough to treat as the same bug." }, { "left": "issue:43596", "right": "issue:44805", "accept": false, - "reason": "Different failure families; one is ZeRO-3 init, the other is mask indexing." + "reason": "ZeRO-3/BERT empty-index loading and a generic mask-shape mismatch are too different to merge." }, { "left": "issue:44805", "right": "issue:45161", "accept": false, - "reason": "Mask shape mismatch is unrelated to GPT-OSS tensor-parallel failure." + "reason": "A mask/tensor shape error is unrelated to GPT-OSS tensor-parallel failure." }, { "left": "issue:44805", "right": "issue:44918", "accept": false, - "reason": "Different training-time bugs and different code paths." + "reason": "Mask shape mismatch and Qwen3.5 embedding unpacking under TRL are different problems." }, { "left": "issue:44661", "right": "issue:44805", "accept": false, - "reason": "Tokenizer mapping/add-new-model issue is unrelated to mask indexing." + "reason": "Tokenizer mapping helper failure is unrelated to the mask-shape index error." }, { "left": "issue:34634", "right": "issue:35545", "accept": false, - "reason": "BarkProcessor preset handling and ModernBERT export are unrelated." + "reason": "BarkProcessor voice preset behavior and ModernBERT ONNX export are unrelated." }, { "left": "issue:30277", "right": "issue:43638", "accept": false, - "reason": "Collective mismatch vs ZeRO-3 out-of-bounds are different bugs." + "reason": "Both involve DeepSpeed, but one is a collective mismatch and the other is a BERT indexing failure." }, { "left": "issue:43596", "right": "issue:43638", "accept": true, - "reason": "Same ZeRO-3 empty-index/out-of-bounds load failure on closely related Bert model paths." + "reason": "Same underlying ZeRO-3/BERT zero-sized indexing bug; 43638 is just a narrower description of the same failure mode." }, { "left": "issue:43975", "right": "issue:44457", "accept": false, - "reason": "Different text detokenization bug vs LoRA merge/save output mismatch." + "reason": "Detokenization in v5 and LoRA merge/save inconsistency are different model-behavior bugs." }, { "left": "issue:43039", "right": "issue:45161", "accept": false, - "reason": "Liger Kernel cross-entropy call and GPT-OSS TP issue are unrelated." + "reason": "Liger Kernel cross-entropy invocation is unrelated to GPT-OSS MoE tensor-parallel behavior." }, { "left": "issue:44292", "right": "issue:44457", "accept": false, - "reason": "NVFP4 runtime error and LoRA merge mismatch are unrelated." + "reason": "Qwen NVFP4 runtime issues and LoRA merge/save output mismatch are not the same bug." }, { "left": "issue:34928", "right": "issue:44805", "accept": false, - "reason": "FSDP/checkpointing tensor-size mismatch is not the same as mask indexing." + "reason": "Activation-checkpointing/FSDP recomputation mismatch is a different failure class from the mask-shape index error." }, { "left": "issue:24643", "right": "issue:30277", "accept": false, - "reason": "Weight dimensionality error in training is unrelated to collective mismatch." + "reason": "Both mention DeepSpeed, but the runtime weight-dimension error and collective mismatch are different code-path failures." }, { "left": "issue:43638", "right": "issue:45161", "accept": false, - "reason": "ZeRO-3 BERT load bug vs GPT-OSS TP failure are different code paths." + "reason": "BERT ZeRO-3 loading failure and GPT-OSS tensor-parallel failure are unrelated." }, { "left": "issue:34928", "right": "issue:36331", "accept": false, - "reason": "Checkpointing/FSDP issue is unrelated to the CustomTrainer kwarg regression." + "reason": "FSDP checkpointing mismatch and CustomTrainer signature change are unrelated issues." }, { "left": "issue:43596", "right": "issue:45161", "accept": false, - "reason": "Different bugs: ZeRO-3 empty-index load failure vs GPT-OSS TP issue." + "reason": "BERT ZeRO-3 index error is unrelated to GPT-OSS MoE TP behavior." }, { "left": "issue:43716", "right": "issue:45507", "accept": false, - "reason": "Image preprocessor dtype mismatch vs invalid method call are unrelated." + "reason": "Image preprocessor dtype mismatch and GraniteMoEHybrid invalid method are different problems." }, { "left": "issue:39290", "right": "issue:45081", "accept": false, - "reason": "Gemma3 sliding_window config regression is unrelated to the Mistral tokenizer patch crash." + "reason": "Gemma3/vLLM sliding-window config error is unrelated to tokenizer regex patch crashes." }, { "left": "issue:43828", - "right": "issue:45685", + "right": "issue:45161", "accept": false, - "reason": "Autocast dtype mismatch and MPS histogram error are different bugs." + "reason": "Autocast dtype mismatch and GPT-OSS TP failure are different concrete bugs." }, { - "left": "issue:44560", - "right": "issue:45685", + "left": "issue:44918", + "right": "issue:45161", "accept": false, - "reason": "Video StopIteration and MPS histogram kernel failure are unrelated." + "reason": "TRL embedding unpacking and GPT-OSS tensor-parallel failure do not share the same bug." }, { "left": "issue:43039", "right": "issue:43828", "accept": false, - "reason": "Liger Kernel cross_entropy issue and autocast dtype mismatch are different." + "reason": "Both are runtime issues, but one is Liger Kernel cross-entropy selection and the other is an autocast dtype mismatch." }, { - "left": "issue:30277", - "right": "issue:36331", + "left": "issue:41762", + "right": "issue:43638", "accept": false, - "reason": "Collective mismatch vs compute_loss signature regression are unrelated." + "reason": "Gemma3 ZeRO-3 out-of-bounds and BERT ZeRO-3 out-of-bounds are similar in shape but not clearly the same bug." }, { - "left": "issue:43039", - "right": "issue:44805", + "left": "issue:43827", + "right": "issue:44512", "accept": false, - "reason": "Liger Kernel bug and mask indexing bug are unrelated." + "reason": "Both are docs regressions after v5 removals, but they target different documentation sections and are not the same change." }, { - "left": "issue:34928", - "right": "issue:36010", + "left": "issue:43716", + "right": "issue:44560", "accept": false, - "reason": "FSDP/checkpointing issue is unrelated to GenerationMixin import failure." - }, - { - "left": "issue:41762", - "right": "issue:43596", - "accept": true, - "reason": "Same DeepSpeed ZeRO-3 index-out-of-bounds failure during model init/load." + "reason": "Mistral image dtype mismatch and Qwen3-vl video StopIteration are unrelated multimodal failures." }, { - "left": "issue:43901", - "right": "issue:44512", + "left": "issue:43828", + "right": "issue:45685", "accept": false, - "reason": "Different docs regressions with different removed references." + "reason": "Autocast dtype mismatch is not the same as an unimplemented MPS histogram kernel." }, { - "left": "issue:39290", - "right": "issue:41720", + "left": "issue:44560", + "right": "issue:45685", "accept": false, - "reason": "Gemma3 vLLM config regression and Qwen3 auto device-mapping assert are unrelated." + "reason": "Video processing StopIteration and MPS Int histogram kernel error are unrelated." }, { "left": "issue:43866", "right": "issue:44863", "accept": false, - "reason": "Checkpoint corruption and NemotronH loading failure are different issues." + "reason": "Checkpoint corruption and NemotronH checkpoint loading are different issues." }, { "left": "issue:43854", "right": "issue:43866", "accept": false, - "reason": "GLM-4.7 unit-test load failure vs Ovis2 checkpoint corruption are unrelated." + "reason": "Unit-test model loading failure and corrupted checkpoint are different root causes." }, { - "left": "issue:43404", - "right": "issue:43866", + "left": "issue:39290", + "right": "issue:43531", "accept": false, - "reason": "Mistral3 lm_head tying bug is unrelated to Ovis2 checkpoint corruption." + "reason": "Gemma3 sliding-window attribute error and Qwen3-MoE sliding_window issue are related only by name, not by the same bug." }, { - "left": "issue:45464", - "right": "issue:45685", + "left": "issue:41093", + "right": "issue:43638", "accept": false, - "reason": "Qwen3.5 streaming chat failure and MPS histogram error are unrelated." + "reason": "Mask/index shape mismatch during one workflow is not the same as ZeRO-3 BERT empty-index loading failure." }, { - "left": "issue:43541", - "right": "issue:45685", + "left": "issue:43425", + "right": "issue:45070", "accept": false, - "reason": "Mixtral grouped_mm tracing error is unrelated to the MPS histogram issue." + "reason": "Torch version incompatibility and pydantic PretrainedConfig field breakage are unrelated." }, { - "left": "issue:43541", - "right": "issue:43828", + "left": "issue:43596", + "right": "issue:45800", "accept": false, - "reason": "Different runtime bugs: grouped_mm tracing vs autocast dtype mismatch." + "reason": "Both mention DeepSpeed-style incompatibilities in a broad sense, but the underlying bugs are different." }, { - "left": "issue:36010", - "right": "issue:39290", + "left": "issue:45362", + "right": "issue:45464", "accept": false, - "reason": "GenerationMixin import regression and Gemma3 sliding_window config bug are unrelated." + "reason": "Qwen3.5 chat crash and streaming inference failure are different runtime issues." }, { - "left": "issue:44841", - "right": "issue:45084", + "left": "issue:43792", + "right": "issue:45070", "accept": false, - "reason": "Voxtral processor failure and non-template node compile error are different bugs." - }, - { - "left": "issue:41762", - "right": "issue:43638", - "accept": true, - "reason": "Same ZeRO-3 empty-tensor/index-out-of-bounds load bug across different BERT/Gemma3 manifests." + "reason": "Whisper load failure and pydantic config breakage are unrelated." }, { - "left": "issue:43828", - "right": "issue:44292", + "left": "issue:43404", + "right": "issue:43866", "accept": false, - "reason": "Different model/runtime bugs with no shared concrete code path." + "reason": "Mistral3 lm_head tying and Ovis2 checkpoint corruption are different defects." }, { - "left": "issue:43425", - "right": "issue:45070", + "left": "issue:45464", + "right": "issue:45685", "accept": false, - "reason": "Torch-version incompatibility and pydantic PretrainedConfig regression are unrelated." + "reason": "Qwen3.5 streaming inference failure and MPS histogram kernel limitation are unrelated." }, { - "left": "issue:43596", - "right": "issue:45800", + "left": "issue:43541", + "right": "issue:45685", "accept": false, - "reason": "Shared versioning theme only; not the same bug." + "reason": "TorchDynamo grouped_mm tracing failure and MPS histogram kernel error are different problems." }, { - "left": "issue:45362", - "right": "issue:45464", + "left": "issue:43541", + "right": "issue:43828", "accept": false, - "reason": "Both touch Qwen3.5 chat, but the entry points and failures differ." + "reason": "Grouped_mm tracing failure and autocast dtype mismatch are not the same code-path bug." }, { - "left": "issue:43792", - "right": "issue:45070", + "left": "issue:36010", + "right": "issue:39290", "accept": false, - "reason": "Whisper load/run failure is unrelated to the PretrainedConfig pydantic regression." + "reason": "GenerationMixin import failure and Gemma3 config attribute error are unrelated." }, { - "left": "issue:41093", - "right": "issue:43638", + "left": "issue:44841", + "right": "issue:45084", "accept": false, - "reason": "Mask/tensor shape mismatch is unrelated to the ZeRO-3 index-out-of-bounds bug." + "reason": "Voxtral processor failure and non-template-node compile error do not share a concrete bug." }, { "left": "issue:41093", "right": "issue:41762", "accept": false, - "reason": "Different bugs: mask shape mismatch vs ZeRO-3 empty-index failure." + "reason": "Both are index/shape errors, but they involve different models and different failure points, so they should not be merged." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is conservative and well grounded in the packet. The ZeRO-3/BERT/Gemma3 index-out-of-bounds family is the only clearly supported duplicate cluster, and the accepted soft edges stay within that family. The rejected pairs are mostly justified by distinct titles/code paths." + "feedback": "The summary is grounded in the packet, and the only accepted duplicate pair is handled conservatively. The broader rejection rationale for the remaining pairs stays within what the titles/evidence support." }, "evaluator_used": true, "retried": false, @@ -671,20 +659,20 @@ "issue:39692|issue:43054", "issue:43828|issue:44928" ], - "prepared_review_unit_hash": "fcd972ca506233a6ac5b76fb25df60dfeae501e471a0a03a52817abd9ecf5cf8", + "prepared_review_unit_hash": "d3c933a61214168108948c3e21d557c1e869298c7e92e416aa804ae5edaceae1", "component_budget": { "node_count": 387, "item_count": 387, "soft_pair_count": 1455, - "serialized_chars": 749552, - "estimated_input_tokens": 187388, - "estimated_eval_tokens": 375032 + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 34, - "serialized_chars": 37556, + "serialized_chars": 37554, "estimated_input_tokens": 9389, "estimated_eval_tokens": 19034 }, @@ -693,222 +681,222 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The Transformers v5 non-persistent-buffer regression remains the clearest duplicate family, with 43950 as the best anchor. Per feedback, the SigLIP2 quality/output complaints are too broad to treat as the same bug, so that soft edge is rejected. Most other soft edges remain conservative rejections across unrelated model- or error-specific failures.", - "confidence": 0.87, - "canonical_issue_reason": "Issue 43950 is the clearest, most general write-up of the v5 non-persistent-buffer regression and has direct near-duplicates in 43644 and 44534.", + "summary": "Mostly a grab bag of Transformers/vLLM regressions, but a few true duplicate clusters stand out. The clearest one is the non-persistent-buffer regression around `from_pretrained()`/v5 (`43950`, `44534`, with `43644` feeding into the same bug). `43054`/`43994` also looks like the same SigLIP2 model-output problem. The `trust_remote_code`/custom-module pair `45632`/`45698` is related thematically, but not strong enough to treat as the same bug.", + "confidence": 0.7, + "canonical_issue_reason": "issue:43950 is the cleanest canonical issue: it describes the core v5 regression around non-persistent buffers clearly and has an obvious near-duplicate in issue:44534.", "canonical_pr_reason": null, - "best_issue_reason": "43950 is the best single issue to anchor the cluster because it states the concrete regression cleanly and represents the accepted duplicate family better than the shorter paraphrases.", + "best_issue_reason": "issue:43950 is the strongest representative because it captures the concrete regression most directly and is supported by the closest duplicate evidence.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43296", "right": "issue:43366", "accept": false, - "reason": "Different failures: PaddleOCR-VL load error vs GGUF gpt-oss support request." + "reason": "Different problems: PaddleOCR-VL/vLLM loading vs GGUF gpt-oss architecture support." }, { "left": "issue:43572", "right": "issue:45070", "accept": false, - "reason": "Unrelated config regressions: StableLm pad_token_idx vs PretrainedConfig pydantic field breakage." + "reason": "Both are config regressions, but missing pad_token_idx and pydantic PretrainedConfig breakage are different bugs." }, { "left": "issue:44387", "right": "issue:45005", "accept": false, - "reason": "Different bugs: int4 quantization OOM vs tied-weights behavior on translation models." + "reason": "Different symptoms and code paths: int4 quantization memory growth vs tied-weights behavior." }, { "left": "issue:43541", "right": "issue:45161", "accept": false, - "reason": "Different code paths: torch dynamo grouped_mm tracing error vs GPT-OSS MoE tensor-parallel issue." + "reason": "Both mention MoE/GPT-OSS-like models, but grouped_mm tracing failure is not the same as tensor-parallelism not working." }, { "left": "issue:43828", "right": "issue:45464", "accept": false, - "reason": "Autocast dtype mismatch on Phi-tiny-MoE is not the same as Qwen3.5 streaming chat/completions failure." + "reason": "Unrelated failures: autocast dtype mismatch in Phi-tiny-MoE vs streaming chat/completions failure in Qwen3.5-0.8B." }, { "left": "issue:44291", "right": "issue:45070", "accept": false, - "reason": "Different regressions: unexpected _is_hf_initialized arg vs pydantic PretrainedConfig field handling." + "reason": "Different regressions: init_empty_weights/_is_hf_initialized issue vs pydantic PretrainedConfig field break." }, { "left": "issue:45020", "right": "issue:45800", "accept": false, - "reason": "Broad remote_code breakage is not the same as a torch 2.4.1 / transformers 5.8.0 incompatibility." + "reason": "Too broad and different: remote_code breakage vs torch 2.4.1/transformers 5.8.0 incompatibility." }, { "left": "issue:43366", "right": "issue:45464", "accept": false, - "reason": "gpt-oss architecture support and Qwen3.5 streaming failures are unrelated." + "reason": "Different scopes: GGUF gpt-oss architecture support vs Qwen3.5 streaming API failure." }, { "left": "issue:43531", "right": "issue:45070", "accept": false, - "reason": "Qwen3-MoE sliding_window handling is unrelated to the PretrainedConfig/pydantic regression." + "reason": "Qwen3-MoE sliding_window issue is unrelated to the PretrainedConfig/pydantic regression." }, { "left": "issue:39290", "right": "issue:43742", "accept": false, - "reason": "Gemma3TextConfig missing sliding_window_pattern and MobileLLM key error are different model-loading bugs." + "reason": "Both are model-loading errors, but Gemma3 sliding-window config and MobileLLM key error are distinct bugs." }, { "left": "issue:43541", "right": "issue:44589", "accept": false, - "reason": "Grouped_mm tracing failure is unrelated to missing Float8 storage support." + "reason": "Different concrete failures: torch dynamo grouped_mm tracing vs missing Float8 storage object." }, { "left": "issue:41093", "right": "issue:43596", "accept": false, - "reason": "Mask length mismatch and zero3 init index error are different tensor-shape failures." + "reason": "Different indexing errors in different contexts; no shared code-path evidence." }, { "left": "issue:43054", "right": "issue:43994", - "accept": false, - "reason": "Both mention SigLIP2, but the titles only describe broad quality/output complaints and do not establish the same concrete bug." + "accept": true, + "reason": "Both report SigLIP2 producing bad/nonsensical outputs, likely the same model/config bug." }, { "left": "issue:45542", "right": "issue:45800", "accept": false, - "reason": "TensorBoard-only tf backend error is unrelated to the torch/transformers version incompatibility." + "reason": "TensorBoard/TensorFlow backend detection bug is unrelated to torch/transformers version incompatibility." }, { "left": "issue:43716", "right": "issue:44918", "accept": false, - "reason": "Mistral-3 image-preprocessor dtype mismatch is not the same as Qwen3.5 SFT trainer unpacking failure." + "reason": "Different model families and failures: Mistral-3 image preprocessor dtype mismatch vs Qwen3.5/TRL embedding unpacking." }, { "left": "issue:44918", "right": "issue:45685", "accept": false, - "reason": "Different issues: TRL unpacking failure vs an MPS histogram implementation gap." + "reason": "Qwen3.5/TRL embedding issue and MPS histogram error are unrelated." }, { "left": "issue:45632", "right": "issue:45698", "accept": false, - "reason": "Both involve custom/remote-code loading, but one is a cache-path collision and the other is loading the wrong custom module after save_pretrained." + "reason": "Both are about custom/remote-code loading, but one is a cache-path collision for local models sharing a leaf directory and the other is a wrong custom module being loaded after save_pretrained; related area, not the same concrete bug." }, { "left": "issue:43404", "right": "issue:45005", "accept": false, - "reason": "Mistral3 tied lm_head weights bug is different from the translation-model tied-weights regression." + "reason": "Both involve tied weights, but they target different models and likely different regressions." }, { "left": "issue:43645", "right": "issue:45399", "accept": false, - "reason": "Jupyter custom-model initialization bug is unrelated to flash-attn2 fallback blocking." + "reason": "Jupyter/custom-model initialization bug is unrelated to flash-attn2 fallback gating." }, { "left": "issue:40990", "right": "issue:44960", "accept": false, - "reason": "GPT-OSS perplexity regression and GLM5 are different model-specific issues." + "reason": "Perplexity anomaly on gpt-oss-20b is not the same as the generic GLM5 issue." }, { "left": "issue:44530", "right": "issue:45005", "accept": false, - "reason": "PagedAttentionCache linear_attention crash is unrelated to tied-weights handling." + "reason": "PagedAttentionCache linear_attention failure is unrelated to tied-weight issues." }, { "left": "issue:42915", "right": "issue:45399", "accept": false, - "reason": "Qwen3-MoE FP8 failure and flash-attn2 fallback logic are separate problems." + "reason": "Qwen3-MoE FineGrainedFP8Config failure is unrelated to flash-attn2 fallback checks." }, { "left": "issue:44360", "right": "issue:44512", "accept": false, - "reason": "DSA indexer ReLU discussion is unrelated to docs mentioning the removed run command." + "reason": "DSA indexer/ReLU discussion is unrelated to the removed run-command docs issue." }, { "left": "issue:44863", "right": "issue:45005", "accept": false, - "reason": "NemotronH checkpoint loading failure is not the same as translation-model tied weights." + "reason": "NemotronH checkpoint loading is unrelated to tied weights on translation models." }, { "left": "issue:43873", "right": "issue:45005", "accept": false, - "reason": "Quantized offloading behavior is unrelated to tied-weights regression." + "reason": "Quantization offloading behavior is different from tied-weight regressions." }, { "left": "issue:39186", "right": "issue:44918", "accept": false, - "reason": "FSDP 2-D weight error and Qwen3.5 embedding unpacking are different failures." + "reason": "FSDP weight-dimension error is unrelated to Qwen3.5/TRL embedding unpacking." }, { "left": "issue:44492", "right": "issue:44512", "accept": false, - "reason": "A typo in cache strategies is unrelated to the removed run-command docs issue." + "reason": "A cache-strategy typo is unrelated to docs mentioning a removed command." }, { "left": "issue:44918", "right": "issue:45507", "accept": false, - "reason": "Qwen3.5 embedding unpacking and GraniteMoEHybrid invalid method calls are different bugs." + "reason": "Different model families and failure modes; no common code-path evidence." }, { "left": "issue:43975", "right": "issue:44292", "accept": false, - "reason": "Detokenization error and Qwen-3-8B-NVFP4 runtime failure are unrelated." + "reason": "Incorrect detokenization and NVFP4 loading errors are distinct issues." }, { "left": "issue:43950", "right": "issue:44534", "accept": true, - "reason": "Same underlying regression: v5 corrupts or fills non-persistent buffers with junk on load." + "reason": "Near-exact duplicate regression: both say v5 fills/corrupts non-persistent buffers with junk." }, { "left": "issue:43644", "right": "issue:43950", "accept": true, - "reason": "Same underlying regression: non-persistent buffers are corrupted in Transformers 5.0.0 / v5 from_pretrained." + "reason": "Same underlying regression in from_pretrained corrupting non-persistent buffers." }, { "left": "issue:44292", "right": "issue:45507", "accept": false, - "reason": "Qwen-3-8B-NVFP4 runtime error is unrelated to GraniteMoEHybrid invalid method calls." + "reason": "Qwen-3-8B-NVFP4 loading failure is unrelated to GraniteMoEHybrid invalid method calls." }, { "left": "issue:39692", "right": "issue:43054", "accept": false, - "reason": "SigLIP2 docs-example failures are not the same as the later model-quality complaint." + "reason": "SigLIP2 docs-example errors are different from the broader SigLIP2 embedding-quality complaint." }, { "left": "issue:43828", "right": "issue:44928", "accept": false, - "reason": "Autocast dtype mismatch and RLHF NaN from 3D position_ids/SDPA fallback are different bugs." + "reason": "Different dtype/NaN problems in different models and training setups; not the same bug." } ] }, "evaluator_result": { "accept": false, - "feedback": "The main non-persistent-buffer family is well grounded, but one soft-edge accept is too speculative: issue:43054 and issue:43994 both mention SigLIP2, yet the titles only show broad quality/output complaints and do not clearly establish the same bug. Treating them as a duplicate family overstates the evidence. The other soft-edge rejections look conservative." + "feedback": "The overall summary is mostly grounded, but one accepted soft-edge verdict is too strong for the evidence in the packet. `issue:45632` and `issue:45698` are both about remote-code/custom-module loading, but the titles alone do not establish that they are the same bug or clearly mergeable into one fix. The other accepted duplicates (`43950`/`44534`, `43644`/`43950`, and likely `43054`/`43994`) are fine, but this custom-module pair makes the cluster too speculative as written." }, "evaluator_used": true, "retried": true, @@ -1017,262 +1005,238 @@ "issue:43940|issue:45313", "issue:43872|issue:44291" ], - "prepared_review_unit_hash": "0c767cb501a4f26326ce41ae80ee45216b92829601dc4ea00c60f06f84d3fdcb", + "prepared_review_unit_hash": "83e778eba049c578dbea81bf7bea6b6393572d3cb393e3e36f711f7ec1128f25", "component_budget": { "node_count": 387, "item_count": 387, "soft_pair_count": 1455, - "serialized_chars": 749552, - "estimated_input_tokens": 187388, - "estimated_eval_tokens": 375032 + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 37, - "serialized_chars": 38721, - "estimated_input_tokens": 9681, - "estimated_eval_tokens": 19618 + "serialized_chars": 38719, + "estimated_input_tokens": 9680, + "estimated_eval_tokens": 19616 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a grab-bag of Transformers v5/model-loading regressions across different models and subsystems. Only a couple of soft duplicates look truly same-bug; most pairs are just similar failure patterns.", - "confidence": 0.79, - "canonical_issue_reason": "issue:45020 is the broadest and most central umbrella here: it explicitly covers recent Transformers versions breaking remote_code models, which matches the dominant loading/regression theme across the cluster.", + "summary": "This cluster is mostly a mix of unrelated model-loading/runtime regressions and docs issues, not one duplicate set. I only see two clear duplicate pairs: the v5 pipeline docs cleanup, and the unexpected `_is_hf_initialized` TypeError during loading/init_empty_weights.", + "confidence": 0.82, + "canonical_issue_reason": "issue:44291 is the clearest concrete bug report in the cluster and the exact duplicate target for issue:43872; it precisely pins the shared `_is_hf_initialized` TypeError during model loading.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45020 is the best single issue anchor because it is broad, active, and closest to the shared theme of version-related model-loading breakages.", + "best_issue_reason": "issue:45020 is the strongest overall umbrella for the cluster because it captures the broader family of recent remote_code / model-loading regressions that many of the other reports resemble, even though most are not true duplicates.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43866", "right": "issue:44403", "accept": false, - "reason": "Different problems: checkpoint corruption versus generic loading noise." + "reason": "Different problems: corrupted checkpoint vs generic loading noise; no shared code-path evidence." }, { "left": "issue:36331", "right": "issue:41762", "accept": false, - "reason": "Training API TypeError and ZeRO-3 model-loading IndexError are unrelated." + "reason": "Trainer API signature error is unrelated to a ZeRO-3 model-loading index failure." }, { "left": "issue:41628", "right": "issue:45800", "accept": false, - "reason": "ImportError for AutoImageProcessor is not the same as torch/transformers version incompatibility." + "reason": "Import/export API issue is not the same as a torch/transformers version incompatibility report." }, { "left": "issue:41553", "right": "issue:43883", "accept": false, - "reason": "Bad tokenizer error messaging for Voxtral and missing all_tied_weights_keys in Molmo are different bugs." + "reason": "Bad tokenizer error message and a missing model attribute are different concrete bugs." }, { "left": "issue:43872", "right": "issue:44589", "accept": false, - "reason": "Both are TypeErrors during loading, but the underlying failure modes are different (_is_hf_initialized vs Float8 storage lookup)." + "reason": "Unexpected `_is_hf_initialized` argument is unrelated to the float8 storage lookup failure." }, { "left": "issue:38175", "right": "issue:43054", "accept": false, - "reason": "Both concern SigLIP/SigLIP2 quality, but one is zero probabilities and the other is poorer text embeddings; not the same bug." + "reason": "Both involve SigLIP2, but one is zero probabilities and the other is embedding quality; too little evidence of the same bug." }, { "left": "issue:41553", "right": "issue:43881", "accept": false, - "reason": "Tokenizer-loading error for Voxtral and GLM-4V load failure are distinct model issues." + "reason": "Voxtral AutoTokenizer loading error and glm-4v-9b loading failure are model-specific and distinct." }, { "left": "issue:43828", "right": "issue:45800", "accept": false, - "reason": "Autocast dtype mismatch in Phi-tiny-MoE is unrelated to torch/transformers version compatibility." + "reason": "Autocast dtype mismatch is not the same as a torch/transformers version compatibility issue." }, { "left": "issue:43054", "right": "issue:45701", "accept": false, - "reason": "SigLIP2 embedding quality and generic tokenization changes are not the same underlying issue." + "reason": "Embedding-quality complaint and tokenization-change report are different failure modes." }, { "left": "issue:42947", "right": "issue:44387", "accept": false, - "reason": "Gradient checkpointing with PEFT LoRA is a different problem from increased CUDA reserved memory under int4 quantization." + "reason": "LoRA gradient-checkpointing behavior is unrelated to int4 CUDA memory growth." }, { "left": "issue:44403", "right": "issue:44451", "accept": false, - "reason": "Generic loading noise and failure to load a specific model are not the same bug." + "reason": "Generic loading noise and a specific ScandiBERT load failure are not the same bug." }, { "left": "issue:43881", "right": "issue:43883", "accept": false, - "reason": "Both are load failures, but one is a model-specific load problem and the other is a missing tied-weights attribute." + "reason": "Different models and different failure symptoms; no concrete overlap." }, { "left": "issue:43643", "right": "issue:45020", "accept": false, - "reason": "Missing fields from AutoConfig with trust_remote_code and broader remote_code breakage overlap only at a high level." + "reason": "Both mention remote_code, but one is missing fields in AutoConfig and the other is a broad breakage report; not the same bug." }, { "left": "issue:43054", "right": "issue:43493", "accept": false, - "reason": "Both mention SigLIP2, but one is embedding degradation and the other is HF-vs-JAX implementation discrepancy." + "reason": "Strongly related SigLIP2 reports, but one is a quality regression and the other is an HF-vs-JAX discrepancy; not proven identical." }, { "left": "issue:43975", "right": "issue:45520", "accept": false, - "reason": "Detokenization regression and Python 3.13 flash_attn import error are unrelated." + "reason": "Detokenization regression is unrelated to a flash_attn import_utils KeyError." }, { "left": "issue:43975", "right": "issue:45237", "accept": false, - "reason": "Detokenization regression and AMD GPU support failure are different problems." + "reason": "Tokenizer detokenization bug and AMD GPU runtime failure are different issues." }, { "left": "issue:44403", "right": "issue:44488", "accept": false, - "reason": "Both involve load-time complaints, but one is generic noise and the other is a specific model load failure." + "reason": "Both are model-load complaints, but they point to different models and likely different root causes." }, { "left": "issue:42947", "right": "issue:43856", "accept": false, - "reason": "PEFT LoRA checkpointing and Qwen3 MoE memory usage are different performance/behavior issues." + "reason": "Gradient checkpointing/LoRA behavior is not the same as Qwen3 MoE training memory usage." }, { "left": "issue:42491", "right": "issue:44368", "accept": false, - "reason": "LoRA incompatibility across hf4.x/hf5.x and a tie_word_embeddings warning are not the same bug." + "reason": "Cross-version LoRA compatibility issue and a tie_word_embeddings warning are not the same underlying bug." }, { "left": "issue:41762", "right": "issue:42915", "accept": false, - "reason": "Gemma3 ZeRO-3 load failure and Qwen3 MoE FP8 failure are different model-specific issues." + "reason": "Both are Qwen-related loading failures, but the specific failure modes and model variants differ too much." }, { "left": "issue:43975", "right": "issue:44451", "accept": false, - "reason": "Different symptoms: incorrect detokenization versus a model load failure." + "reason": "Different model/tokenization regressions with no shared concrete code-path." }, { "left": "issue:44661", "right": "issue:45020", "accept": false, - "reason": "Tokenizer mapping restriction in add-new-model-like is not the same as remote_code load breakage." + "reason": "Tokenizer-mapping failure in add-new-model-like is not the same as broad remote_code loading regressions." }, { "left": "issue:43824", "right": "issue:43872", "accept": false, - "reason": "Importing Qwen2.5-VL and bitsandbytes _is_hf_initialized TypeError are unrelated." + "reason": "ImportError for Qwen2.5-VL and `_is_hf_initialized` TypeError are unrelated." }, { "left": "issue:44991", "right": "issue:45399", "accept": false, - "reason": "Tokenizer load failure for a specific model and flash-attn fallback blocking are not the same underlying issue." + "reason": "Tokenizer loading regression and flash-attn fallback blocking are different issues." }, { "left": "issue:45020", "right": "issue:45356", "accept": false, - "reason": "Broad remote_code breakage and a Kimi-K2.5 tokenizer regression are different issues." + "reason": "Broad remote_code breakage and a Kimi tokenizer codec/warning regression are not the same bug." }, { "left": "issue:43782", "right": "issue:44960", "accept": false, - "reason": "Qwen3VL weight_only error and a generic GLM5 issue do not match." + "reason": "Different model families and different loading errors." }, { "left": "issue:45081", "right": "issue:45399", "accept": false, - "reason": "Mistral regex patch crash and flash-attn fallback checks are separate code paths." + "reason": "Mistral regex patch crash is unrelated to flash-attn fallback logic." }, { "left": "issue:43827", "right": "issue:44509", "accept": true, - "reason": "Same doc regression: both say v5 removed pipeline tasks are still referenced in docs." + "reason": "Same docs regression: both say v5 removed pipeline tasks are still referenced in docs, just with different wording." }, { "left": "issue:43792", "right": "issue:44220", "accept": false, - "reason": "Whisper not running and a generic fbank feature extraction issue are too underspecified to treat as the same bug." + "reason": "Whisper runtime failure is too broad to equate with the specific `_torch_extract_fbank_features()` issue." }, { "left": "issue:43404", "right": "issue:43950", "accept": false, - "reason": "Untied lm_head in Mistral3 and silent corruption of non-persistent buffers are different model-state bugs." + "reason": "Weight-tying bug in Mistral3 and non-persistent buffer corruption are distinct model/state-loading problems." }, { "left": "issue:43742", "right": "issue:45399", "accept": false, - "reason": "MobileLLM loading failure and flash-attn fallback gating are unrelated." - }, - { - "left": "issue:42915", - "right": "issue:45020", - "accept": false, - "reason": "Qwen3 MoE FP8 failure and broad remote_code breakage are not the same concrete bug." - }, - { - "left": "issue:43257", - "right": "issue:43866", - "accept": false, - "reason": "Qwen3 MoE weight conversion under accelerate+deepspeed and Ovis2 checkpoint corruption are unrelated." - }, - { - "left": "issue:41553", - "right": "issue:44991", - "accept": false, - "reason": "Voxtral tokenizer error messaging and est-roberta tokenizer loading failure are different model-specific issues." - }, - { - "left": "issue:43742", - "right": "issue:45020", - "accept": false, - "reason": "A specific MobileLLM key error is not the same as the broad remote_code regression umbrella." + "reason": "MobileLLM-125M key error and flash-attn fallback blocking are not the same failure." }, { "left": "issue:43940", "right": "issue:45313", "accept": false, - "reason": "Both are ZeRO-3 Qwen-family load failures, but they target different model variants and likely different code paths." + "reason": "Both are ZeRO-3 loading failures, but they affect different Qwen generations and the concrete symptoms differ." }, { "left": "issue:43872", "right": "issue:44291", "accept": true, - "reason": "Same underlying _is_hf_initialized TypeError during model loading; these look like the same constructor-signature regression." + "reason": "These describe the same `_is_hf_initialized`/init_empty_weights TypeError during loading; 44291 is the clearer full report." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is grounded in the packet: issue:45020 is a reasonable umbrella for the remote_code/version-regression theme, and the cluster-wide characterization as mostly heterogeneous loading/regression bugs fits the titles. The two accepted soft duplicates also look mergeable into single fixes (docs regression; the _is_hf_initialized TypeError). I don\u2019t see an overstatement of evidence in the pair verdicts." + "feedback": "Grounded and conservative overall. The two accepted duplicate pairs are supported by the packet, and the other soft-edge rejections avoid overclaiming despite surface similarity. The cluster summary does not overstate duplication." }, "evaluator_used": true, "retried": false, @@ -1376,20 +1340,20 @@ "issue:38175|issue:39692", "issue:45313|issue:45357" ], - "prepared_review_unit_hash": "f3cc03c82c93d8f091c5243d5d9294604adde4b719d63da9383c3b7d9867cfcb", + "prepared_review_unit_hash": "cc27aef110d8c6c96dfb6199d1efa7db385bd17e45f4ad2ba4f1dc6a621c081f", "component_budget": { "node_count": 387, "item_count": 387, "soft_pair_count": 1455, - "serialized_chars": 749552, - "estimated_input_tokens": 187388, - "estimated_eval_tokens": 375032 + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 32, - "serialized_chars": 37466, + "serialized_chars": 37467, "estimated_input_tokens": 9367, "estimated_eval_tokens": 18990 }, @@ -1398,213 +1362,213 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Mostly separate regression reports. Issue:43957 is the best umbrella issue because its title explicitly describes a model-loading regression on `torch.device(\"meta\")` in transformers 5.+, with concrete examples in the title; that makes it the broadest visible loading-path issue in the cluster.", - "confidence": 0.79, - "canonical_issue_reason": "issue:43957 is the clearest umbrella issue from the visible titles: it names a transformers 5.+ model-loading regression on `torch.device(\"meta\")` and includes concrete examples in the title. That makes it broader and more representative than the narrower loading failures elsewhere.", + "summary": "This set is mostly a grab-bag of unrelated bug reports. Only three soft pairs look like true duplicates: the custom-model-init regression in Transformers 5.0.0, the SigLIP2 bad-output reports, and the mask/tensor shape mismatch error.", + "confidence": 0.83, + "canonical_issue_reason": "No single issue represents the whole list, but 43646 is the best canonical anchor among the confirmed duplicates: it states the broad Transformers 5.0 custom model initialization regression, while 43645 is just a notebook-specific repro.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43957 is the most useful representative issue because its title captures the shared loading regression pattern in a broad but still concrete way, while staying anchored to an explicit code path (`torch.device(\"meta\")`) and example models.", + "best_issue_reason": "43646 is the strongest overall issue to keep as the representative issue: it is concise, root-cause oriented, and broad enough to subsume the narrower 43645 report.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43854", "right": "issue:44863", "accept": false, - "reason": "Both are load failures, but the titles point to different models and different failure descriptions." + "reason": "Different model families and failure modes; both are loading bugs, but not the same underlying defect." }, { "left": "issue:42915", "right": "issue:43872", "accept": false, - "reason": "Qwen3Moe FP8 failure and bitsandbytes constructor incompatibility are different bugs." + "reason": "Different quantization-related regressions with different error surfaces and code paths." }, { "left": "issue:44928", "right": "issue:45588", "accept": false, - "reason": "Different models, different code paths, and different crash modes." + "reason": "Both touch attention-related code, but the symptoms and underlying causes are different." }, { "left": "issue:41553", "right": "issue:44488", "accept": false, - "reason": "One is a bad error message for Voxtral loading; the other is a load failure for a different model." + "reason": "Both are tokenizer-loading complaints, but they concern different models and likely different root causes." }, { "left": "issue:43257", "right": "issue:45305", "accept": false, - "reason": "Both mention DeepSpeed, but the titles describe different problems: MoE weight conversion vs gradient averaging." + "reason": "Distinct DeepSpeed issues: one is MOE weight conversion, the other is gradient averaging with GAS." }, { "left": "issue:36331", "right": "issue:41093", "accept": false, - "reason": "Unexpected trainer argument and mask/index shape mismatch are unrelated failures." + "reason": "Unrelated trainer API breakage versus tensor mask shape mismatch." }, { "left": "issue:43872", "right": "issue:45538", "accept": false, - "reason": "bitsandbytes incompatibility and CLIPTokenizer max-length behavior are separate issues." + "reason": "Different bug classes entirely: bitsandbytes parameter incompatibility vs tokenizer max-length handling." }, { "left": "issue:44530", "right": "issue:44863", "accept": false, - "reason": "PagedAttentionCache group-type crash and NemotronH checkpoint loading are different bugs." + "reason": "Different models and different failures; no shared concrete bug." }, { "left": "issue:41093", "right": "issue:44805", "accept": true, - "reason": "The titles describe the same mask-vs-indexed-tensor shape mismatch pattern." + "reason": "Same mask/index shape-mismatch error; 44805 is essentially the same bug with generalized placeholders." }, { "left": "issue:44479", "right": "issue:44560", "accept": false, - "reason": "Both are video-related, but the titles indicate different regressions and different failure points." + "reason": "Both are video-related, but the reported failures and model scopes differ enough to avoid merging." }, { "left": "issue:24643", "right": "issue:30064", "accept": false, - "reason": "DeepSpeed training weight-shape error and void segmentation-map processing are unrelated." + "reason": "Completely different subsystems and symptoms." }, { "left": "issue:43366", "right": "issue:43828", "accept": false, - "reason": "GGUF gpt-oss support and torch.autocast dtype mismatch are different changes/failures." + "reason": "GGUF architecture support and autocast dtype mismatch are unrelated problems." }, { "left": "issue:43366", "right": "issue:43531", "accept": false, - "reason": "GGUF support and a Qwen3-MoE sliding_window issue are not the same underlying bug." + "reason": "Different Qwen3-MoE and GGUF-related issues; same general area, not the same bug." }, { "left": "issue:43550", "right": "issue:44910", "accept": false, - "reason": "Different model families and different runtime failures." + "reason": "Both involve attention kernels, but the concrete model, trigger, and failure mode differ." }, { "left": "issue:43645", "right": "issue:43646", "accept": true, - "reason": "Both titles say Transformers 5.0.0 breaks custom model initialization; one adds the Jupyter context." + "reason": "Same Transformers 5.0 custom-model-initialization regression; 43645 is a notebook-specific reproduction of 43646." }, { "left": "issue:43366", "right": "issue:44589", "accept": false, - "reason": "GGUF architecture support and missing Float8 storage are separate compatibility issues." + "reason": "Unrelated GGUF/model-architecture support issue vs missing Float8 storage type." }, { "left": "issue:43950", "right": "issue:45325", "accept": false, - "reason": "Non-persistent buffer corruption and rope-index scaling are different regressions." + "reason": "Different regressions: non-persistent buffer corruption versus Qwen2.5-VL rope indexing." }, { "left": "issue:44560", "right": "issue:44918", "accept": false, - "reason": "The titles describe different Qwen3.5/video/trainer failures, not one shared bug." + "reason": "Different Qwen3.5 problems with different reproduction paths and symptoms." }, { "left": "issue:42617", "right": "issue:43366", "accept": false, - "reason": "3d_parallel.py runtime failure and GGUF gpt-oss support are unrelated." + "reason": "Unrelated runner failure versus GGUF architecture support." }, { "left": "issue:42915", "right": "issue:43957", "accept": false, - "reason": "The titles point to different loading problems for different model families and different mechanisms." + "reason": "Different loading regressions: FineGrainedFP8Config versus meta-device initialization." }, { "left": "issue:43824", "right": "issue:43957", "accept": false, - "reason": "Missing import of a Qwen2.5-VL class is a separate export/import issue from meta-device model loading." + "reason": "Import error for a missing class is not the same as the meta-device loading regression." }, { "left": "issue:36010", "right": "issue:42915", "accept": false, - "reason": "GenerationMixin import error and Qwen3Moe FP8 failure are unrelated." + "reason": "Completely different import-path regression versus quantization config failure." }, { "left": "issue:38175", "right": "issue:43994", "accept": true, - "reason": "Both titles center on siglip2-base-patch16-224 producing wrong outputs through different entry points." + "reason": "Same SigLIP2 model producing invalid outputs; these are two reports of the same underlying inference problem." }, { "left": "issue:43646", "right": "issue:43950", "accept": false, - "reason": "Custom model initialization and non-persistent buffer corruption do not match as the same bug." + "reason": "Custom model initialization regression and non-persistent buffer corruption are separate issues." }, { "left": "issue:44155", "right": "issue:45584", "accept": false, - "reason": "AudioFlamingo3 batched inference leakage and Whisper empty-transcription generation are different issues." + "reason": "Different models, different workflows, and different failure mechanisms." }, { "left": "issue:44877", "right": "issue:45030", "accept": false, - "reason": "Strict config handling for granite_speech and glm4v config validation changes are different loader regressions." + "reason": "Both are config-validation related, but they affect different models and are not the same bug." }, { "left": "issue:44560", "right": "issue:44805", "accept": false, - "reason": "Video StopIteration and mask/tensor shape mismatch are not the same failure mode." + "reason": "Unrelated video StopIteration bug versus mask/tensor shape mismatch." }, { "left": "issue:43638", "right": "issue:44661", "accept": false, - "reason": "ZeRO-3/BertModel index error and tokenizer-mapping add-new-model-like failure are unrelated." + "reason": "Different failure modes: DeepSpeed/BERT zero3 init versus tokenizer-mapping-name handling." }, { "left": "issue:43596", "right": "issue:44661", "accept": false, - "reason": "Same reasoning: different code paths and different errors." + "reason": "Same as above; not the same underlying defect." }, { "left": "issue:43317", "right": "issue:44756", "accept": false, - "reason": "gpu+cpu offload load failure and Strix Halo mmap/OOM are different runtime-environment bugs." + "reason": "Different memory/loading problems with different triggers and remedies." }, { "left": "issue:38175", "right": "issue:39692", "accept": false, - "reason": "They share the SigLIP2 model family, but the titles describe different scenarios: wrong outputs vs docs-example errors with model/processor mismatch and quantization failure." + "reason": "Same model family, but this is a docs/example and quantization failure, not the same bug as the bad-output report." }, { "left": "issue:45313", "right": "issue:45357", "accept": false, - "reason": "ZeRO-3 weight loading for `language_model` and incorrect visual-encoder keys in `save_pretrained` are different parts of the workflow." + "reason": "Different Qwen3.5 regressions: ZeRO-3 weight loading versus save_pretrained key serialization." } ] }, "evaluator_result": { - "accept": false, - "feedback": "The clustering judgment is mostly conservative, but the issue rationale overreaches the packet in places. In particular, it claims issue:43957 \"gives multiple affected models,\" which is not supported by the packet metadata shown here. Please keep the canonical/best-issue explanation grounded in the visible title-level evidence only, and avoid adding specifics that are not present in the packet." + "accept": true, + "feedback": "Grounded overall. The three accepted soft pairs are defensible from the packet, and the many rejections stay conservative by avoiding broad same-bug claims across different models/subsystems. The only slightly borderline call is the SigLIP2 pair, but it is still supportable from the titles alone, so the cluster summary remains precise." }, "evaluator_used": true, - "retried": true, + "retried": false, "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null @@ -1691,14 +1655,14 @@ "issue:45362|issue:45406", "issue:42491|issue:45292", "issue:44488|issue:44991", - "issue:43404|issue:43454", "issue:36010|issue:39404", + "issue:43404|issue:43454", "issue:45276|issue:45335", "issue:44164|issue:44291", "issue:44402|issue:44488", "issue:39692|issue:44960", - "issue:43335|issue:45663", "issue:43425|issue:43957", + "issue:43335|issue:45663", "issue:42947|issue:45704", "issue:45663|issue:45759", "issue:43873|issue:44387", @@ -1706,244 +1670,244 @@ "issue:44387|issue:44756", "issue:44291|issue:44756" ], - "prepared_review_unit_hash": "d7d1530eee5bbf4ab5aa1c76f29b5fb8d87c4c7c52e93e140dae17d54c862ba0", + "prepared_review_unit_hash": "71bd6f313bb62a9786dddfbb9ef4945e54d03427c6c946d0476213906830759f", "component_budget": { "node_count": 387, "item_count": 387, "soft_pair_count": 1455, - "serialized_chars": 749552, - "estimated_input_tokens": 187388, - "estimated_eval_tokens": 375032 + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 }, "budget": { "node_count": 47, "item_count": 47, "soft_pair_count": 34, - "serialized_chars": 37318, - "estimated_input_tokens": 9330, - "estimated_eval_tokens": 18916 + "serialized_chars": 37322, + "estimated_input_tokens": 9331, + "estimated_eval_tokens": 18918 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Mostly a mixed cluster of unrelated Transformers bug reports. I only accept the clear near-duplicate regression pair around non-persistent buffers, plus the closely related resize_token_embeddings embedding-propagation pair; the BERT model-loading pair 44451/44488 is not strong enough evidence to merge.", - "confidence": 0.46, - "canonical_issue_reason": "No single issue cleanly represents the whole cluster. If one anchor is needed, 43644 is the best representative because it is the clearest concrete regression in the set and has the strongest near-duplicate match in 44534.", + "summary": "The cluster is mostly a set of unrelated Transformers issues. The strongest duplicate signal is issue:43644 \u2194 issue:44534, which appear to report the same Transformers v5 non-persistent-buffer regression; this should be treated as a likely/strong soft duplicate rather than a confirmed duplicate. The rest of the soft pairs are not safe to merge as duplicates.", + "confidence": 0.87, + "canonical_issue_reason": "issue:43644 is the best anchor because it describes a concrete Transformers 5.0.0 regression with clear symptoms, and issue:44534 is a strong soft duplicate candidate for the same non-persistent-buffer bug.", "canonical_pr_reason": null, - "best_issue_reason": "43644 is the best overall issue anchor: it is specific, reproducible, and has the most defensible duplicate-style counterpart, so it best reflects the cluster\u2019s strongest shared regression theme.", + "best_issue_reason": "issue:43644 is the most representative and actionable issue in the cluster: it is specific, reproducible, and central to the only clear duplicate pair.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43644", "right": "issue:44534", "accept": true, - "reason": "Same underlying Transformers v5 regression: non-persistent buffers being filled with junk; the reports are effectively paraphrases." + "reason": "Strong soft duplicate candidate: both titles describe the same Transformers v5 bug where non-persistent buffers are filled with junk." }, { "left": "issue:44960", "right": "issue:44977", "accept": false, - "reason": "Both are model-specific generation/flash-attention complaints, but they describe different failure modes and are not the same bug." + "reason": "Both are Qwen-related generation/flash-attention failures, but they describe different models and different failure modes." }, { "left": "issue:44451", "right": "issue:44488", "accept": false, - "reason": "Only evidence is that both are BERT model-loading failures for different models; that is too weak to conclude a shared regression." + "reason": "Both are model-loading problems, but they involve different checkpoints and no shared concrete bug is evident." }, { "left": "issue:42491", "right": "issue:43720", "accept": false, - "reason": "Different concrete bugs: Qwen3 MoE LoRA compatibility vs packed-weight loading under accelerate." + "reason": "Different subsystems and symptoms: Qwen3 MoE LoRA incompatibility vs BitNet packed-weight unpacking during accelerate loading." }, { "left": "issue:43708", "right": "issue:44075", "accept": false, - "reason": "Unrelated code paths: resume/max_steps calculation vs SGD arguments not being applied." + "reason": "These are unrelated training/resume max_steps logic and SGD-argument usage bugs." }, { "left": "issue:43854", "right": "issue:43931", "accept": false, - "reason": "Both are loading failures, but for different models and different symptoms; not the same underlying defect." + "reason": "Different model-loading failures on different models; no clear shared code path." }, { "left": "issue:41720", "right": "issue:42915", "accept": false, - "reason": "Same broad model family, but one is a CUDA assert with auto device mapping and the other is a FineGrainedFP8Config failure." + "reason": "Both mention Qwen3-family models, but one is device-mapping/cuda assert and the other is FineGrainedFP8Config; not the same bug." }, { "left": "issue:42491", "right": "issue:45276", "accept": false, - "reason": "Different problems: Qwen3 MoE LoRA loading vs resize_token_embeddings not updating Gemma4 embeddings." + "reason": "LoRA/model-compatibility issue vs resize_token_embeddings propagation bug; different code paths." }, { "left": "issue:43856", "right": "issue:43873", "accept": false, - "reason": "Both involve memory, but one is training memory usage and the other is offloading with quantization; too different to merge." + "reason": "Both are memory-related, but one is Qwen3 MoE training inefficiency and the other is offloading/quantization behavior." }, { "left": "issue:43257", "right": "issue:43854", "accept": false, - "reason": "Different loading paths and models: Qwen3 MoE weight conversion vs GLM-4.7-Flash unit-test loading." + "reason": "Different model families and different loading paths; not the same underlying bug." }, { "left": "issue:39692", "right": "issue:43994", "accept": false, - "reason": "Same model family only superficially; one is a docs/example plus quantization issue, the other is nonsensical outputs with AutoModel/pipeline." + "reason": "Both involve SigLIP2, but the reported failures are different: docs/example mismatch vs nonsensical outputs with AutoModel/pipeline." }, { "left": "issue:42915", "right": "issue:44756", "accept": false, - "reason": "Unrelated issues: Qwen3 MoE FP8 failure vs disabling mmap on Strix Halo to avoid OOM." + "reason": "Unrelated issues: Qwen3Moe FP8 failure vs Strix Halo mmap/OOM." }, { "left": "issue:42491", "right": "issue:43931", "accept": false, - "reason": "Different model-specific failures; no evidence they share the same bug." + "reason": "Different loading failures with different models and symptoms." }, { "left": "issue:43856", "right": "issue:44387", "accept": false, - "reason": "Both are memory-related, but the concrete regressions and triggers are different." + "reason": "One is Qwen3 MoE training memory inefficiency, the other is int4 reserved-memory growth causing OOM." }, { "left": "issue:44661", "right": "issue:44683", "accept": false, - "reason": "Different subsystems entirely: tokenizer mapping/model registration vs compiled flex_attention on torch >= 2.9." + "reason": "Tokenizer-mapping bug vs compiled flex_attention failure on newer torch; unrelated." }, { "left": "issue:43644", "right": "issue:43645", "accept": false, - "reason": "Both are Transformers 5.0.0 regressions, but they affect different initialization paths and are not the same bug." + "reason": "Both are Transformers 5.x regressions, but they affect different behaviors and code paths." }, { "left": "issue:44910", "right": "issue:45588", "accept": false, - "reason": "Different flash-attention crashes with different triggers and failure modes." + "reason": "Different flash-attention bugs with different triggers and failure modes." }, { "left": "issue:45362", "right": "issue:45406", "accept": false, - "reason": "Different model/component failures: Qwen3.5 chat crash vs Gemma4Processor missing _tokenizer." + "reason": "Different model-specific crashes affecting different components (chat vs serve/processor)." }, { "left": "issue:42491", "right": "issue:45292", "accept": false, - "reason": "Different issues: Qwen3 MoE LoRA compatibility vs resize_token_embeddings not updating output embeddings." + "reason": "LoRA/model-compatibility issue vs resize_token_embeddings not updating output embeddings." }, { "left": "issue:44488", "right": "issue:44991", "accept": false, - "reason": "Both are load failures, but the affected models and likely causes differ." + "reason": "Both are tokenizer/loading failures, but for different models and without evidence of a shared root cause." }, { - "left": "issue:43404", - "right": "issue:43454", + "left": "issue:36010", + "right": "issue:39404", "accept": false, - "reason": "Related to weight tying, but they are different models with different manifestations." + "reason": "ImportError for GenerationMixin and Whisper return_language pipeline regression are unrelated." }, { - "left": "issue:36010", - "right": "issue:39404", + "left": "issue:43404", + "right": "issue:43454", "accept": false, - "reason": "Different breakages: missing GenerationMixin import vs Whisper pipeline return_language regression." + "reason": "Both mention lm_head tying, but they affect different model classes/configs and are not the same code path." }, { "left": "issue:45276", "right": "issue:45335", - "accept": true, - "reason": "Same underlying resize_token_embeddings propagation bug across closely related model implementations." + "accept": false, + "reason": "Similar resize_token_embeddings wording, but different architectures and affected submodules." }, { "left": "issue:44164", "right": "issue:44291", "accept": false, - "reason": "Different loading/saving problems: extra_state handling vs unexpected _is_hf_initialized during init_empty_weights." + "reason": "save/from_pretrained extra_state handling and init_empty_weights argument handling are distinct bugs." }, { "left": "issue:44402", "right": "issue:44488", "accept": false, - "reason": "Tokenizer vocab-size mismatch vs a separate model-loading failure for another model." + "reason": "Tokenizer vocab-size mismatch vs model loading failure; different symptoms and likely different fixes." }, { "left": "issue:39692", "right": "issue:44960", "accept": false, - "reason": "Only loosely related by vision/model family; the concrete bugs differ." + "reason": "Different models and different errors; no shared concrete bug." }, { - "left": "issue:43335", - "right": "issue:45663", + "left": "issue:43425", + "right": "issue:43957", "accept": false, - "reason": "Different structural bugs in different models: sparse-layer config vs FSDP2 shared_kv_states handling." + "reason": "Torch version incompatibility is not the same as meta-device loading breakage in Transformers 5.x." }, { - "left": "issue:43425", - "right": "issue:43957", + "left": "issue:43335", + "right": "issue:45663", "accept": false, - "reason": "Unrelated: torch 2.10 compatibility vs model loading with torch.device(\"meta\")." + "reason": "Sparse-layer config bug vs Gemma-4 FSDP2 attention-state KeyError; unrelated." }, { "left": "issue:42947", "right": "issue:45704", "accept": false, - "reason": "Different topics: gradient checkpointing with PEFT LoRA vs an apex.FusedRMSNorm memory leak." + "reason": "Gradient checkpointing with PEFT LoRA is unrelated to the apex RMSNorm memory leak." }, { "left": "issue:45663", "right": "issue:45759", "accept": false, - "reason": "Different model-family issues: Gemma-4 FSDP2 attention state vs unwrapping text_config for composite Qwen configs." + "reason": "Different code paths: Gemma-4 FSDP2 attention state bug vs Qwen multimodal config unwrapping bug." }, { "left": "issue:43873", "right": "issue:44387", "accept": false, - "reason": "Both are memory-related, but the concrete bugs are different and not mergeable as one fix." + "reason": "Both touch memory usage, but the mechanisms and fixes are different." }, { "left": "issue:43854", "right": "issue:45216", "accept": false, - "reason": "Different regressions: GLM-4.7-Flash loading in tests vs Qwen3.5 save_pretrained checkpoint corruption." + "reason": "Different Qwen3-related issues: model loading in tests vs incorrect saved checkpoints after a version change." }, { "left": "issue:44387", "right": "issue:44756", "accept": false, - "reason": "Different performance/memory problems with different triggers; no shared underlying bug." + "reason": "Unrelated OOM scenarios: int4 quantization reserved memory vs Strix Halo mmap." }, { "left": "issue:44291", "right": "issue:44756", "accept": false, - "reason": "Different failure modes: init_empty_weights type error vs mmap-related OOM mitigation." + "reason": "Init-empty-weights argument handling is unrelated to Strix Halo mmap/OOM." } ] }, "evaluator_result": { "accept": false, - "feedback": "The cluster summary is mostly grounded, but one soft-edge acceptance is too strong: issue:44451 vs issue:44488. From the packet we only know both are model-loading failures for different BERT models; that is not enough to conclude they are the same regression. The other accepted pairs are more defensible, but this overstates the evidence, so I cannot accept the review." + "feedback": "The cluster-level choice is mostly grounded, and the accepted duplicate (issue:43644 \u2194 issue:44534) looks reasonable from the titles. However, the writeup overstates the evidence by calling that duplicate \u201cconfirmed\u201d; the packet only shows a high-similarity soft pair, not an explicit confirmation. Please soften that language to something like \u201cstrong soft duplicate candidate\u201d or \u201clikely duplicate.\u201d" }, "evaluator_used": true, "retried": true, @@ -2034,8 +1998,8 @@ "issue:43425|issue:43828", "issue:44589|issue:44756", "issue:44479|issue:45663", - "issue:43906|issue:44530", "issue:39404|issue:45701", + "issue:43906|issue:44530", "issue:43611|issue:43645", "issue:43425|issue:43716", "issue:43931|issue:44863", @@ -2051,259 +2015,259 @@ "issue:44589|issue:45084", "issue:40990|issue:42947" ], - "prepared_review_unit_hash": "31a9adc47ac252851a14391345bb3396f086dc5769a292cc2300512ca850e2e4", + "prepared_review_unit_hash": "7c0f34728ef80cc4f31b35faa8675901804b63deefd9bc39ff6125fe320ea4d6", "component_budget": { "node_count": 387, "item_count": 387, "soft_pair_count": 1455, - "serialized_chars": 749552, - "estimated_input_tokens": 187388, - "estimated_eval_tokens": 375032 + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 36, - "serialized_chars": 38396, - "estimated_input_tokens": 9599, - "estimated_eval_tokens": 19454 + "serialized_chars": 38397, + "estimated_input_tokens": 9600, + "estimated_eval_tokens": 19456 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Mostly a conservative cluster of Transformers v5/model-loading regressions, with only the resize_token_embeddings family showing clear duplicate-level overlap. The 44402/44451 tokenizer-vocab pair should stay rejected given the limited evidence for 44451.", - "confidence": 0.71, - "canonical_issue_reason": "issue:43611 is the clearest center of gravity: a concrete Transformers 5 loading regression in custom-model handling via base_model_prefix, representative of the broader cluster without being overly model-specific.", + "summary": "Mostly a heterogeneous issue bucket with a few nearby Transformers v5 loading regressions. Only one soft edge looks like a true duplicate-style pair; the rest are related by subsystem or model family but not the same concrete bug.", + "confidence": 0.57, + "canonical_issue_reason": "Issue 43611 is the best representative of the broad Transformers 5 loading-regression theme: it is generic, concrete, and closer to several nearby loading/init reports than the more model-specific outliers.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43611 is the best anchor for the issue cluster because it is concrete, broadly representative of the repeated v5 custom-model loading failures, and less narrow than the model-specific variants.", + "best_issue_reason": "43611 is the strongest single anchor for the cluster because it captures the most reusable underlying failure mode (v5 model-loading breakage) without being overly narrow to one model or environment.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44292", "right": "issue:44912", "accept": false, - "reason": "Both are loading/quantization-related, but they concern different models and different quantization behaviors; too broad to treat as the same bug." + "reason": "Different models and quantization paths; both are load/quantization regressions, but not the same concrete bug." }, { "left": "issue:42947", "right": "issue:43381", "accept": false, - "reason": "Both mention gradient checkpointing, but one is about LoRA ineffectiveness and the other is about eval-mode restriction; different failures." + "reason": "Both concern gradient checkpointing, but one is PEFT/LoRA ineffectiveness and the other is eval-mode incompatibility." }, { "left": "issue:45684", "right": "issue:45698", "accept": false, - "reason": "Both involve save/from_pretrained and custom modules, but one is about permissions and the other is loading the wrong module; distinct code paths." + "reason": "Both involve save/from_pretrained workflows, but one is read-only permission propagation and the other is loading the wrong custom module." }, { "left": "issue:45276", "right": "issue:45292", "accept": true, - "reason": "Same underlying resize_token_embeddings propagation bug: both report updated embeddings not being reflected after resizing, with Gemma4 as a model-specific instance." + "reason": "Same resize_token_embeddings bug family: both report missing propagation to output embeddings; 45276 is the broader Gemma4 manifestation." }, { "left": "issue:43381", "right": "issue:44928", "accept": false, - "reason": "Gradient checkpointing in eval mode is unrelated to the Qwen3.5 RLHF NaN/position_ids regression." + "reason": "Unrelated failures: eval-mode checkpointing restriction versus Qwen3.5 RLHF NaNs from 3D position_ids/SDPA fallback." }, { "left": "issue:44164", "right": "issue:44756", "accept": false, - "reason": "One is a save/from_pretrained extra_state bug; the other is an mmap/OOM issue. Different code paths and symptoms." + "reason": "Different areas entirely: extra_state save/load handling versus mmap OOM on Strix Halo." }, { "left": "issue:45588", "right": "issue:45663", "accept": false, - "reason": "Flash-attention sinkless-model crash and Gemma-4 FSDP2 shared_kv_states KeyError are distinct runtime/model failures." + "reason": "Both are attention-related, but one is a flash_attention null-aux crash and the other is a Gemma4 FSDP2 shared_kv_states bug." }, { "left": "issue:41720", "right": "issue:44155", "accept": false, - "reason": "Different models, different symptoms, and no concrete evidence of a shared root cause beyond a vague inference/batched-inference overlap." + "reason": "Different model families and symptoms: Qwen3 auto device map assert versus AudioFlamingo3 batched inference token leakage." }, { "left": "issue:43646", "right": "issue:43708", "accept": false, - "reason": "Custom model initialization in notebooks and resume_from_checkpoint max_steps accounting are unrelated bugs." + "reason": "Both mention custom model behavior, but one is initialization breakage and the other is Trainer checkpoint-resume step accounting." }, { "left": "issue:44479", "right": "issue:44877", "accept": false, - "reason": "A video-input regression and strict config handling for granite_speech are different issues." + "reason": "Video-input regression versus strict config loading; different code paths and failure modes." }, { "left": "issue:43854", "right": "issue:44960", "accept": false, - "reason": "GLM-4.7-Flash load failure and GLM5 are different model issues with no concrete shared bug shown." + "reason": "Model loading in tests versus a generic GLM5 issue; too little evidence of the same underlying bug." }, { "left": "issue:43957", "right": "issue:44756", "accept": false, - "reason": "Meta-device loading regressions and Strix Halo mmap OOM are different failure mechanisms." + "reason": "Meta-device loading breakage and mmap OOM are both loading-related but clearly distinct problems." }, { "left": "issue:45292", "right": "issue:45335", - "accept": true, - "reason": "Same resize_token_embeddings bug: output_embeddings and decoder.embed_tokens both fail to update after resizing, across model variants." + "accept": false, + "reason": "Related resize_token_embeddings theme, but the affected modules and model families differ enough to avoid merging as one bug." }, { "left": "issue:44402", "right": "issue:44451", "accept": false, - "reason": "The shared tokenizer/vocab_size explanation is not sufficiently grounded for issue:44451; its title is too generic to confirm the same underlying bug." + "reason": "Different models with similar load failures; no clear shared code-path from the titles alone." }, { "left": "issue:43645", "right": "issue:44164", "accept": false, - "reason": "Notebook custom-model initialization and extra_state save/load handling are not the same bug." + "reason": "Custom-model Jupyter initialization versus extra_state save/load handling are different bugs." }, { "left": "issue:43828", "right": "issue:43957", "accept": false, - "reason": "Autocast dtype mismatch in Phi-tiny-MoE is unrelated to meta-device loading breakage." + "reason": "Different dtype/autocast and meta-device loading issues; both are compatibility bugs but not the same one." }, { "left": "issue:43257", "right": "issue:44863", "accept": false, - "reason": "Both are checkpoint-loading problems, but they involve different model families and distinct root causes." + "reason": "Different model-loading failures: Qwen3 MoE with accelerate+deepspeed versus NemotronH checkpoint loading." }, { "left": "issue:43425", "right": "issue:43828", "accept": false, - "reason": "Torch version incompatibility and a specific autocast dtype mismatch are different issues." + "reason": "Torch version incompatibility versus autocast dtype mismatch; related only at a very high level." }, { "left": "issue:44589", "right": "issue:44756", "accept": false, - "reason": "Float8 storage lookup failure and Strix Halo mmap OOM are different errors." + "reason": "Float8 storage lookup error is unrelated to the Strix Halo mmap/OOM report." }, { "left": "issue:44479", "right": "issue:45663", "accept": false, - "reason": "Video-input regression and Gemma-4 FSDP2 attention-state bug are unrelated." + "reason": "Different regressions: video input handling versus Gemma4 FSDP2 shared_kv_states rebuild." }, { - "left": "issue:43906", - "right": "issue:44530", + "left": "issue:39404", + "right": "issue:45701", "accept": false, - "reason": "Isolated reproduction of an older issue and a PagedAttentionCache linear_attention crash are not the same bug." + "reason": "Whisper pipeline return_language regression and tokenizer-version behavior changes are unrelated." }, { - "left": "issue:39404", - "right": "issue:45701", + "left": "issue:43906", + "right": "issue:44530", "accept": false, - "reason": "Whisper return_language pipeline behavior and tokenizer changes across versions are unrelated." + "reason": "Isolated reproduction of an earlier bug versus a Qwen3.5 cache/group-type crash; not the same issue." }, { "left": "issue:43611", "right": "issue:43645", "accept": false, - "reason": "Both are Transformers 5 custom-model regressions, but one is base_model_prefix loading and the other is notebook-specific model definition/init behavior." + "reason": "Both are Transformers 5 loading regressions, but base_model_prefix loading and notebook-defined custom model init are different failures." }, { "left": "issue:43425", "right": "issue:43716", "accept": false, - "reason": "Torch compatibility issues and Mistral-3 image-preprocessor dtype mismatch are different issues." + "reason": "Torch compatibility issue versus a Mistral-3 image preprocessor/model dtype mismatch." }, { "left": "issue:43931", "right": "issue:44863", "accept": false, - "reason": "Qwen3-VL weight-shape mismatch and NemotronH checkpoint loading failure are different model-specific bugs." + "reason": "Both are model loading errors, but Qwen3-VL shape mismatch and NemotronH checkpoint loading are distinct." }, { "left": "issue:44530", "right": "issue:45084", "accept": false, - "reason": "PagedAttentionCache group-type crash and template-node compile error are different code paths." + "reason": "PagedAttentionCache linear_attention crash and non-template-node compile error do not share the same bug." }, { "left": "issue:43901", "right": "issue:44509", "accept": false, - "reason": "Both are docs-related, but they point at different outdated pipeline topics, not the same documentation bug." + "reason": "Docs/API mismatch around return_all_scores is unrelated to removed pipeline-task documentation." }, { "left": "issue:39404", "right": "issue:44912", "accept": false, - "reason": "Whisper pipeline language-return regression is unrelated to MXFP4/bf16 quantization fallback." + "reason": "Different models and APIs: Whisper pipeline language return versus MXFP4 quantization fallback." }, { "left": "issue:44863", "right": "issue:45663", "accept": false, - "reason": "NemotronH checkpoint loading and Gemma-4 FSDP2 attention-state rebuilding are unrelated model issues." + "reason": "Different loading paths and symptoms; no evidence they are the same underlying bug." }, { "left": "issue:44265", "right": "issue:44898", "accept": false, - "reason": "torch.export with torch_compilable_check and Perceiver non-default-resolution failure are different problems." + "reason": "torch.export/torch_compilable_check failure is unrelated to Perceiver non-default-resolution resizing." }, { "left": "issue:43299", "right": "issue:43931", "accept": false, - "reason": "Both involve Qwen3-VL loading, but one is a v5.dev0 loading regression and the other is a shape-mismatch report; not enough evidence they are the same bug." + "reason": "Same model family, but one is a broad Transformers 5.0 loading regression and the other is a specific weight-shape mismatch; not enough to treat as one bug." }, { "left": "issue:44479", "right": "issue:44805", "accept": false, - "reason": "Video input regression and mask/tensor shape IndexError are unrelated." + "reason": "Video-input regression versus a mask/tensor shape IndexError are not the same failure." }, { "left": "issue:43425", "right": "issue:44918", "accept": false, - "reason": "Torch compatibility issues and Qwen3.5 embedding unpacking in TRL SFT are different failures." + "reason": "Torch 2.10 compatibility and Qwen3.5 embedding unpacking are unrelated." }, { "left": "issue:43856", "right": "issue:44756", "accept": false, - "reason": "Qwen3 MoE training memory inefficiency and Strix Halo mmap OOM are different resource problems." + "reason": "Memory inefficiency in Qwen3 MoE training is different from mmap-triggered OOM on Strix Halo." }, { "left": "issue:44589", "right": "issue:45084", "accept": false, - "reason": "Float8 storage lookup failure and compile error on non-template nodes are unrelated." + "reason": "Float8 storage lookup and template-node compilation errors are unrelated." }, { "left": "issue:40990", "right": "issue:42947", "accept": false, - "reason": "Extremely high perplexity on gpt-oss-20b and gradient-checkpointing/LoRA behavior are unrelated." + "reason": "High perplexity on gpt-oss-20b is unrelated to gradient checkpointing with PEFT LoRA." } ] }, "evaluator_result": { - "accept": false, - "feedback": "The overall cluster framing is mostly conservative, but one soft-edge acceptance is overstated: issue:44402 vs issue:44451. The packet only gives a generic title for 44451 (\u201cLatest version cannot load 'vesteinn/ScandiBERT'\u201d), so the claim that both are the same tokenizer/vocab_size mismatch is not grounded enough from the provided evidence. The resize_token_embeddings acceptances are well-supported, but this one should stay rejected or be marked tentative." + "accept": true, + "feedback": "The summary is grounded in the packet and stays conservative overall. The single accepted soft edge (45276/45292) is well supported by the titles and appears mergeable as the same resize_token_embeddings bug family. The other pairwise rejections are cautious and do not overclaim duplicate status. The canonical/best-issue framing is subjective but reasonable for this heterogeneous cluster." }, "evaluator_used": true, - "retried": true, + "retried": false, "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null @@ -2404,239 +2368,230 @@ "issue:36032|issue:43645", "issue:45084|issue:45507" ], - "prepared_review_unit_hash": "3900e3440555db2056f9842c517ee1b51d90580dc7824ef4689aa6fcf1839b3a", + "prepared_review_unit_hash": "de8280b2fb7c09912c0e1c31e9b806256111b844c9b17f9287be5e1143152ee7", "component_budget": { "node_count": 387, "item_count": 387, "soft_pair_count": 1455, - "serialized_chars": 749552, - "estimated_input_tokens": 187388, - "estimated_eval_tokens": 375032 + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 }, "budget": { "node_count": 47, "item_count": 47, "soft_pair_count": 33, - "serialized_chars": 37028, - "estimated_input_tokens": 9257, - "estimated_eval_tokens": 18770 + "serialized_chars": 37023, + "estimated_input_tokens": 9256, + "estimated_eval_tokens": 18768 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is thematically similar but not duplicate-tight: it mixes many separate Transformers v5 regressions, model-specific loading bugs, and a few unrelated runtime issues. None of the soft pairs look like the same underlying bug/change, so they should stay separate.", - "confidence": 0.82, - "canonical_issue_reason": "Issue 43644 is the broadest representative of the cluster: a core Transformers v5 regression affecting model initialization/state handling across models, which makes it the best umbrella issue among these mostly separate bugs.", + "summary": "The set is mostly a collection of unrelated closed issue reports across many subsystems. I only see one plausible duplicate family: the GGUF loading regressions. No PRs are present.", + "confidence": 0.66, + "canonical_issue_reason": "issue:43452 is the best anchor because it describes the shared broken `gguf_file`/`from_pretrained` loading path directly and covers both tokenizer and CausalLM loading, making it the most general and actionable report in the clear duplicate family.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 43644 is the strongest single anchor because it is a framework-level regression with wider impact than the more model-specific reports.", + "best_issue_reason": "issue:43452 is the strongest standalone issue here: it names the concrete loader regression rather than a single model-specific symptom, and the Qwen2.5 GGUF report can naturally attach to it.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43425", "right": "issue:45685", "accept": false, - "reason": "Different failures: Torch version compatibility vs an MPS-specific histogram/int dtype error." + "reason": "Both are v5-era failures, but one is Torch-version incompatibility and the other is an MPS histogram/int-type bug; different code paths." }, { "left": "issue:39404", "right": "issue:44671", "accept": false, - "reason": "Whisper pipeline return_language regression and CamemBERT masked-LM predictions are unrelated code paths." + "reason": "Whisper `return_language` pipeline behavior and CamemBERT masked-LM predictions are unrelated model-specific bugs." }, { "left": "issue:44671", "right": "issue:44991", "accept": false, - "reason": "One is incorrect masked-LM output; the other is tokenizer loading failure for a different model." + "reason": "Incorrect CamemBERT predictions and tokenizer loading failure for `EMBIDDIA/est-roberta` are different problems." }, { "left": "issue:43994", "right": "issue:44079", "accept": false, - "reason": "SigLIP inference quality bug vs ModelOutput key assignment logic bug are not the same issue." + "reason": "SigLIP2 nonsensical outputs and `ModelOutput` key assignment with prior `None` are distinct issues." }, { "left": "issue:36032", "right": "issue:43452", "accept": false, - "reason": "Tokenizer special-token method conflict is unrelated to gguf_file loading breakage." + "reason": "T5 tokenizer special-token method conflict and GGUF loader breakage are different bug classes." }, { "left": "issue:44164", "right": "issue:45759", "accept": false, - "reason": "Save/load extra_state handling and composite config unwrapping are different serialization/config bugs." + "reason": "`extra_state` save/load handling and `AutoModelForCausalLM.from_config` text_config unwrapping are unrelated serialization/config bugs." }, { "left": "issue:43596", "right": "issue:44683", "accept": false, - "reason": "DeepSpeed ZeRO-3/BertModel indexing error is unrelated to compiled flex_attention on newer Torch." + "reason": "Deepspeed ZeRO3 init index error and compiled flex_attention failing on newer Torch versions are different failures." }, { "left": "issue:44991", "right": "issue:45081", "accept": false, - "reason": "Different model-loading regressions: a repo-specific tokenizer failure vs a Mistral regex patch backend attribute error." + "reason": "Both are loading regressions, but one is a tokenizer load failure and the other is a Mistral regex patch crash; not the same code path." }, { "left": "issue:32090", "right": "issue:39404", "accept": false, - "reason": "Trainer GPU broadcast NoneType error and Whisper return_language pipeline failure are unrelated." + "reason": "Trainer GPU broadcast NoneType error and Whisper pipeline `return_language` regression are unrelated." }, { "left": "issue:44164", "right": "issue:44479", "accept": false, - "reason": "extra_state save/load handling is not the same as the Qwen video-input regression." + "reason": "`extra_state` serialization and Qwen video-input regression do not share a concrete bug." }, { "left": "issue:44488", "right": "issue:45081", "accept": false, - "reason": "Both are load-time failures, but they involve different models and different root causes." + "reason": "Both involve model/tokenizer loading, but they affect different models and different failure mechanisms." }, { "left": "issue:40990", "right": "issue:43421", "accept": false, - "reason": "High perplexity on a model is unrelated to tokenizer post-processor updates for runtime special-token changes." + "reason": "High perplexity on GPT-OSS and runtime special-token post-processor updates are not the same issue." }, { "left": "issue:43582", "right": "issue:43632", "accept": false, - "reason": "Apple Silicon warmup TypeError and the _is_hf_initialized flag regression are separate behaviors." + "reason": "Apple Silicon allocator warmup TypeError and `_is_hf_initialized` regression are different problems." }, { "left": "issue:43883", "right": "issue:44488", "accept": false, - "reason": "Missing tied-weights attribute in Molmo is unrelated to a specific BERT model loading failure." + "reason": "Missing `all_tied_weights_keys` on Molmo and failure to load `cjvt/sleng-bert` are unrelated." }, { "left": "issue:43611", "right": "issue:43644", "accept": false, - "reason": "Both are v5 regressions, but one is about base_model_prefix loading and the other about non-persistent buffer initialization." + "reason": "Both are v5 regression reports, but one is about `base_model_prefix` loading and the other about non-persistent buffer serialization; not the same bug." }, { "left": "issue:44206", "right": "issue:44220", "accept": false, - "reason": "Audio feature-extraction failures, but not the same concrete bug: unsupported argument vs fbanks helper error." + "reason": "Unsupported `center` arg in LasrFeatureExtractor and `_torch_extract_fbank_features()` are too different to call duplicate from titles alone." }, { "left": "issue:43716", "right": "issue:45084", "accept": false, - "reason": "Mistral image dtype mismatch is unrelated to a template compilation error." + "reason": "Mistral image-preprocessor dtype mismatch and template compilation TypeError are unrelated." }, { "left": "issue:44488", "right": "issue:44987", "accept": false, - "reason": "Two different repository-specific loading failures with no clear shared code-path." + "reason": "Different model-loading failures for different models; no clear shared code-path." }, { "left": "issue:34567", "right": "issue:40444", "accept": false, - "reason": "TrainerState token counting and multimodal IterableDataset finetuning failures are unrelated." + "reason": "TrainerState token-count tracking and multi-image IterableDataset finetuning failure are unrelated." }, { "left": "issue:43643", "right": "issue:44530", "accept": false, - "reason": "trust_remote_code config field loss is a config serialization issue, not the PagedAttentionCache linear_attention crash." + "reason": "`trust_remote_code` missing fields and Qwen PagedAttentionCache group-type crash are different bugs." }, { "left": "issue:44811", "right": "issue:45584", "accept": false, - "reason": "Both involve Whisper text handling, but one is batch_decode skip_special_tokens behavior and the other is empty-transcription generation after alignment." + "reason": "Whisper skip-special-tokens decode behavior and empty-transcription generation failure are both Whisper-related but not the same concrete defect." }, { "left": "issue:43644", "right": "issue:43646", "accept": false, - "reason": "Shared v5 theme, but distinct bugs: junk non-persistent buffers vs custom model initialization failure." + "reason": "Both are v5 regressions in model initialization/serialization, but they describe different root causes and fixes." }, { "left": "issue:43883", "right": "issue:44991", "accept": false, - "reason": "Different models and symptoms: missing all_tied_weights_keys vs tokenizer loading regression." + "reason": "Molmo tied-weights attribute missing and tokenizer loading failure for `est-roberta` are unrelated." }, { "left": "issue:36032", "right": "issue:42915", "accept": false, - "reason": "Tokenizer special-token method conflict is unrelated to Qwen3Moe FineGrainedFP8Config failure." + "reason": "T5 tokenizer special-token conflict and Qwen3Moe FP8 config failure are different issues." }, { "left": "issue:44451", "right": "issue:44987", "accept": false, - "reason": "Different model-loading failures for different repos; no evidence of the same underlying defect." + "reason": "Separate model-loading regressions affecting different repositories/models." }, { "left": "issue:43541", "right": "issue:45084", "accept": false, - "reason": "Torch Dynamo grouped_mm error on Mixtral is unrelated to a non-template-node compile error." + "reason": "Mixtral Dynamo tracing error and template-node compilation error are distinct code-path failures." }, { "left": "issue:43452", "right": "issue:43482", - "accept": false, - "reason": "Both concern GGUF loading, but one is a generic gguf_file API break and the other is a specific Qwen2.5-GGUF v5 failure." + "accept": true, + "reason": "Both point to the same GGUF loading regression in the `gguf_file`/`from_pretrained` path; the Qwen2.5 report looks like a specific instance of the broader breakage." }, { "left": "issue:42915", "right": "issue:44530", "accept": false, - "reason": "FP8 config failure and PagedAttentionCache linear_attention crash are different issues." + "reason": "Qwen3Moe FP8 config failure and Qwen3.5 PagedAttentionCache crash are different problems." }, { "left": "issue:43317", "right": "issue:43482", "accept": false, - "reason": "GPU/CPU offload dequantization failure is not the same as a Qwen2.5-GGUF loading regression." + "reason": "GPU+CPU offload loading failure and GGUF loading failure are different loader paths." }, { "left": "issue:44849", "right": "issue:45325", "accept": false, - "reason": "Qwen3.5 hidden_states bug and Qwen2.5-VL rope-index scaling bug are separate model behaviors." + "reason": "Qwen3.5 hidden-state output bug and Qwen2.5-VL rope-index scaling bug are separate model-specific regressions." }, { "left": "issue:43404", "right": "issue:44466", "accept": false, - "reason": "Un-tied lm_head weights in Mistral3 and inconsistent serialization of lm_head.weight are related in topic but not the same fixable bug." - }, - { - "left": "issue:36032", - "right": "issue:43645", - "accept": false, - "reason": "Tokenizer method naming conflict is unrelated to Jupyter custom-model initialization breakage." - }, - { - "left": "issue:45084", - "right": "issue:45507", - "accept": false, - "reason": "Template compilation failure and GraniteMoEHybrid invalid method call are different code paths." + "reason": "Both involve tied `lm_head.weight`, but one is a model-class tying bug and the other is device-dependent serialization inconsistency; not the same issue." } ] }, - "evaluator_result": null, - "evaluator_used": false, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet and stays conservative: most pairs are clearly unrelated from the titles, and the only accepted soft edge (43452/43482) is a plausible shared GGUF loading regression. No PRs are present, and nothing is overstated." + }, + "evaluator_used": true, "retried": false, - "accepted_nontrivial_soft_edge": false, + "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null }, @@ -2722,8 +2677,8 @@ "issue:43632|issue:44756", "issue:43296|issue:43531", "issue:44062|issue:44987", - "issue:42915|issue:44291", "issue:36032|issue:43742", + "issue:42915|issue:44291", "issue:43531|issue:44877", "issue:43611|issue:44534", "issue:42915|issue:44589", @@ -2731,223 +2686,223 @@ "issue:44479|issue:45325", "issue:43317|issue:43856", "issue:43425|issue:44403", - "issue:44079|issue:45663", - "issue:42175|issue:45542" + "issue:42175|issue:45542", + "issue:44079|issue:45663" ], - "prepared_review_unit_hash": "64a0042d1e3eb0bc17bbaf05a45dd6c1b9bf9da945985fad309882d18134b1a1", + "prepared_review_unit_hash": "a3188fa4952733c47c24bfe3ac309303f95669c26cae32d6c72e54b28ae3d1c3", "component_budget": { "node_count": 387, "item_count": 387, "soft_pair_count": 1455, - "serialized_chars": 749552, - "estimated_input_tokens": 187388, - "estimated_eval_tokens": 375032 + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 30, - "serialized_chars": 36625, - "estimated_input_tokens": 9157, - "estimated_eval_tokens": 18570 + "serialized_chars": 36624, + "estimated_input_tokens": 9156, + "estimated_eval_tokens": 18568 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Mostly a heterogeneous set of issue reports, with only a couple of clear duplicate-like pairs. The strongest duplicate candidates are the Qwen3.5 save_pretrained regression pair and the TensorFlow-backend dependency pair; the rest are separate model-specific or subsystem-specific bugs.", - "confidence": 0.74, - "canonical_issue_reason": "issue:43611 is the best overall anchor because it describes a broad, high-signal Transformers v5 loading regression with wide potential impact; most other items are narrower model-specific failures or unrelated issues.", + "summary": "This cluster is mostly heterogeneous model-loading/regression reports, so most soft edges are false positives. I only see two plausible duplicate families: Qwen save_pretrained checkpoint corruption and Qwen video/rope-index regression. No PRs are present.", + "confidence": 0.71, + "canonical_issue_reason": "Issue 44479 is the broadest concrete regression here: it captures a shared Qwen video-input failure across multiple variants, and it cleanly represents the narrower Qwen2.5-VL temporal-position bug in 45325.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43611 is the strongest representative issue among the set, but only as a broad v5-loading regression anchor rather than a true umbrella duplicate for the whole cluster.", + "best_issue_reason": "44479 is the best standalone representative because it describes a clear versioned regression with a shared code path and multiple affected models, making it the most central issue among the few genuinely related reports.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43262", "right": "issue:43377", "accept": false, - "reason": "Different bug classes: audio chat-template sampling-rate default vs MIMI batch/single padding-mask mismatch." + "reason": "Both are audio-related, but one is a processor chat-template sampling-rate default and the other is a MIMI padding-mask batching bug; different code paths." }, { "left": "issue:43646", "right": "issue:44534", "accept": false, - "reason": "Both are Transformers v5 regressions, but one is custom model init while the other is non-persistent buffer initialization; different code paths and symptoms." + "reason": "One is custom model initialization in v5, the other is non-persistent buffers being filled with junk; related release, different bug." }, { "left": "issue:44987", "right": "issue:44991", "accept": false, - "reason": "Both are loading failures, but they affect different models and likely different causes; no clear shared code-path bug." + "reason": "Both fail on transformers>=5, but one is model loading for `physical-intelligence/fast` and the other is tokenizer loading for EMBEDDIA/est-roberta; not the same failure." }, { "left": "issue:43994", "right": "issue:44610", "accept": false, - "reason": "Different multimodal regressions: nonsensical outputs vs wrong processor image size." + "reason": "Different multimodal models and different symptoms: nonsensical outputs vs processor/model input-size mismatch." }, { "left": "issue:43322", "right": "issue:43540", "accept": false, - "reason": "Unrelated model families and failures: Llava Next segfault vs Qwen3OmniMoe video ValueError." + "reason": "Both are multimodal, but one is a Llava Next load segfault and the other is a Qwen3OmniMoe video ValueError; no shared bug evidence." }, { "left": "issue:44403", "right": "issue:45081", "accept": false, - "reason": "General loading noise vs a specific Mistral regex patch crash; not the same bug." + "reason": "Noise on transformer loading versus a Mistral tokenizer regex crash; unrelated code paths." }, { "left": "issue:43377", "right": "issue:43819", "accept": false, - "reason": "Both are audio-related, but one is padding-mask batching behavior and the other is DAC latent/STE mismatch; different underlying issues." + "reason": "Both involve audio model outputs, but one is batched-vs-single padding behavior and the other is DAC STE mismatch; distinct root causes." }, { "left": "issue:45216", "right": "issue:45357", "accept": true, - "reason": "Same underlying Qwen3.5 save_pretrained regression: incorrect checkpoint/visual encoder key serialization across versions." + "reason": "Both report the same Qwen3.5 `save_pretrained` checkpoint corruption/incorrect visual-encoder key saving across versions; one is the broader regression, the other a specific manifestation." }, { "left": "issue:43645", "right": "issue:44534", "accept": false, - "reason": "Different v5 regressions affecting custom model initialization vs buffer loading state." + "reason": "Custom model initialization in notebooks is unrelated to filling non-persistent buffers with junk, despite both being v5 regressions." }, { "left": "issue:44451", "right": "issue:45081", "accept": false, - "reason": "Different failures: a model load failure for ScandiBERT vs a Mistral tokenizer regex patch crash." + "reason": "Different model families and failures: a model load failure for ScandiBERT versus a Mistral tokenizer regex patch crash." }, { "left": "issue:42915", "right": "issue:45084", "accept": false, - "reason": "Different issues: FineGrainedFP8Config load failure vs a template compilation TypeError." + "reason": "FineGrainedFP8Config load failure and a template-node compile TypeError are separate problems." }, { "left": "issue:44265", "right": "issue:44610", "accept": false, - "reason": "torch.export/torch_compilable_check failure is unrelated to OmDet processor size mismatch." + "reason": "One is `torch.export` with `torch_compilable_check`, the other is an OmDet-Turbo processor size mismatch; different layers of the stack." }, { "left": "issue:44220", "right": "issue:44683", "accept": false, - "reason": "Audio feature extraction vs compiled flex_attention on new torch versions; different subsystems." + "reason": "Audio feature extraction vs compiled flex_attention on newer torch; no shared code path." }, { "left": "issue:44987", "right": "issue:45081", "accept": false, - "reason": "Different models and distinct failure modes; no shared concrete bug." + "reason": "A model-loading regression and a tokenizer regex patch crash are unrelated." }, { "left": "issue:43404", "right": "issue:45127", "accept": false, - "reason": "Both involve weights/ties, but one is a multimodal lm_head tying bug and the other is LoRA merge collapse after vocab extension; not the same change." + "reason": "Both touch weight tying, but one is Mistral3 `lm_head` tying and the other is LoRA merge collapse with extended vocab; not the same bug." }, { "left": "issue:41553", "right": "issue:43582", "accept": false, - "reason": "Bad AutoTokenizer error message vs Apple Silicon warmup TypeError; unrelated." + "reason": "Tokenizer-loading error for Voxtral versus Apple Silicon allocator warmup TypeError; unrelated." }, { "left": "issue:43632", "right": "issue:44756", "accept": false, - "reason": "_is_hf_initialized breakage and mmap OOM are unrelated regressions." + "reason": "Both are v5/runtime issues, but `_is_hf_initialized` flag breakage and mmap OOM on Strix Halo are different bugs." }, { "left": "issue:43296", "right": "issue:43531", "accept": false, - "reason": "PaddleOCR-VL load failure in vLLM vs Qwen3-MoE sliding_window bug; unrelated." + "reason": "PaddleOCR-VL loading in vLLM and Qwen3-MoE sliding-window behavior are unrelated." }, { "left": "issue:44062", "right": "issue:44987", "accept": false, - "reason": "Tokenizer AddedToken argument duplication vs model loading failure for a specific repo; not the same bug." + "reason": "Tokenizer AddedToken `special` keyword conflict is not the same as a model load failure for `physical-intelligence/fast`." }, { - "left": "issue:42915", - "right": "issue:44291", + "left": "issue:36032", + "right": "issue:43742", "accept": false, - "reason": "Both involve model loading, but one is FP8 config handling and the other is init_empty_weights/_is_hf_initialized constructor compatibility." + "reason": "Both are model/tokenizer load errors, but T5 `add_special_tokens` conflict and MobileLLM key error have different causes." }, { - "left": "issue:36032", - "right": "issue:43742", + "left": "issue:42915", + "right": "issue:44291", "accept": false, - "reason": "T5 tokenizer add_special_tokens conflict and MobileLLM key error are different code paths." + "reason": "Both involve model loading under v5 FP8/empty-init scenarios, but one is Qwen3Moe FineGrainedFP8Config and the other is unexpected `_is_hf_initialized` arg; distinct issues." }, { "left": "issue:43531", "right": "issue:44877", "accept": false, - "reason": "Qwen3-MoE sliding_window handling and granite_speech strict config loading are unrelated." + "reason": "Qwen3-MoE sliding-window behavior and strict config rejection for granite_speech are not the same bug." }, { "left": "issue:43611", "right": "issue:44534", "accept": false, - "reason": "Both are v5 regressions, but one is base_model_prefix loading and the other is non-persistent buffer junk; not the same underlying bug." + "reason": "Base-model-prefix loading regression and junk-filled non-persistent buffers are separate v5 breakages." }, { "left": "issue:42915", "right": "issue:44589", "accept": false, - "reason": "Both touch float8/quantization, but one is FineGrainedFP8Config and the other is missing Float8 storage support; different concrete failures." + "reason": "FP8 config load failure and missing Float8 storage type are related only loosely by dtype support, not by the same failure mode." }, { "left": "issue:40990", "right": "issue:44811", "accept": false, - "reason": "High perplexity on a language model is unrelated to Whisper batch_decode special-token handling." + "reason": "High perplexity on gpt-oss-20b and Whisper `batch_decode` ignoring `skip_special_tokens` are unrelated." }, { "left": "issue:44479", "right": "issue:45325", - "accept": false, - "reason": "Both are Qwen VL-family regressions, but one is a broad video-input regression and the other is a specific still-image rope_index scaling bug." + "accept": true, + "reason": "45325 is a precise Qwen2.5-VL rope-index/temporal-position regression that matches the broader Qwen video-input regression in 44479; same shared code path, one broader and one specific report." }, { "left": "issue:43317", "right": "issue:43856", "accept": false, - "reason": "device_map/offload loading failure and MoE training memory usage are different problems." + "reason": "Dequantized model offload failure and Qwen3 MoE training memory inefficiency are different performance/load bugs." }, { "left": "issue:43425", "right": "issue:44403", "accept": false, - "reason": "Torch version incompatibility is unrelated to loading-noise output." + "reason": "Torch 2.10 incompatibility is a dependency-compat issue, while unnecessary loading noise is a logging/UX issue." }, { - "left": "issue:44079", - "right": "issue:45663", + "left": "issue:42175", + "right": "issue:45542", "accept": false, - "reason": "ModelOutput key assignment bug and Gemma4 shared_kv_states KeyError are unrelated." + "reason": "Different packaging/backend problems: TensorFlow missing from `.[torch]` versus tensorboard installed without TensorFlow causing backend errors." }, { - "left": "issue:42175", - "right": "issue:45542", - "accept": true, - "reason": "Same underlying TensorFlow backend packaging/detection problem: TensorFlow is missing while the install exposes a TF-related backend state through optional extras." + "left": "issue:44079", + "right": "issue:45663", + "accept": false, + "reason": "`ModelOutput` key assignment bug and Gemma4 FSDP2 shared_kv_states `KeyError` are unrelated." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is grounded in the packet and the soft-edge calls stay conservative. The two accepted duplicate-like pairs are plausible from the issue titles, and the rest are appropriately treated as separate bug classes. The canonical anchor choice is also reasonable as a broad v5 loading regression, not a claim that the whole cluster is one duplicate." + "feedback": "The summary is grounded in the packet: it correctly characterizes the cluster as mostly heterogeneous and only identifies two plausible duplicate families that are supported by the issue titles. The accepted soft-edge pairs are conservative enough based on the available evidence, and the rejected pairs do not overclaim shared root causes." }, "evaluator_used": true, "retried": false, @@ -3019,11 +2974,11 @@ ], "soft_pairs": [ "issue:43421|issue:44568", - "issue:43818|issue:45357", "issue:36032|issue:41720", + "issue:43818|issue:45357", "issue:43632|issue:44534", - "issue:44683|issue:45020", "issue:43792|issue:44589", + "issue:44683|issue:45020", "issue:44403|issue:44991", "issue:33666|issue:40444", "issue:43482|issue:45538", @@ -3033,8 +2988,8 @@ "issue:44291|issue:44589", "issue:43377|issue:45491", "issue:41720|issue:45081", - "issue:44265|issue:45072", "issue:42175|issue:43296", + "issue:44265|issue:45072", "issue:43632|issue:45072", "issue:43611|issue:43646", "issue:43531|issue:44291", @@ -3044,28 +2999,28 @@ "issue:43881|issue:44488", "issue:43828|issue:44683", "issue:42915|issue:43482", - "issue:44079|issue:45072", "issue:43792|issue:43957", + "issue:44079|issue:45072", "issue:43644|issue:44079", "issue:40444|issue:45381", "issue:45325|issue:45381", "issue:43296|issue:45800", "issue:44387|issue:44617" ], - "prepared_review_unit_hash": "afc176463741799ceec5b9a3a450c0df55995564e0fa6da63996021f17f32de1", + "prepared_review_unit_hash": "d193d5342c8ec7b6520225a37ead9b2975679b1d4c6cc94421caebfa075e4c27", "component_budget": { "node_count": 387, "item_count": 387, "soft_pair_count": 1455, - "serialized_chars": 749552, - "estimated_input_tokens": 187388, - "estimated_eval_tokens": 375032 + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 33, - "serialized_chars": 37775, + "serialized_chars": 37774, "estimated_input_tokens": 9444, "estimated_eval_tokens": 19144 }, @@ -3074,210 +3029,210 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a loose collection of unrelated Transformers regressions and model-specific bugs, not a set of true duplicates. The only strong commonality is that many issues involve recent v5-era loading/init regressions, but the concrete failure modes differ.", - "confidence": 0.61, - "canonical_issue_reason": "issue:45020 is the broadest umbrella item here ('Recent transformers versions break models using `remote_code`') and best represents the shared v5-breakage theme, though it is not a true duplicate of most others.", + "summary": "The cluster is heterogeneous: most items are unrelated bug reports spanning different models, regressions, and subsystems. I found no soft-edge pair that clearly describes the same underlying bug or change, so all soft links should stay unmerged.", + "confidence": 0.92, + "canonical_issue_reason": "Issue 45020 is the broadest umbrella-style report in the set (general regressions in recent transformers versions affecting remote_code models), so it is the best representative issue if one must be chosen.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45020 is the most globally representative issue for the cluster because it covers the widest scope of recent-version loading regressions; the rest are mostly narrower, distinct manifestations.", + "best_issue_reason": "Issue 45020 is also the most suitable global issue representative because it is open, broadly scoped, and closer to a cross-cutting regression than the more model-specific reports.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43421", "right": "issue:44568", "accept": false, - "reason": "Both involve tokenizer special tokens, but one is about runtime post-processor updates and the other about BOS/EOS not being added; different bugs." + "reason": "Both involve special tokens, but one is about TokenizersBackend post-processor updates while the other is a specific mdeberta-v3 add_special_tokens regression." }, { - "left": "issue:43818", - "right": "issue:45357", + "left": "issue:36032", + "right": "issue:41720", "accept": false, - "reason": "Different models and failure modes: Video-LLaVA tower architecture vs Qwen3.5 visual-encoder save_pretrained keys." + "reason": "Completely different failures: a T5 tokenizer method conflict versus a Qwen3 auto device mapping CUDA assert." }, { - "left": "issue:36032", - "right": "issue:41720", + "left": "issue:43818", + "right": "issue:45357", "accept": false, - "reason": "Tokenizer class method conflict vs CUDA assert during auto device mapping; no shared underlying bug." + "reason": "Both touch vision models, but one is a Video-LLaVA tower/weight issue and the other is Qwen3.5 save_pretrained visual key serialization." }, { "left": "issue:43632", "right": "issue:44534", "accept": false, - "reason": "Both are v5 init-related regressions, but one is the `_is_hf_initialized` flag and the other is non-persistent buffer contents." + "reason": "Both are Transformers v5 regressions, but they describe different code paths: _is_hf_initialized versus non-persistent buffer initialization." }, { - "left": "issue:44683", - "right": "issue:45020", + "left": "issue:43792", + "right": "issue:44589", "accept": false, - "reason": "Compiled flex_attention on torch>=2.9 is unrelated to remote_code model-loading regressions." + "reason": "Unrelated runtime failures: Whisper loading versus a Float8 storage lookup error." }, { - "left": "issue:43792", - "right": "issue:44589", + "left": "issue:44683", + "right": "issue:45020", "accept": false, - "reason": "Whisper run failure and missing Float8 storage are distinct runtime issues." + "reason": "Different problem areas: torch 2.9 flex_attention compilation versus remote_code model loading regressions." }, { "left": "issue:44403", "right": "issue:44991", "accept": false, - "reason": "Generic loading noise vs tokenizer load failure for a specific model; not the same defect." + "reason": "One is about noisy logging during model loading; the other is a tokenizer loading failure for a specific model." }, { "left": "issue:33666", "right": "issue:40444", "accept": false, - "reason": "Both are multimodal training issues, but one is Qwen2-VL multi-GPU training and the other is iterable-dataset multi-image batching." + "reason": "Both are Qwen2-VL-family training issues, but they concern different model generations and different failure modes." }, { "left": "issue:43482", "right": "issue:45538", "accept": false, - "reason": "Qwen2.5-GGUF loading under v5 and CLIPTokenizer model_max_length are unrelated tokenizer/loading problems." + "reason": "GGUF loading under Transformers v5 is unrelated to CLIPTokenizer model_max_length behavior." }, { "left": "issue:44829", "right": "issue:45479", "accept": false, - "reason": "Different training pathologies: flash_attention_3 degenerate training vs single-label classification zero-loss behavior." + "reason": "Different bugs in different heads: Perceiver image resolution handling versus sequence-classification zero-loss behavior." }, { "left": "issue:40444", "right": "issue:45325", "accept": false, - "reason": "Different Qwen2.5-VL bugs: iterable-dataset multi-image prompt handling vs still-image temporal position IDs scaling." + "reason": "Both are Qwen2.5-VL-related, but one is an iterable-dataset multi-image training failure and the other is a still-image temporal position-id scaling bug." }, { "left": "issue:43482", "right": "issue:43525", "accept": false, - "reason": "Different loading failures affecting different model/config classes; no shared root cause evident." + "reason": "Unrelated: GGUF model loading failure versus missing pad_token_id on Llama4Config." }, { "left": "issue:44291", "right": "issue:44589", "accept": false, - "reason": "`init_empty_weights` argument handling and Float8 storage lookup are separate v5 compatibility problems." + "reason": "Different initialization issues: unexpected _is_hf_initialized argument versus missing Float8 storage object." }, { "left": "issue:43377", "right": "issue:45491", "accept": false, - "reason": "Both are batching-related, but MIMI padding-mask mismatch and Gemma3 sliding-window NaNs are different code paths." + "reason": "Both involve batching/masking, but they are different models and distinct failure mechanisms." }, { "left": "issue:41720", "right": "issue:45081", "accept": false, - "reason": "CUDA assert in Qwen3 auto device mapping is unrelated to Mistral tokenizer regex patching." + "reason": "A CUDA assert during Qwen3 device mapping is unrelated to the Mistral tokenizer backend_tokenizer crash." }, { - "left": "issue:44265", - "right": "issue:45072", + "left": "issue:42175", + "right": "issue:43296", "accept": false, - "reason": "`torch.export`/`torch_compilable_check` failure and dtype mismatches are unrelated." + "reason": "Packaging/backend inclusion differs from a model-loading failure in vLLM." }, { - "left": "issue:42175", - "right": "issue:43296", + "left": "issue:44265", + "right": "issue:45072", "accept": false, - "reason": "Different package/model-loading regressions: missing TensorFlow backend vs PaddleOCR-VL load failure." + "reason": "torch.export/torch_compilable_check is unrelated to bf16 dtype mismatches in inference." }, { "left": "issue:43632", "right": "issue:45072", "accept": false, - "reason": "`_is_hf_initialized` breakage and bfloat16 dtype mismatch are separate regressions." + "reason": "Different Transformers v5 regressions with no shared concrete code path." }, { "left": "issue:43611", "right": "issue:43646", "accept": false, - "reason": "Both involve v5 model loading/init, but one is `base_model_prefix` loading and the other is custom initialization; not the same bug." + "reason": "Both are model-init/load regressions in v5, but one is about base_model_prefix handling and the other about custom initialization; not the same bug." }, { "left": "issue:43531", "right": "issue:44291", "accept": false, - "reason": "Sliding-window Qwen3-MoE behavior and `init_empty_weights` argument errors are unrelated." + "reason": "Sliding-window behavior in Qwen3-MoE is unrelated to init_empty_weights/_is_hf_initialized." }, { "left": "issue:44079", "right": "issue:44534", "accept": false, - "reason": "ModelOutput key assignment and non-persistent buffer junk are different internal state bugs." + "reason": "ModelOutput key assignment and non-persistent buffer junk are separate internal state bugs." }, { "left": "issue:43994", "right": "issue:44898", "accept": false, - "reason": "Different vision model failures: SigLIP2 AutoModel/pipeline output vs Perceiver non-default resolution classification." + "reason": "Different models and symptoms: SigLIP2 bad outputs versus Perceiver resolution failure." }, { "left": "issue:43454", "right": "issue:44466", "accept": false, - "reason": "Both mention lm_head/tied weights, but one is failed weight tying and the other is serialization inconsistency depending on device." + "reason": "Both mention lm_head, but one is failed weight tying causing bad generation and the other is serialization inconsistency depending on device." }, { "left": "issue:43881", "right": "issue:44488", "accept": false, - "reason": "Two separate model loading failures with no concrete shared mechanism." + "reason": "Unrelated loading failures for different models." }, { "left": "issue:43828", "right": "issue:44683", "accept": false, - "reason": "Autocast dtype mismatch in Phi-tiny-MoE is unrelated to compiled flex_attention on torch>=2.9." + "reason": "Autocast dtype mismatch is a different issue from compiled flex_attention failing on torch 2.9." }, { "left": "issue:42915", "right": "issue:43482", "accept": false, - "reason": "Qwen3Moe FineGrainedFP8Config and Qwen2.5-GGUF loading failure are different model/config issues." + "reason": "Qwen3MoE FP8 configuration failure is unrelated to GGUF loading under Transformers v5." }, { - "left": "issue:44079", - "right": "issue:45072", + "left": "issue:43792", + "right": "issue:43957", "accept": false, - "reason": "ModelOutput key handling and dtype mismatch are unrelated." + "reason": "Whisper loading failure and meta-device model loading breakage are different problems." }, { - "left": "issue:43792", - "right": "issue:43957", + "left": "issue:44079", + "right": "issue:45072", "accept": false, - "reason": "Whisper loading/runtime failure and meta-device loading breakage are different problems." + "reason": "ModelOutput key handling is unrelated to dtype mismatches in bf16 inference." }, { "left": "issue:43644", "right": "issue:44079", "accept": false, - "reason": "Non-persistent buffer initialization junk and ModelOutput key assignment are different subsystems." + "reason": "Both are v5 internals issues, but one is buffer serialization and the other is ModelOutput key bookkeeping." }, { "left": "issue:40444", "right": "issue:45381", "accept": false, - "reason": "Both are Qwen2.5-VL-related, but one is iterable-dataset multi-image batching and the other is video vision_position_ids." + "reason": "Both are Qwen2.5-VL issues, but one is multi-image iterable-dataset training and the other is video vision_position_ids ordering." }, { "left": "issue:45325", "right": "issue:45381", "accept": false, - "reason": "Same model family, but different concrete bugs: still-image temporal position scaling vs video vision_position_ids." + "reason": "Both concern Qwen2.5-VL position IDs, but one is still-image temporal scaling and the other is video input position IDs; not the same concrete bug." }, { "left": "issue:43296", "right": "issue:45800", "accept": false, - "reason": "PaddleOCR-VL loading under 4.53.0 and torch/transformers version incompatibility are not the same issue." + "reason": "Different compatibility issues: PaddleOCR-VL loading in vLLM versus torch 2.4.1/transformers 5.8.0 incompatibility." }, { "left": "issue:44387", "right": "issue:44617", "accept": false, - "reason": "Int4 quantization memory regression and Sam3Video OOM are distinct OOM reports with different causes." + "reason": "Both mention OOM, but one is an int4/Transformers v5 memory regression and the other is a model-specific CUDA OOM." } ] }, @@ -3380,225 +3335,237 @@ "issue:43404|issue:45357", "issue:44423|issue:45406", "issue:44186|issue:45072", - "issue:34634|issue:35707", "issue:43577|issue:43708", + "issue:34634|issue:35707", "issue:35707|issue:41720", "issue:43526|issue:43881" ], - "prepared_review_unit_hash": "90f1f1393dc0d43685e5d4acf0c6c45f8cd0c812ad80d5f4b5ff036d60b4d26e", + "prepared_review_unit_hash": "4f08226759877c8617e9c37a39e48627232ec9b9057bd2a4c17d39841a942d02", "component_budget": { "node_count": 387, "item_count": 387, "soft_pair_count": 1455, - "serialized_chars": 749552, - "estimated_input_tokens": 187388, - "estimated_eval_tokens": 375032 + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 33, - "serialized_chars": 37931, - "estimated_input_tokens": 9483, - "estimated_eval_tokens": 19222 + "serialized_chars": 37933, + "estimated_input_tokens": 9484, + "estimated_eval_tokens": 19224 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "No soft pair looks like a true duplicate: the items span unrelated bugs in audio, vision, tokenizer/config loading, offloading/quantization, generation, and serialization. I would reject all candidate merges.", - "confidence": 0.98, - "canonical_issue_reason": null, + "summary": "This cluster is not a true duplicate set; it mixes unrelated issues across offloading/loading, multimodal preprocessing, tokenizer/config regressions, and training/generation bugs. If one issue must anchor it, 43873 is the broadest offloading/quantization report, but the soft edges do not describe the same underlying bug.", + "confidence": 0.83, + "canonical_issue_reason": "Issue 43873 is the closest umbrella because it is the broadest offloading/quantization failure and overlaps the nearby device/offload-loading reports more than any other item; the rest are mostly narrower, model-specific, or unrelated bugs.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43873 is the best anchor if one must pick: it is open, has the most discussion/inbound references here, and is framed broadly around offloading with quantization.", + "best_issue_reason": "Issue 43873 is the best global representative only in a loose triage sense: it is the most general device/offloading problem in the set and is the only one that can plausibly subsume related offload/quantization failures, though it is not a duplicate of the others.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:35707", "right": "issue:40444", "accept": false, - "reason": "Different failures: progressive generation with inputs_embeds/past_key_values vs Qwen2.5-VL iterable dataset handling for multiple images." + "reason": "Different bugs: generation with inputs_embeds/past_key_values versus multimodal finetuning on multiple images per prompt." }, { "left": "issue:44291", "right": "issue:44877", "accept": false, - "reason": "Unrelated root causes: init_empty_weights/_is_hf_initialized argument handling vs strict config blocking granite_speech loading." + "reason": "Constructor/init_empty_weights incompatibility is unrelated to strict config loading for granite_speech." }, { "left": "issue:43526", "right": "issue:43883", "accept": false, - "reason": "Different code paths: BeitImageProcessorFast reduce_labels bug vs Molmo missing all_tied_weights_keys attribute." + "reason": "Image label reduction bug in a processor is unrelated to a model missing all_tied_weights_keys metadata." }, { "left": "issue:43606", "right": "issue:44265", "accept": false, - "reason": "One is a Bark CPU offload device-mismatch bug; the other is torch.export failing on torch_compilable_check. Not the same bug." + "reason": "CPU offload device mismatch is not the same code path as torch.export failures with torch_compilable_check." }, { "left": "issue:43317", "right": "issue:43873", "accept": false, - "reason": "Both mention offloading, but one is dequantized model loading with device_map=auto and the other is quantization/offloading behavior generally; not the same concrete failure." + "reason": "Both involve offloading/quantization, but one is a dequantized model load failure and the other is a broader offloading regression; not the same concrete bug." }, { "left": "issue:43881", "right": "issue:44991", "accept": false, - "reason": "Completely different: glm-4v-9b load failure vs tokenizer loading regression for EMBEDDIA/est-roberta." + "reason": "A model loading failure for glm-4v-9b is unrelated to a tokenizer-loading regression for EMBEDDIA/est-roberta." }, { "left": "issue:43701", "right": "issue:45325", "accept": false, - "reason": "Resume-from-checkpoint key mismatch is unrelated to Qwen2.5-VL rope index temporal position scaling." + "reason": "Resume-from-checkpoint key mismatch is unrelated to Qwen2.5-VL rope index temporal scaling." }, { "left": "issue:44186", "right": "issue:44265", "accept": false, - "reason": "Tokenizer padding/NER crash is unrelated to export failing on torch_compilable_check." + "reason": "Tokenizer crashes on NER/padding are unrelated to export-time failures in torch_compilable_check." }, { "left": "issue:43632", "right": "issue:45563", "accept": false, - "reason": "Transformers v5 _is_hf_initialized flag regression is unrelated to stale num_return_sequences warning in paged generate." + "reason": "A v5 HF initialization flag regression is unrelated to a stale warning in paged generate()." }, { "left": "issue:43606", "right": "issue:43819", "accept": false, - "reason": "Bark CPU offload device mismatch is not the same as DAC.from_latents missing STE behavior." + "reason": "Device mismatch during CPU offload is not the same as DAC.from_latents diverging from forward due to missing STE." }, { "left": "issue:43262", "right": "issue:43746", "accept": false, - "reason": "Audio apply_chat_template sampling-rate default bug is unrelated to GraniteSpeech PEFT checkpoint loading." + "reason": "Audio chat-template sampling-rate defaulting is unrelated to GraniteSpeech PEFT checkpoint loading." }, { "left": "issue:44610", "right": "issue:45325", "accept": false, - "reason": "OmDet-Turbo processor image-size mismatch is unrelated to Qwen2.5-VL temporal position id scaling." + "reason": "Processor image-size mismatch is unrelated to Qwen2.5-VL temporal position_id scaling." }, { "left": "issue:43454", "right": "issue:43550", "accept": false, - "reason": "lm_head weight tying bug in AyaVision is unrelated to Bamba torch.compile with SDPA." + "reason": "Tied lm_head weights causing bad text generation is unrelated to a torch.compile/SDPA failure." }, { "left": "issue:43377", "right": "issue:44610", "accept": false, - "reason": "MIMI batched-vs-single padding-mask bug is unrelated to OmDet-Turbo processor output size mismatch." + "reason": "Batch-vs-single encoder output mismatch from padding-mask handling is unrelated to the OmDet-Turbo processor image-size bug." }, { "left": "issue:43454", "right": "issue:45072", "accept": false, - "reason": "Weight tying/serialization issue in AyaVision is unrelated to dtype mismatches in SwitchTransformers/TimmWrapperModel." + "reason": "Weight-tying serialization bugs are not the same as dtype mismatches in bfloat16 inference." }, { "left": "issue:43377", "right": "issue:44898", "accept": false, - "reason": "MIMI padding-mask batching bug is unrelated to Perceiver non-default-resolution preprocessing failure." + "reason": "Missing padding-mask support in MIMI encoder is unrelated to Perceiver non-default resolution failures." }, { "left": "issue:40990", "right": "issue:45584", "accept": false, - "reason": "High perplexity on gpt-oss-20b is not the same as Whisper empty-transcription generation failure." + "reason": "High perplexity on gpt-oss-20b is unrelated to Whisper empty-transcription generation after token alignment." }, { "left": "issue:43572", "right": "issue:45759", "accept": false, - "reason": "StableLm pad_token_idx regression is unrelated to AutoModelForCausalLM.from_config not unwrapping text_config." + "reason": "A missing config field in StableLmConfig is unrelated to unwrapping text_config for composite Qwen multimodal configs." }, { "left": "issue:39692", "right": "issue:40444", "accept": false, - "reason": "SigLIP2 doc-example errors are unrelated to Qwen2.5-VL fine-tuning failures with multi-image iterable datasets." + "reason": "Broken docs/example mismatch for SigLIP2 is unrelated to Qwen2.5-VL finetuning with multiple images per prompt." }, { "left": "issue:44466", "right": "issue:45357", "accept": false, - "reason": "Both concern serialization, but one is lm_head.weight device-dependent behavior and the other is incorrect visual encoder keys in Qwen3.5 save_pretrained." + "reason": "Device-dependent lm_head serialization is not the same bug as Qwen3.5 saving incorrect visual encoder keys." }, { "left": "issue:43296", "right": "issue:44991", "accept": false, - "reason": "PaddleOCR-VL/vLLM model loading failure is unrelated to tokenizer loading for est-roberta." + "reason": "vLLM/PaddleOCR-VL model load failure is unrelated to tokenizer loading for EMBEDDIA/est-roberta." }, { "left": "issue:43454", "right": "issue:43720", "accept": false, - "reason": "AyaVision lm_head tying is unrelated to BitNet packed-weight unpacking during accelerate loading." + "reason": "lm_head weight tying is unrelated to BitNet packed-weight unpacking during accelerate loading." }, { "left": "issue:43792", "right": "issue:44683", "accept": false, - "reason": "Whisper model run failure is unrelated to compiled flex_attention breakage on torch >= 2.9." + "reason": "Whisper runtime failure is unrelated to compiled flex_attention breakage on torch >= 2.9." }, { "left": "issue:43873", "right": "issue:45701", "accept": false, - "reason": "Quantization/offloading behavior and tokenization changes are different issues with different failure modes." + "reason": "Quantized offloading problems are unrelated to a version-dependent tokenization change." }, { "left": "issue:41720", "right": "issue:44977", "accept": false, - "reason": "Qwen3 auto device-mapping cudaErrorAssert is unrelated to Qwen3.5 flash-attention generation failure." + "reason": "Qwen3 auto device mapping cuda asserts are unrelated to Qwen3.5 flash-attention generation failures." }, { "left": "issue:45632", "right": "issue:45684", "accept": false, - "reason": "Cache-path collision for trust_remote_code is unrelated to read-only permission propagation in save_pretrained/register_for_auto_class." + "reason": "Cache-path collisions for trust_remote_code are unrelated to read-only permission propagation in save_pretrained." }, { "left": "issue:43404", "right": "issue:45357", "accept": false, - "reason": "Mistral3 lm_head tying bug is unrelated to Qwen3.5 incorrect visual encoder keys on save_pretrained." + "reason": "Mistral3 lm_head tying is unrelated to Qwen3.5 visual encoder key serialization." }, { "left": "issue:44423", "right": "issue:45406", "accept": false, - "reason": "Continuous-batching multimodal crash with str.to is unrelated to Gemma4Processor missing _tokenizer." + "reason": "Continuous-batching multimodal crash from a string .to() call is unrelated to Gemma4Processor missing _tokenizer." }, { "left": "issue:44186", "right": "issue:45072", "accept": false, - "reason": "LayoutLMv2 tokenizer crash is unrelated to dtype mismatches in SwitchTransformers/TimmWrapperModel." + "reason": "Tokenizer padding/truncation crashes are unrelated to dtype mismatches in SwitchTransformers/TimmWrapperModel." + }, + { + "left": "issue:43577", + "right": "issue:43708", + "accept": false, + "reason": "dtype remaining float32 in Blip2 is unrelated to resume_from_checkpoint max_steps calculation." }, { "left": "issue:34634", "right": "issue:35707", "accept": false, - "reason": "BarkProcessor voice_preset failure is unrelated to progressive generation with inputs_embeds/past_key_values." + "reason": "Bark voice_preset not working is unrelated to progressive generation with inputs_embeds and past_key_values." }, { - "left": "issue:43577", - "right": "issue:43708", + "left": "issue:35707", + "right": "issue:41720", + "accept": false, + "reason": "Progressive generation issues are unrelated to Qwen3 auto device mapping cuda asserts." + }, + { + "left": "issue:43526", + "right": "issue:43881", "accept": false, - "reason": "Blip2 dtype remaining float32 is unrelated to resume_from_checkpoint max_steps miscalculation." + "reason": "BeitImageProcessorFast label reduction bug is unrelated to glm-4v-9b model loading failure." } ] }, @@ -3688,8 +3655,8 @@ "issue:42915|issue:43317", "issue:43550|issue:43606", "issue:40990|issue:45701", - "issue:41720|issue:43322", "issue:43295|issue:45701", + "issue:41720|issue:43322", "issue:41720|issue:45161", "issue:44671|issue:44743", "issue:43531|issue:43572", @@ -3703,249 +3670,219 @@ "issue:43421|issue:45701", "issue:30064|issue:34634", "issue:44492|issue:44509", - "issue:43388|issue:45059", "issue:43606|issue:43720", + "issue:43388|issue:45059", "issue:42915|issue:43322", "issue:43550|issue:43653" ], - "prepared_review_unit_hash": "8db125c2827bd55ea7f0b94690f921b657063edc8e43056cb0c126fa4867d9ad", + "prepared_review_unit_hash": "65990d0226dd1208b38b64868d7fd9382c178abaa883219793d9cf21ca844659", "component_budget": { "node_count": 387, "item_count": 387, "soft_pair_count": 1455, - "serialized_chars": 749552, - "estimated_input_tokens": 187388, - "estimated_eval_tokens": 375032 + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 35, - "serialized_chars": 37807, - "estimated_input_tokens": 9452, - "estimated_eval_tokens": 19160 + "serialized_chars": 37799, + "estimated_input_tokens": 9450, + "estimated_eval_tokens": 19156 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is a noisy mix of mostly unrelated issues; the soft-similarity edges are superficial and should not be merged as duplicates. The strongest anchor is the broad offloading/quantization regression in issue:43873.", + "summary": "These items are mostly unrelated issues from many different subsystems. None of the soft pairs look like true duplicates, so I rejected them all. If forced to anchor the cluster, the broadest loading/offloading-related report is the best fit, but there is no strong canonical match overall.", "confidence": 0.81, - "canonical_issue_reason": "issue:43873 is the broadest and most central loading/offloading regression here, with substantial discussion and inbound references; it\u2019s a better anchor than the more model-specific or narrower reports.", + "canonical_issue_reason": "No true canonical duplicate cluster is evident; issue 43873 is the broadest open report and the closest thing to an anchor because it covers the general offloading/quantization failure space.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43873 is the best representative issue in the set because it has the clearest cross-cutting runtime/loading impact and enough community activity to serve as a canonical tracker.", + "best_issue_reason": "Issue 43873 is the best global representative of the only recurring theme in the set (loading/offloading/memory behavior), even though most items are still distinct.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43873", "right": "issue:44756", "accept": false, - "reason": "Both involve loading/offloading, but one is a quantization/offload regression and the other is an mmap OOM workaround for Strix Halo; different bug and fix." + "reason": "Both mention offloading/memory, but one is quantization offload behavior and the other is mmap/OOM on Strix Halo; different root causes and code paths." }, { "left": "issue:43749", "right": "issue:44534", "accept": false, - "reason": "FSDP CPU RAM efficient loading is a specific loading regression; non-persistent buffers being junk is a separate state/buffer initialization problem." + "reason": "Both touch model loading/memory, but FSDP CPU RAM-efficient loading and non-persistent buffer corruption are different bugs." }, { "left": "issue:44877", "right": "issue:44991", "accept": false, - "reason": "Config loading for granite_speech and tokenizer loading for est-roberta are unrelated failures in different subsystems." + "reason": "Both are loading regressions after v5, but one is a config-loading strictness issue and the other is tokenizer loading for a specific model." }, { "left": "issue:42617", "right": "issue:42915", "accept": false, - "reason": "3d_parallel.py startup failure and Qwen3Moe FineGrainedFP8Config failure are different model/runtime issues." + "reason": "Different failures: running 3d_parallel.py versus Qwen3MoE with FineGrainedFP8Config; no shared code-path bug." }, { "left": "issue:43450", "right": "issue:45701", "accept": false, - "reason": "Batched video processor shape bugs and version-dependent tokenization changes are not the same underlying defect." + "reason": "Video batching shape bug versus version-dependent tokenization changes; unrelated behavior." }, { "left": "issue:43550", "right": "issue:43720", "accept": false, - "reason": "Torch.compile+SDPA failure in Bamba and packed-weight unpacking during accelerate loading in BitNet are separate code paths." + "reason": "Both are model/runtime failures, but torch.compile with SDPA and packed-weight unpacking during loading are distinct problems." }, { "left": "issue:36331", "right": "issue:42915", "accept": false, - "reason": "Unexpected compute_loss kwargs in a trainer override is unrelated to Qwen3Moe FP8 loading/runtime behavior." + "reason": "Trainer API signature mismatch versus Qwen3MoE FP8 loading failure; not the same bug." }, { "left": "issue:44610", "right": "issue:44898", "accept": false, - "reason": "Both are about image resolution, but they affect different models and different preprocessing/model-expectation mismatches." + "reason": "Both are image-size/preprocessing issues, but they affect different models and different failure mechanisms." }, { "left": "issue:44186", "right": "issue:44610", "accept": false, - "reason": "Tokenizer crashes on NER/padding vs an image processor output-size mismatch are different bugs." + "reason": "Tokenizer crash on NER padding/truncation is unrelated to OmDet-Turbo image-size mismatch." }, { "left": "issue:36331", "right": "issue:42617", "accept": false, - "reason": "Trainer API mismatch and 3d_parallel launch failure do not describe the same bug." + "reason": "Unexpected trainer keyword argument and 3d_parallel execution failure are unrelated." }, { "left": "issue:44387", "right": "issue:45704", "accept": false, - "reason": "Int4 quantization memory regression and apex FusedRMSNorm memory leak are distinct memory issues with different root causes." + "reason": "CUDA memory growth under int4 quantization is not the same as the apex RMSNorm memory leak." }, { "left": "issue:44479", "right": "issue:44918", "accept": false, - "reason": "Video-input regression in Qwen VL models and TRL embedding-unpacking failure in Qwen3.5 are separate problems." + "reason": "Video-input regression in Qwen VL models and Qwen3.5 embedding unpacking with TRL are different bugs." }, { "left": "issue:43316", "right": "issue:44913", "accept": false, - "reason": "A config API discrepancy and a rotary_pct reload regression are not the same defect." + "reason": "Both are config-related, but one is Gemma3TextConfig API mismatch and the other is GPTNeoX rotary_pct reload loss." }, { "left": "issue:42915", "right": "issue:43317", "accept": false, - "reason": "Qwen3Moe FP8 failure and dequantized model offload load failure are different runtime paths." + "reason": "Qwen3MoE FP8 failure and dequantized model offload loading failure are different code paths." }, { "left": "issue:43550", "right": "issue:43606", "accept": false, - "reason": "Bamba torch.compile/SDPA failure and Bark CPU-offload device mismatch are unrelated model/runtime bugs." + "reason": "Both involve model execution/loading, but Bamba torch.compile SDPA and Bark CPU offload device mismatch are not the same bug." }, { "left": "issue:40990", "right": "issue:45701", "accept": false, - "reason": "High perplexity on GPT-OSS and tokenization changes across versions are not the same concrete issue." + "reason": "Perplexity on gpt-oss-20b is a model/eval issue, not the tokenizer-version regression described in 45701." }, { - "left": "issue:41720", - "right": "issue:43322", + "left": "issue:43295", + "right": "issue:45701", "accept": false, - "reason": "CUDA assert on Qwen3 auto device mapping and Llava Next segfault on load are different failures." + "reason": "Processor.tokenizer regression and tokenization changes across versions are related only superficially; different concrete failures." }, { - "left": "issue:43295", - "right": "issue:45701", + "left": "issue:41720", + "right": "issue:43322", "accept": false, - "reason": "Processor.tokenizer regression and generic tokenization changes are too broad and not the same bug." + "reason": "Qwen3 auto device mapping CUDA assert and Llava Next segfault are separate model-loading failures." }, { "left": "issue:41720", "right": "issue:45161", "accept": false, - "reason": "These concern different GPT/Qwen model parallelism behaviors and do not share a concrete code-path failure." + "reason": "Both mention Qwen/generation scaling issues, but one is CUDA assert under device mapping and the other is tensor-parallel not working." }, { "left": "issue:44671", "right": "issue:44743", "accept": false, - "reason": "CamemBERT masked LM prediction regression and Qwen3.5 recurrent-state reset are unrelated model-specific bugs." + "reason": "CamemBERT masked-LM output regression and Qwen3.5 recurrent-state reset are unrelated model bugs." }, { "left": "issue:43531", "right": "issue:43572", "accept": false, - "reason": "Sliding-window behavior in Qwen3-MoE and missing pad_token_idx in StableLmConfig are separate config/runtime issues." + "reason": "Qwen3-MoE sliding-window behavior and StableLmConfig pad_token_idx loss are different issues." }, { "left": "issue:44292", "right": "issue:44479", "accept": false, - "reason": "Qwen-3-8B-NVFP4 runtime error and a video-input regression in Qwen VL models are different failures." + "reason": "Qwen-3-8B-NVFP4 runtime error and Qwen VL video-input regression are different failures." }, { "left": "issue:41720", - "right": "issue:42617", - "accept": false, - "reason": "Auto device mapping CUDA assert and 3d_parallel startup failure are unrelated." - }, - { - "left": "issue:36010", - "right": "issue:42617", - "accept": false, - "reason": "GenerationMixin import failure and 3d_parallel.py execution failure are not the same bug." - }, - { - "left": "issue:44186", - "right": "issue:44361", - "accept": false, - "reason": "Two tokenizer crashes, but on different tokenizers, tasks, and failure modes." - }, - { - "left": "issue:43540", - "right": "issue:43782", - "accept": false, - "reason": "Qwen3OmniMoe video-processing ValueError and Qwen3VL weight_only loading error are different code paths." - }, - { - "left": "issue:45406", - "right": "issue:45464", - "accept": false, - "reason": "Gemma4Processor missing _tokenizer in serve and Qwen3.5 streaming chat/completions failure are unrelated." - }, - { - "left": "issue:44387", - "right": "issue:45701", + "right": "issue:43873", "accept": false, - "reason": "CUDA memory growth under int4 quantization and tokenization behavior changes are not the same issue." + "reason": "Both involve Qwen/model loading, but the specific failures are device-mapping CUDA assert versus quantization/offload behavior." }, { "left": "issue:43421", "right": "issue:45701", "accept": false, - "reason": "Runtime post-processor updates for special tokens and generic version-driven tokenization changes are different bugs." + "reason": "Runtime post-processor update bug is not the same as version-dependent tokenization changes." }, { "left": "issue:30064", "right": "issue:34634", "accept": false, - "reason": "Void segmentation map processing and Bark voice_preset behavior are unrelated processor issues." + "reason": "Image segmentation-map processing and Bark voice_preset are unrelated processor bugs." }, { "left": "issue:44492", "right": "issue:44509", "accept": false, - "reason": "A typo in cache strategy text and stale docs mentioning removed pipelines are documentation issues, but not the same one." + "reason": "A typo in cache-strategy docs is unrelated to removed pipeline-task docs." }, { - "left": "issue:43388", - "right": "issue:45059", + "left": "issue:43606", + "right": "issue:43720", "accept": false, - "reason": "Metric gathering dropping tuple labels and SAM3 PCS text/bounding-box behavior are different problems." + "reason": "Both are loading/offload related, but the failing mechanism differs: CPU offload device mismatch versus packed weights not unpacked." }, { - "left": "issue:43606", - "right": "issue:43720", + "left": "issue:43388", + "right": "issue:45059", "accept": false, - "reason": "CPU-offload device mismatch in Bark and packed-weight handling in BitNet are separate accelerate/loading bugs." + "reason": "Metric-gathering label truncation and SAM3 text/bounding-box behavior are different code paths." }, { "left": "issue:42915", "right": "issue:43322", "accept": false, - "reason": "Qwen3Moe FineGrainedFP8Config failure and Llava Next load segfault do not share a concrete root cause." + "reason": "Qwen3MoE FP8 failure and Llava Next segmentation fault are distinct runtime problems." }, { "left": "issue:43550", "right": "issue:43653", "accept": false, - "reason": "Bamba SDPA/compile failure and BigBirdTokenizer special-token registration bug are unrelated." + "reason": "Torch.compile/SDPA failure and BigBirdTokenizer special-token registration bug are unrelated." } ] }, @@ -4030,8 +3967,8 @@ "issue:44451|issue:44991", "issue:44898|issue:45072", "issue:44062|issue:44451", - "issue:45588|issue:45759", "issue:43421|issue:43653", + "issue:45588|issue:45759", "issue:33666|issue:35707", "issue:44987|issue:45042", "issue:42886|issue:44843", @@ -4040,8 +3977,8 @@ "issue:43381|issue:43844", "issue:43421|issue:44186", "issue:44918|issue:44936", - "issue:44625|issue:45479", "issue:45081|issue:45356", + "issue:44625|issue:45479", "issue:43525|issue:43531", "issue:43782|issue:45381", "issue:43994|issue:44671", @@ -4051,226 +3988,214 @@ "issue:42915|issue:45759", "issue:43596|issue:44220" ], - "prepared_review_unit_hash": "b5b67fbdb8cf3761e310f7ec0e8f9877b8f20a46ebcdeac64a3fe098a4cb17a2", + "prepared_review_unit_hash": "8ace39b3185234ee1c92db22afa53139bf2e054aa423396589cf3f43cbeb4b73", "component_budget": { "node_count": 387, "item_count": 387, "soft_pair_count": 1455, - "serialized_chars": 749552, - "estimated_input_tokens": 187388, - "estimated_eval_tokens": 375032 + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 31, - "serialized_chars": 37260, - "estimated_input_tokens": 9315, - "estimated_eval_tokens": 18886 + "serialized_chars": 37263, + "estimated_input_tokens": 9316, + "estimated_eval_tokens": 18888 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This set is mostly a mix of unrelated issues. I found two true duplicate pairs: the Qwen2_5_VLProcessor batched padding=False crash, and the HF_HUB_OFFLINE tokenizer loading regression caused by _patch_mistral_regex/model_info().", - "confidence": 0.82, - "canonical_issue_reason": "issue:44843 is the clearest representative of the offline tokenizer regression because it names the concrete failing code path and root cause, not just the downstream symptom.", + "summary": "Mostly a heterogeneous cluster of unrelated bug reports. Two soft edges look like true duplicates: the Qwen2_5_VLProcessor batched `padding=False` crash, and the HF_HUB_OFFLINE tokenizer regression tied to `_patch_mistral_regex` / `model_info()`.", + "confidence": 0.84, + "canonical_issue_reason": "issue:44843 is the clearest root-cause report in the set: it pinpoints the offline failure to an unconditional `model_info()` call inside `_patch_mistral_regex`, which also explains the broader symptom in issue:42886.", "canonical_pr_reason": null, - "best_issue_reason": "issue:44843 is the strongest overall issue candidate among the provided items: it is specific, actionable, and directly captures the underlying bug rather than a generic symptom.", + "best_issue_reason": "Best anchor for the cluster because it states the concrete failing code path and environment trigger, making it the most actionable representative of the offline tokenizer bug.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44514", "right": "issue:44545", "accept": true, - "reason": "Same underlying bug: `Qwen2_5_VLProcessor.apply_chat_template` crashes on batched input when `padding=False`." + "reason": "Same Qwen2_5_VLProcessor `apply_chat_template` crash on batched input with `padding=False`; the titles describe the same failure." }, { "left": "issue:43531", "right": "issue:45759", "accept": false, - "reason": "Different problems: one is a Qwen3-MoE sliding-window issue, the other is an `AutoModelForCausalLM.from_config` text_config unwrap bug." + "reason": "Both are Qwen3-family bugs, but one is `sliding_window` handling in Qwen3-MoE and the other is `from_config`/`text_config` unwrapping for composite multimodal configs." }, { "left": "issue:43606", "right": "issue:43653", "accept": false, - "reason": "Unrelated subsystems: Bark CPU-offload device mismatch vs. BigBirdTokenizer special-token decode behavior." + "reason": "Different bugs: Bark CPU offload device mismatch vs BigBirdTokenizer special-token decode/registration issue." }, { "left": "issue:43421", "right": "issue:43927", "accept": false, - "reason": "Both are tokenizer/config bugs, but the concrete failures differ: runtime post-processor updates vs. save/load token ID loss causing generation IndexError." + "reason": "One is runtime post-processor updates for special tokens; the other is saving/loading custom token IDs in `DiaConfig`." }, { "left": "issue:34567", "right": "issue:35707", "accept": false, - "reason": "Different code paths: TrainerState token counting vs. progressive generation with `inputs_embeds` and `past_key_values`." + "reason": "Trainer state token counting and progressive generation with `inputs_embeds`/`past_key_values` are unrelated code paths." }, { "left": "issue:44368", "right": "issue:45357", "accept": false, - "reason": "Different Qwen3.5 issues: a tied-embeddings warning in LoRA training vs. incorrect visual encoder keys on save/load." + "reason": "A warning about `tie_word_embeddings` in Qwen3.5 fine-tuning is not the same bug as incorrect visual encoder keys being saved in `save_pretrained`." }, { "left": "issue:41720", "right": "issue:43746", "accept": false, - "reason": "Different models and failures: Qwen3 device-mapping CUDA assert vs. GraniteSpeech PEFT checkpoint loading." + "reason": "Different failures: Qwen3 auto device mapping on A800 vs GraniteSpeech PEFT adapter loading from local checkpoints." }, { "left": "issue:43873", "right": "issue:45704", "accept": false, - "reason": "Different underlying bugs: quantization/offloading behavior vs. T5 using apex RMSNorm with a memory leak." + "reason": "Quantization/offloading behavior and a T5 RMSNorm memory leak are separate issues." }, { "left": "issue:44451", "right": "issue:44991", "accept": false, - "reason": "Both are tokenizer-loading complaints, but they involve different models and likely different root causes." + "reason": "Both are tokenizer-loading regressions, but for different models and likely different root causes." }, { "left": "issue:44898", "right": "issue:45072", "accept": false, - "reason": "Different inference issues: Perceiver non-default-resolution image processing vs. dtype mismatch in SwitchTransformers/TimmWrapperModel." + "reason": "Perceiver resolution handling and bfloat16 dtype mismatches in SwitchTransformers/TimmWrapperModel are unrelated." }, { "left": "issue:44062", "right": "issue:44451", "accept": false, - "reason": "Different failures: AddedToken special-argument collision vs. a model-loading regression for ScandiBERT." + "reason": "`AddedToken(..., special=...)` argument duplication is a different tokenizer bug from loading `vesteinn/ScandiBERT`." }, { "left": "issue:45588", "right": "issue:45759", "accept": false, - "reason": "Different areas: flash_attention integration crash with `s_aux=None` vs. Qwen multimodal config unwrapping." - }, - { - "left": "issue:43421", - "right": "issue:43653", - "accept": false, - "reason": "Both tokenizer-related, but one is runtime post-processor updates and the other is mask-token registration/empty decode output." + "reason": "Flash-attention `s_aux=None` handling and Qwen `text_config` unwrapping are different code-path problems." }, { "left": "issue:33666", "right": "issue:35707", "accept": false, - "reason": "Different training/model features: Qwen2-VL multi-GPU training vs. TrainerState token accounting." + "reason": "Multi-GPU training for Qwen2-VL is unrelated to progressive generation with embeddings and cache reuse." }, { "left": "issue:44987", "right": "issue:45042", "accept": false, - "reason": "Different regression types: a model load failure for physical-intelligence/fast vs. PIL image processors incorrectly requiring torchvision." + "reason": "A model-loading failure for `physical-intelligence/fast` is not the same as PIL image processors incorrectly requiring torchvision." }, { "left": "issue:42886", "right": "issue:44843", "accept": true, - "reason": "Same offline-mode tokenizer regression; `44843` pinpoints the `_patch_mistral_regex`/`model_info()` call that breaks `HF_HUB_OFFLINE=1`." + "reason": "Same HF_HUB_OFFLINE tokenizer regression; 44843 identifies the concrete cause (`model_info()` called unconditionally in `_patch_mistral_regex`) behind 42886's cache/offline failure." }, { "left": "issue:43531", "right": "issue:44589", "accept": false, - "reason": "Different bugs: Qwen3-MoE sliding_window behavior vs. a Float8 storage lookup TypeError." + "reason": "`sliding_window` in Qwen3-MoE and missing `Float8_e4m3fnStorage` lookup are unrelated." }, { "left": "issue:39692", "right": "issue:44610", "accept": false, - "reason": "Both involve vision/model examples, but the specific failures are unrelated: SigLIP2 docs errors vs. OmDet-Turbo size mismatch." + "reason": "SigLIP2 doc-example errors and OmDet-Turbo processor/model resolution mismatch are different model-specific issues." }, { "left": "issue:43381", "right": "issue:43844", "accept": false, - "reason": "Different training/runtime issues: checkpointing in eval mode vs. gradient blow-up with HfDeepSpeedConfig + ZeRO-3." + "reason": "Gradient checkpointing being disallowed in eval mode is not the same as gradients exploding under ZeRO-3." }, { "left": "issue:43421", "right": "issue:44186", "accept": false, - "reason": "Different tokenizer failures: runtime special-token post-processing vs. LayoutLMv2 crashes on NER/batched padding inputs." + "reason": "Runtime special-token post-processor updates and LayoutLMv2Tokenizer NER/padding crashes are separate tokenizer problems." }, { "left": "issue:44918", "right": "issue:44936", "accept": false, - "reason": "Different trainer bugs: Qwen3.5 embedding unpacking in TRL SFT vs. `trainer.evaluate()` failing after `trainer.train()`." + "reason": "Qwen3.5 embedding unpacking with TRL SFT and `trainer.evaluate()` failing after `trainer.train()` are different trainer/model interactions." }, { - "left": "issue:44625", - "right": "issue:45479", + "left": "issue:45081", + "right": "issue:45356", "accept": false, - "reason": "Different classification/config problems: Qwen3.5 `num_labels` propagation vs. zero-loss behavior with `num_labels=1`." + "reason": "Both touch `fix_mistral_regex`, but one is a hard crash in `_patch_mistral_regex` while the other is a Kimi-K2.5 codec-handling regression with a misleading warning; not the same concrete bug." }, { - "left": "issue:45081", - "right": "issue:45356", + "left": "issue:44625", + "right": "issue:45479", "accept": false, - "reason": "Related to Mistral/Qwen tokenizer handling, but the concrete failures differ: backend_tokenizer attribute crash vs. Kimi-K2.5 codec-handling regression." + "reason": "`num_labels` propagation in Qwen3.5 configs and degenerate zero loss for `num_labels=1` are distinct classification bugs." }, { "left": "issue:43525", "right": "issue:43531", "accept": false, - "reason": "Different Qwen model-config issues: missing `pad_token_id` on Llama4Config vs. Qwen3-MoE sliding-window behavior." + "reason": "Missing `pad_token_id` on `Llama4Config` is unrelated to Qwen3-MoE `sliding_window` behavior." }, { "left": "issue:43782", "right": "issue:45381", "accept": false, - "reason": "Both are Qwen VL bugs, but one is a `from_pretrained(weight_only=True)` failure and the other is incorrect video `vision_position_ids`." + "reason": "Qwen3VL `weight_only=True` loading failure and Qwen2.5-VL video `vision_position_ids` are different multimodal bugs." }, { "left": "issue:43994", "right": "issue:44671", "accept": false, - "reason": "Different model families and symptoms: SigLIP2 pipeline output issues vs. CamemBERT masked LM predictions in v5." + "reason": "SigLIP2 nonsensical output with `AutoModel`/pipeline is not the same as CamemBERT masked-LM prediction regression." }, { "left": "issue:34634", "right": "issue:41720", "accept": false, - "reason": "Different domains: BarkProcessor voice preset handling vs. Qwen3 auto device mapping CUDA assert." + "reason": "BarkProcessor `voice_preset` failure and Qwen3 auto device-mapping cuda assert are unrelated." }, { "left": "issue:43381", "right": "issue:45704", "accept": false, - "reason": "No shared underlying bug: gradient-checkpointing eval-mode restriction vs. T5/apex RMSNorm memory leak." - }, - { - "left": "issue:39692", - "right": "issue:43873", - "accept": false, - "reason": "Different themes: documentation/example errors for SigLIP2 vs. quantization/offloading not working as expected." + "reason": "Eval-mode gradient checkpointing and a T5 apex RMSNorm memory leak are separate issues." }, { "left": "issue:42915", "right": "issue:45759", "accept": false, - "reason": "Different issues: Qwen3Moe FineGrainedFP8Config failure vs. Qwen multimodal config unwrapping." + "reason": "Qwen3Moe failing with `FineGrainedFP8Config` is unrelated to `AutoModelForCausalLM.from_config` not unwrapping `text_config`." }, { "left": "issue:43596", "right": "issue:44220", "accept": false, - "reason": "Different failures: deepspeed zero3 init on BertModel vs. `_torch_extract_fbank_features()` behavior." + "reason": "A Zero-3/BertModel indexing crash and `_torch_extract_fbank_features()` are not the same bug." } ] }, "evaluator_result": { "accept": true, - "feedback": "The analyst\u2019s conclusion is grounded in the packet and conservative. The two accepted pairs are strong duplicates: issue:44514/44545 are effectively the same Qwen2_5_VLProcessor padding=False crash, and issue:42886/44843 describe the same HF_HUB_OFFLINE tokenizer load regression. The rest of the pair judgments are appropriately cautious, and the summary does not overstate evidence." + "feedback": "The summary is grounded in the packet and stays conservative: it treats the cluster as mostly heterogeneous and only marks the two strongest title-aligned pairs as duplicates (`44514/44545` and `42886/44843`). The canonical anchor choice for `44843` is supported by the issue title, and the false soft-edge rejections do not overstate similarity." }, "evaluator_used": true, "retried": false, @@ -4345,12 +4270,12 @@ "issue:44220|issue:44661", "issue:43618|issue:43697", "issue:43720|issue:43819", - "issue:44077|issue:44625", "issue:43575|issue:45704", + "issue:44077|issue:44625", "issue:30064|issue:36010", "issue:43335|issue:43756", - "issue:43653|issue:43720", "issue:43334|issue:45759", + "issue:43653|issue:43720", "issue:43716|issue:44008", "issue:42915|issue:43746", "issue:43746|issue:45381", @@ -4370,214 +4295,208 @@ "issue:43782|issue:44977", "issue:44561|issue:45020" ], - "prepared_review_unit_hash": "abbfd4aaf19349a285137579743e5e6aeb3c64117b3bdb954806ede12dd19873", + "prepared_review_unit_hash": "af742e957b7802b1094e04e82734b3a97daa4b98caccb73116303e28cb2e9970", "component_budget": { "node_count": 387, "item_count": 387, "soft_pair_count": 1455, - "serialized_chars": 749552, - "estimated_input_tokens": 187388, - "estimated_eval_tokens": 375032 + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 }, "budget": { "node_count": 47, "item_count": 47, "soft_pair_count": 29, - "serialized_chars": 36144, - "estimated_input_tokens": 9036, - "estimated_eval_tokens": 18328 + "serialized_chars": 36139, + "estimated_input_tokens": 9035, + "estimated_eval_tokens": 18326 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly heterogeneous: the issues span unrelated bugs across tokenizers, multimodal configs, model loading, dtype handling, and inference paths. I do not see any soft pair that is safe to merge as a true duplicate on the available evidence.", - "confidence": 0.66, - "canonical_issue_reason": "issue:45020 is the broadest umbrella report in the set, covering recent-version regressions affecting remote_code models; it is the best representative issue even though the cluster overall is not a tight duplicate group.", + "summary": "This cluster is mostly a grab bag of unrelated Transformers bug reports. The closest thing to a center is the broad remote-code/config-loading regression theme, but most soft pairs are not true duplicates and should stay separate.", + "confidence": 0.36, + "canonical_issue_reason": "issue:45020 is the broadest and most central issue here: it captures the general class of recent-version regressions affecting remote_code models, which overlaps the strongest theme in the cluster.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45020 has the widest scope and most potential to subsume neighboring remote_code-loading regressions, so it is the most useful issue to keep as the representative anchor for this cluster.", + "best_issue_reason": "issue:45020 is the best representative issue because it is the most general umbrella report and can subsume adjacent config/remote-code loading regressions better than the narrower model-specific bugs.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45183", "right": "issue:45479", "accept": false, - "reason": "Different bugs: input/embedding validation errors versus a degenerate zero-loss classification bug. No shared code path is evident." + "reason": "Different failures: input validation/XOR error messaging vs degenerate zero-loss classification behavior." }, { "left": "issue:43540", "right": "issue:44008", "accept": false, - "reason": "Different models and failure modes: Qwen3OmniMoe video processing versus Gemma3n audio mask name collision." + "reason": "Different models and code paths: Qwen3OmniMoe video processing vs Gemma 3n audio tensor name collision." }, { "left": "issue:44220", "right": "issue:44661", "accept": false, - "reason": "Unrelated subsystems: speech feature extraction helper versus tokenizer-mapping model discovery logic." + "reason": "Unrelated tokenizer/audio feature bug vs model-addition failure in TOKENIZER_MAPPING_NAMES." }, { "left": "issue:43618", "right": "issue:43697", "accept": false, - "reason": "CLIP attention output regression and RTDetrV2 output drift are distinct model behaviors with different likely fixes." + "reason": "Different regressions: CLIPOutput attentions field omission vs RTDetrV2 output mismatch in v5." }, { "left": "issue:43720", "right": "issue:43819", "accept": false, - "reason": "BitNet packed-weight loading and DAC STE consistency are unrelated code paths." + "reason": "Different subsystems: BitNet packed-weight loading vs DAC latent decoding/STE mismatch." }, { - "left": "issue:44077", - "right": "issue:44625", + "left": "issue:43575", + "right": "issue:45704", "accept": false, - "reason": "PatchTSMixer post_init policy and Qwen3.5 num_labels propagation are different configuration bugs." + "reason": "Qwen2 tensor-parallel OOM is unrelated to T5's apex RMSNorm memory leak." }, { - "left": "issue:43575", - "right": "issue:45704", + "left": "issue:44077", + "right": "issue:44625", "accept": false, - "reason": "TP loading OOM for Qwen2 versus T5 using a leaking RMSNorm implementation are not the same bug." + "reason": "PatchTSMixer post_init API constraint and Qwen3.5 num_labels propagation are different bugs." }, { "left": "issue:30064", "right": "issue:36010", "accept": false, - "reason": "ImageProcessor segmentation-map handling and a GenerationMixin import error are unrelated." + "reason": "Image processor void segmentation handling is unrelated to the GenerationMixin import error." }, { "left": "issue:43335", "right": "issue:43756", "accept": false, - "reason": "Sparse-layer creation in SwitchTransformers and Smollm3 RoPE-layer dropping are different architectural/config issues." + "reason": "SwitchTransformers sparse-layer creation and Smollm3 RoPE layer dropping are unrelated architecture/config bugs." }, { - "left": "issue:43653", - "right": "issue:43720", + "left": "issue:43334", + "right": "issue:45759", "accept": false, - "reason": "BigBird special-token registration and BitNet accelerate-loading are unrelated." + "reason": "Both involve Qwen composite configs, but one is a missing pad_token_id at load time while the other is from_config unwrapping behavior; too different to merge." }, { - "left": "issue:43334", - "right": "issue:45759", + "left": "issue:43653", + "right": "issue:43720", "accept": false, - "reason": "Both touch Qwen multimodal config loading, but one is a Qwen3-VL pad_token_id attribute error and the other is AutoModelForCausalLM.from_config not unwrapping text_config; too different to treat as the same concrete bug." + "reason": "BigBird special-token registration and BitNet accelerate loading are unrelated." }, { "left": "issue:43716", "right": "issue:44008", "accept": false, - "reason": "Mistral-3 dtype mismatch and Gemma3n variable collision are separate problems." + "reason": "Mistral-3 dtype mismatch and Gemma 3n tensor collision are different model bugs." }, { "left": "issue:42915", "right": "issue:43746", "accept": false, - "reason": "FineGrainedFP8Config failure for Qwen3Moe and GraniteSpeech PEFT checkpoint loading are unrelated." + "reason": "Qwen3Moe FP8 config failure and GraniteSpeech PEFT checkpoint loading are unrelated." }, { "left": "issue:43746", "right": "issue:45381", "accept": false, - "reason": "Local PEFT adapter loading and qwen2.5-vl video position IDs are different bugs." + "reason": "PEFT local checkpoint loading and qwen2.5-vl video position IDs are different code paths." }, { "left": "issue:43746", "right": "issue:43873", "accept": false, - "reason": "PEFT checkpoint loading and quantization offloading behavior are distinct issues." + "reason": "Checkpoint loading with PEFT is unrelated to quantization/offloading behavior." }, { "left": "issue:43550", "right": "issue:45072", "accept": false, - "reason": "torch.compile+SDPA failure in Bamba and dtype mismatches in unrelated models are not the same underlying bug." + "reason": "torch.compile/SDPA failure in Bamba and dtype mismatch in SwitchTransformers/TimmWrapperModel are different issues." }, { "left": "issue:43550", "right": "issue:43819", "accept": false, - "reason": "Different model families and failure modes; no plausible shared fix." + "reason": "Different models and failures: compile-time SDPA bug vs DAC latent conversion mismatch." }, { "left": "issue:44186", "right": "issue:44898", "accept": false, - "reason": "Tokenizer NER padding/truncation crash and Perceiver image-resolution failure are unrelated." + "reason": "Tokenizer crash on NER padding/truncation is unrelated to Perceiver image-size handling." }, { "left": "issue:43421", "right": "issue:44811", "accept": false, - "reason": "Runtime post-processor updates and Whisper batch_decode skip_special_tokens behavior are different tokenizer/processor issues." + "reason": "Runtime special-token post-processor updates and Whisper batch_decode skip_special_tokens behavior are not the same bug." }, { "left": "issue:44843", "right": "issue:45081", "accept": false, - "reason": "Both involve _patch_mistral_regex, but one is an offline hub-info call and the other is a backend_tokenizer attribute crash; different concrete failure modes and fixes." + "reason": "Both touch _patch_mistral_regex, but one is an offline model_info call and the other is an attribute crash; related area, not the same concrete bug." }, { "left": "issue:44062", "right": "issue:44488", "accept": false, - "reason": "AddedToken special-argument duplication and failure to load a specific tokenizer/model are not clearly the same root cause." + "reason": "AddedToken argument duplication is unrelated to a model-not-loading report." }, { "left": "issue:43643", "right": "issue:44877", "accept": false, - "reason": "trust_remote_code missing-fields behavior and strict granite_speech config loading are related only at a very high level, not the same bug." + "reason": "Both are config-related, but one is missing fields from trust_remote_code and the other is strict GraniteSpeech config loading; not the same defect." }, { "left": "issue:43531", "right": "issue:43643", "accept": false, - "reason": "Sliding-window behavior in Qwen3-MoE is unrelated to AutoConfig missing fields under trust_remote_code." + "reason": "Qwen3-MoE sliding_window behavior and remote_code config field loss are unrelated." }, { "left": "issue:43577", "right": "issue:43818", "accept": false, - "reason": "Blip2 dtype propagation and Video-LLaVA video_tower/temporal-attention issues are different code paths." + "reason": "BLIP2 dtype propagation and Video-LLaVA missing temporal attention/shared weights are different model issues." }, { "left": "issue:43540", "right": "issue:45381", "accept": false, - "reason": "One is video input processing for Qwen3OmniMoe; the other is qwen2.5-vl video position IDs. Similar domain, but not the same concrete bug." + "reason": "Both involve video inputs, but they affect different Qwen model families and different failures." }, { "left": "issue:43653", "right": "issue:43819", "accept": false, - "reason": "Tokenizer special-token registration and DAC latent decoding mismatch are unrelated." + "reason": "Tokenizer special-token registration and DAC forward/from_latents mismatch are unrelated." }, { "left": "issue:44448", "right": "issue:44671", "accept": false, - "reason": "Pegasus v4/v5 output drift and CamemBERT masked-LM regression are separate model-specific regressions." + "reason": "Two separate v5 output-regression reports for different models; not one concrete bug." }, { "left": "issue:43782", "right": "issue:44977", "accept": false, - "reason": "Qwen3VL weight_only loading error and Qwen3.5 flash-attention generation issue are different failure modes." + "reason": "Qwen3VL weight_only loading error and Qwen3.5 flash-attention generation issue are different problems." }, { "left": "issue:44561", "right": "issue:45020", "accept": false, - "reason": "issue:45020 is a broad umbrella for remote_code regressions, while issue:44561 is one specific cause (removal of is_torch_fx_available); not the same concrete bug." - }, - { - "left": "issue:43531", - "right": "issue:43643", - "accept": false, - "reason": "No meaningful overlap beyond both being model/config-related." + "reason": "44561 is one specific remote_code regression cause; 45020 is a broad umbrella report, so they are related but not the same concrete bug." } ] }, @@ -4679,20 +4598,20 @@ "issue:43329|issue:45600", "issue:43644|issue:43749" ], - "prepared_review_unit_hash": "1eeb9e69fbd7682ffe1e467e4c448c46aed6171bedfd11415746756c3c7e8372", + "prepared_review_unit_hash": "12b4eac731f53f2a701f03b34234840bb152335a1eb8dc049b62ae6c1782bd59", "component_budget": { "node_count": 387, "item_count": 387, "soft_pair_count": 1455, - "serialized_chars": 749552, - "estimated_input_tokens": 187388, - "estimated_eval_tokens": 375032 + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 27, - "serialized_chars": 36027, + "serialized_chars": 36025, "estimated_input_tokens": 9007, "estimated_eval_tokens": 18270 }, @@ -4701,174 +4620,174 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a set of loosely related Transformers bugs and regressions, not true duplicates. The only near-theme overlap is around v5 pipeline removals, but even those issues target different symptoms (runtime messaging vs stale docs).", - "confidence": 0.33, - "canonical_issue_reason": "issue:43749 is the strongest single representative: it is a concrete, high-impact loading regression with the most follow-up activity/inbound references, whereas most other items are narrower, model-specific, or doc-only bugs.", + "summary": "The cluster is highly heterogeneous: most items are unrelated transformer/model regressions, tokenizer/load errors, or docs cleanup issues. None of the soft-edge pairs look like the same underlying bug/change, so all should be rejected.", + "confidence": 0.94, + "canonical_issue_reason": "issue:43749 is the strongest standalone representative in this set: it is a concrete, high-impact regression with substantial discussion and inbound references. It is not a duplicate of the other items, but it is the best anchor issue for the cluster.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43749 is the best global issue to anchor triage because it has broad impact and the clearest evidence of ongoing user pain; the rest of the set is too heterogeneous to serve as a duplicate canonical.", + "best_issue_reason": "issue:43749 is the most suitable global issue representative because it has the most signal (discussion, refs, and a clearly defined failure mode) among otherwise mixed topics.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43322", "right": "issue:45685", "accept": false, - "reason": "Different failures: Llava Next loading segfault vs an MPS histogram kernel type error in moe." + "reason": "Different failures: Llava Next model load segfault vs MPS histogram kernel not implemented for int." }, { "left": "issue:43825", "right": "issue:43994", "accept": false, - "reason": "One is a v5 error-message mismatch about removed pipeline tasks; the other is incorrect SigLIP2 outputs with AutoModel/pipeline." + "reason": "One is a v5 pipeline error-message regression; the other is nonsensical outputs for SigLIP2. Different bugs and code paths." }, { "left": "issue:44821", "right": "issue:45084", "accept": false, - "reason": "Unrelated areas: loading AutoImageProcessor from URL vs a Jinja/template compilation TypeError." + "reason": "AutoImageProcessor URL loading and template-node compilation are unrelated." }, { "left": "issue:30064", "right": "issue:36331", "accept": false, - "reason": "Different code paths and symptoms: void segmentation maps vs CustomTrainer.compute_loss signature breakage." + "reason": "Void segmentation map handling and CustomTrainer.compute_loss signature mismatch are separate issues." }, { "left": "issue:43825", "right": "issue:43827", "accept": false, - "reason": "Related v5 theme, but not the same bug: runtime error wording vs stale documentation references." + "reason": "Related v5 pipeline-removal theme, but one is runtime messaging and the other is docs cleanup; not the same bug." }, { "left": "issue:44485", "right": "issue:44509", "accept": false, - "reason": "GLM-5 RoPE implementation discussion is unrelated to stale pipeline-task docs." + "reason": "GLM-5 RoPE implementation discussion vs docs mentioning removed pipeline tasks. Different changes." }, { "left": "issue:43577", "right": "issue:43582", "accept": false, - "reason": "Different problems: Blip2 dtype handling vs Apple Silicon caching allocator warmup TypeError." + "reason": "BLIP2 dtype propagation and Apple Silicon warmup TypeError are unrelated." }, { "left": "issue:44186", "right": "issue:44442", "accept": false, - "reason": "Both are tokenizer issues, but one is LayoutLMv2 NER/padding crashes and the other is AutoTokenizer loading a specific tokenizer class." + "reason": "LayoutLMv2 tokenizer crashes and FastSpeech2ConformerTokenizer loading failure are different tokenizer bugs." }, { "left": "issue:43606", "right": "issue:45072", "accept": false, - "reason": "Both mention inference mismatches, but they affect different models and root causes (CPU offload device mismatch vs bfloat16 dtype mismatch)." + "reason": "CPU offload device mismatch for Bark-small vs dtype mismatch in SwitchTransformers/TimmWrapperModel." }, { "left": "issue:34634", "right": "issue:36010", "accept": false, - "reason": "Bark voice_preset malfunction is unrelated to the GenerationMixin import error." + "reason": "BarkProcessor voice_preset behavior and GenerationMixin import failure are unrelated." }, { "left": "issue:43452", "right": "issue:44991", "accept": false, - "reason": "Different tokenizer/model-loading regressions: gguf_file handling vs EMBEDDIA/est-roberta tokenizer loading on v5." + "reason": "GGUF loading breakage and tokenizer loading failure for est-roberta are different regressions." }, { "left": "issue:44805", "right": "issue:44936", "accept": false, - "reason": "Mask/tensor shape IndexError is not the same as trainer.evaluate() failing after train()." + "reason": "Mask-shape IndexError and trainer.evaluate() after train() failure are distinct problems." }, { "left": "issue:44610", "right": "issue:45072", "accept": false, - "reason": "Different issues: OmDet-Turbo preprocessing size mismatch vs dtype mismatches in inference." + "reason": "OmDet input-size mismatch and dtype mismatch in inference are not the same code-path bug." }, { "left": "issue:42175", "right": "issue:45081", "accept": false, - "reason": "Package/backend install issue is unrelated to the Mistral regex patch crash." + "reason": "Missing TensorFlow in torch install and Mistral tokenizer backend attribute error are unrelated." }, { "left": "issue:43650", "right": "issue:43906", "accept": false, - "reason": "'ADD THE DATA' is not the same as an isolated reproduction of another issue." + "reason": "The placeholder issue and the isolated reproduction report do not describe the same concrete failure." }, { "left": "issue:43720", "right": "issue:45072", "accept": false, - "reason": "BitNet packed-weight loading bug is unrelated to generic bfloat16 dtype mismatch behavior." + "reason": "BitNet packed-weights loading and SwitchTransformers/TimmWrapper dtype mismatch are different loading issues." }, { "left": "issue:45081", "right": "issue:45406", "accept": false, - "reason": "Different tokenizer failures: Mistral regex patch AttributeError vs Gemma4Processor missing _tokenizer in serve." + "reason": "Mistral tokenizer backend_tokenizer attribute error and Gemma4Processor missing _tokenizer are different attribute/access bugs." }, { "left": "issue:44671", "right": "issue:45701", "accept": false, - "reason": "CamemBERT MLM output regression is not the same as version-dependent tokenization changes." + "reason": "CamemBERT masked-LM regression and a broad tokenization-version report are not the same underlying bug." }, { "left": "issue:43404", "right": "issue:43782", "accept": false, - "reason": "Both involve model loading/weights, but they are different models and different failure modes." + "reason": "Mistral3 weight tying and Qwen3VL weight_only load error are unrelated model-specific issues." }, { "left": "issue:44060", "right": "issue:44368", "accept": false, - "reason": "Both are tied-weights warnings, but they concern different Qwen variants and different incorrect tie behavior." + "reason": "Qwen3-Next tied-weights warning bug and ms-swift LoRA warning about tie_word_embeddings are different problems." }, { "left": "issue:43582", "right": "issue:44857", "accept": false, - "reason": "AppleSilicon warmup TypeError and float16 AMP crash are separate runtime issues." + "reason": "Apple Silicon warmup TypeError and LwDetrImageLoss AMP crash are unrelated runtime failures." }, { "left": "issue:45459", "right": "issue:45803", "accept": false, - "reason": "Both are exception-handling bugs, but in different functions with different trigger conditions and fixes." + "reason": "Both involve bad exception handling, but they affect different functions and errors; not the same bug." }, { "left": "issue:43262", "right": "issue:43819", "accept": false, - "reason": "Audio chat-template sampling-rate default is unrelated to DAC.from_latents vs forward mismatch." + "reason": "Audio chat-template sampling-rate defaulting and DAC.from_latents STE mismatch are unrelated." }, { "left": "issue:43526", "right": "issue:43697", "accept": false, - "reason": "BeitImageProcessorFast label reduction bug does not match the RTDetrV2 output regression." + "reason": "BeitImageProcessorFast label reduction and RTDetrV2 v5 output differences are different model/processor regressions." }, { "left": "issue:45084", "right": "issue:45538", "accept": false, - "reason": "Template compilation TypeError and CLIPTokenizer model_max_length are unrelated." + "reason": "Template compilation TypeError and CLIPTokenizer max-length sentinel are unrelated." }, { "left": "issue:43329", "right": "issue:45600", "accept": false, - "reason": "Undefined video-branch variables in multimodal token counting is a different bug from removed auto_mappings configs." + "reason": "Undefined variables in multimodal token counting and stale auto_mappings references are separate issues." }, { "left": "issue:43644", "right": "issue:43749", "accept": false, - "reason": "Both are v5 loading regressions, but one is junk-filled non-persistent buffers and the other is FSDP CPU RAM-efficient loading; different code paths." + "reason": "Both touch loading, but one reports junk-filled non-persistent buffers and the other FSDP CPU RAM efficient loading; not the same concrete bug." } ] }, @@ -4946,10 +4865,10 @@ "issue:43746|issue:44811", "issue:44589|issue:45042", "issue:38175|issue:44610", - "issue:43854|issue:44977", "issue:43454|issue:43577", - "issue:43526|issue:44554", + "issue:43854|issue:44977", "issue:36010|issue:43645", + "issue:43526|issue:44554", "issue:34634|issue:36331", "issue:39692|issue:44912", "issue:44568|issue:45488", @@ -4971,208 +4890,208 @@ "issue:43697|issue:45357", "issue:43452|issue:43582" ], - "prepared_review_unit_hash": "16b6e786a0487ba25a8aeea1c842203e2f75b4b3ee673dba2dc45e05f3a165b6", + "prepared_review_unit_hash": "fdb33c136ddae4466fa1dd1f1a07adfa05963ab1dbbd453cd90b226d48c70f0f", "component_budget": { "node_count": 387, "item_count": 387, "soft_pair_count": 1455, - "serialized_chars": 749552, - "estimated_input_tokens": 187388, - "estimated_eval_tokens": 375032 + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 }, "budget": { "node_count": 47, "item_count": 47, "soft_pair_count": 29, - "serialized_chars": 36232, - "estimated_input_tokens": 9058, - "estimated_eval_tokens": 18372 + "serialized_chars": 36224, + "estimated_input_tokens": 9056, + "estimated_eval_tokens": 18368 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is highly heterogeneous: the soft pairs mostly share only broad subsystem terms (tokenizer, loading, video, vision, trainer) but not the same concrete bug or code path. I would not merge any of the soft edges.", - "confidence": 0.11, - "canonical_issue_reason": "Issue 45488 is the most concrete, actionable, and broadly impactful standalone bug report in the set, but it does not appear duplicate to the others.", + "summary": "The cluster is mostly heterogeneous; the soft edges share broad subsystems but not the same concrete bug/change, so none should be merged. If forced to pick an issue representative, the tokenizer regression in issue 45488 is the clearest actionable exemplar, but it does not unify the whole set.", + "confidence": 0.68, + "canonical_issue_reason": "Issue 45488 is the strongest single representative because it describes a concrete, user-visible tokenizer regression with broad impact; the other items are too mixed to serve as one duplicate anchor.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 45488 has a clear failure mode, impacted model family, and user-visible regression, making it the strongest representative issue here.", + "best_issue_reason": "Issue 45488 is the best overall issue candidate: specific enough to be actionable, broad enough to represent a real regression class, and open rather than already closed.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43653", "right": "issue:44912", "accept": false, - "reason": "Tokenizer special-token decoding and MXFP4 quantization fallback are unrelated bugs." + "reason": "Different bugs: tokenizer special-token registration vs quantization/load fallback." }, { "left": "issue:43638", "right": "issue:44936", "accept": false, - "reason": "Zero3 index error in training a Bert model is not the same as evaluate-after-train state breakage." + "reason": "Different training failures: ZeRO-3 empty-index crash vs post-train evaluate failure." }, { "left": "issue:43746", "right": "issue:44811", "accept": false, - "reason": "PEFT local checkpoint loading and Whisper batch_decode special-token handling are different code paths." + "reason": "Unrelated code paths: PEFT checkpoint loading vs Whisper batch decode flag handling." }, { "left": "issue:44589", "right": "issue:45042", "accept": false, - "reason": "Float8 storage lookup failure is unrelated to PIL backend torchvision requirements." + "reason": "Different runtime problems: missing Float8 storage vs PIL image processors requiring torchvision." }, { "left": "issue:38175", "right": "issue:44610", "accept": false, - "reason": "SigLIP2 zero-probability output and OmDet-Turbo image-size mismatch are distinct model/processor issues." - }, - { - "left": "issue:43854", - "right": "issue:44977", - "accept": false, - "reason": "GLM-4.7 loading failure and Qwen3.5 flash-attention generation bug are different model-specific regressions." + "reason": "Different models and failures: zero probabilities vs processor/model input-size mismatch." }, { "left": "issue:43454", "right": "issue:43577", "accept": false, - "reason": "Weight tying failure in AyaVision and Blip2 dtype not propagating are separate loading/generation problems." + "reason": "Different multimodal load bugs: lm_head tying vs dtype propagation to Blip2/QFormer." }, { - "left": "issue:43526", - "right": "issue:44554", + "left": "issue:43854", + "right": "issue:44977", "accept": false, - "reason": "BeitImageProcessor label reduction and MPS attention correctness are unrelated." + "reason": "Different model-specific regressions and code paths; not the same generation bug." }, { "left": "issue:36010", "right": "issue:43645", "accept": false, - "reason": "Missing GenerationMixin import and Jupyter custom-model initialization failure are not the same bug." + "reason": "Import-path regression vs Jupyter custom-model initialization issue." + }, + { + "left": "issue:43526", + "right": "issue:44554", + "accept": false, + "reason": "Image label reduction bug is unrelated to attention correctness on MPS." }, { "left": "issue:34634", "right": "issue:36331", "accept": false, - "reason": "BarkProcessor voice_preset and CustomTrainer.compute_loss kwargs are unrelated trainer/processor issues." + "reason": "Different APIs entirely: BarkProcessor voice preset vs CustomTrainer compute_loss signature." }, { "left": "issue:39692", "right": "issue:44912", "accept": false, - "reason": "SigLIP2 doc example problems do not match the MXFP4 quantization loading regression." + "reason": "Docs/example errors and quantization failure are not the same underlying bug." }, { "left": "issue:44568", "right": "issue:45488", "accept": false, - "reason": "mDeBERTa special-token insertion and LlamaTokenizer ByteLevel/Metaspace override are different tokenizer regressions." + "reason": "Both are tokenizer regressions, but they affect different tokenizers and different code paths." }, { "left": "issue:43819", "right": "issue:43927", "accept": false, - "reason": "DAC latent mismatch is unrelated to DiaConfig losing custom token IDs on save/load." + "reason": "Different bugs: DAC forward/from_latents mismatch vs DiaConfig save/load token-ID loss." }, { "left": "issue:43257", "right": "issue:43329", "accept": false, - "reason": "Qwen3 MoE accelerate/deepspeed weight conversion and multimodal token counting are different failures." + "reason": "Qwen3 MoE weight conversion during loading is unrelated to multimodal video token counting." }, { "left": "issue:43720", "right": "issue:44610", "accept": false, - "reason": "BitNet packed-weight unpacking and OmDet processor image resolution mismatch are unrelated." + "reason": "Packed-weight unpacking during accelerate loading is unrelated to OmDet processor image sizing." }, { "left": "issue:43749", "right": "issue:44291", "accept": false, - "reason": "FSDP CPU RAM efficient loading and init_empty_weights _is_hf_initialized argument handling are distinct loading issues." + "reason": "Different model-loading failures: FSDP CPU RAM efficient loading vs unexpected init arg in init_empty_weights." }, { "left": "issue:43550", "right": "issue:44912", "accept": false, - "reason": "Bamba SDPA torch.compile failure is unrelated to MXFP4 quantization fallback." + "reason": "torch.compile/SDPA bug is unrelated to MXFP4 quantization fallback." }, { "left": "issue:43262", "right": "issue:44811", "accept": false, - "reason": "Audio apply_chat_template sample-rate defaulting and Whisper batch_decode behavior are different processor bugs." + "reason": "Audio chat-template sampling-rate default is a different processor bug than Whisper decode flags." }, { "left": "issue:44568", "right": "issue:45356", "accept": false, - "reason": "These are both tokenizer regressions, but the failure modes differ: BOS/EOS omission versus codec/fix_mistral_regex handling." + "reason": "Different tokenizer regressions affecting different models and symptoms." }, { "left": "issue:43531", "right": "issue:45042", "accept": false, - "reason": "Qwen3-MoE sliding_window behavior and PIL torchvision dependency are unrelated." + "reason": "Sliding-window model behavior is unrelated to the PIL backend torchvision dependency issue." }, { "left": "issue:43643", "right": "issue:45759", "accept": false, - "reason": "trust_remote_code config field loss and composite config text_config unwrapping are different config-loading bugs." + "reason": "Missing fields from trust_remote_code config loading is not the same as unwrapping text_config in from_config." }, { "left": "issue:36010", "right": "issue:36331", "accept": false, - "reason": "Importing GenerationMixin and CustomTrainer.compute_loss kwargs are not the same code-path problem." + "reason": "Import failure from generation module is unrelated to a custom trainer loss signature change." }, { "left": "issue:43329", "right": "issue:43450", "accept": false, - "reason": "Undefined multimodal token helpers and batched video processor shape errors are separate video-path bugs." + "reason": "Both concern video handling, but one is undefined helpers in token counting while the other is batched output shape; not the same bug." }, { "left": "issue:34567", "right": "issue:40990", "accept": false, - "reason": "TrainerState token counter not updating is unrelated to high perplexity on gpt-oss-20b." + "reason": "TrainerState token accounting bug and high perplexity on gpt-oss are different problems." }, { "left": "issue:43329", "right": "issue:44479", "accept": false, - "reason": "Both mention video, but one is undefined helper usage in token counting and the other is a broader generation regression across models." + "reason": "Related video area, but one is an undefined-helper token-count bug and the other is a broader v5.3 video-input regression." }, { "left": "issue:44849", "right": "issue:45701", "accept": false, - "reason": "Qwen3.5 hidden-states output bug and general tokenization-version changes are not the same underlying issue." + "reason": "Qwen3.5 hidden-state behavior and generic tokenization-version changes are not the same concrete issue." }, { "left": "issue:43257", "right": "issue:43931", "accept": false, - "reason": "MoE weight conversion/loading and Qwen3-VL shape mismatch are unrelated model-loading failures." + "reason": "Different load failures: MoE weight conversion with accelerate/deepspeed vs Qwen3-VL shape mismatch." }, { "left": "issue:43697", "right": "issue:45357", "accept": false, - "reason": "RTDetrV2 output drift in v5 and Qwen3.5 visual-encoder save_pretrained keys are different regressions." + "reason": "Different regressions: RTDetrV2 output divergence vs Qwen3.5 save_pretrained visual-key bug." }, { "left": "issue:43452", "right": "issue:43582", "accept": false, - "reason": "gguf_file tokenizer/model loading and Apple Silicon warmup TypeError are unrelated." + "reason": "Tokenizer/model loading with gguf_file is unrelated to Apple Silicon warmup TypeError." } ] }, @@ -5249,15 +5168,15 @@ "issue:44291|issue:45042", "issue:42915|issue:43742", "issue:33357|issue:34567", - "issue:44479|issue:45381", "issue:43317|issue:45538", + "issue:44479|issue:45381", "issue:44987|issue:45538", "issue:43540|issue:44560", "issue:42915|issue:44821", "issue:43540|issue:44479", "issue:43927|issue:44912", - "issue:43257|issue:43756", "issue:43577|issue:45072", + "issue:43257|issue:43756", "issue:42915|issue:45538", "issue:43257|issue:43696", "issue:45137|issue:45405", @@ -5278,232 +5197,232 @@ "issue:43526|issue:45479", "issue:44291|issue:44821" ], - "prepared_review_unit_hash": "61a637f49b029e3774b9e579a6a7c9afd9f24af7228bee6e136c10ab835e2fbd", + "prepared_review_unit_hash": "3386c60f1e048385fabff1ba9500f086983d44931a2399d5e755fc2cbb189e47", "component_budget": { "node_count": 387, "item_count": 387, "soft_pair_count": 1455, - "serialized_chars": 749552, - "estimated_input_tokens": 187388, - "estimated_eval_tokens": 375032 + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 }, "budget": { "node_count": 47, "item_count": 47, "soft_pair_count": 32, - "serialized_chars": 36861, - "estimated_input_tokens": 9216, - "estimated_eval_tokens": 18688 + "serialized_chars": 36869, + "estimated_input_tokens": 9218, + "estimated_eval_tokens": 18692 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This set is mostly unrelated issues. Only two soft pairs look like true duplicates: the Qwen video-input regression pair and the SigLIP2 behavior discrepancy pair. I\u2019d use the broad Qwen video regression as the representative issue.", - "confidence": 0.69, - "canonical_issue_reason": "Issue 44479 is the broadest, most representative bug in the set: a v5.3.0 video-input regression affecting multiple Qwen variants, with a clear duplicate-style companion in 45381.", + "summary": "This is a noisy mixed cluster, but two duplicate-like pockets stand out: Qwen video-input regressions and a SigLIP2 output mismatch. Most other pairs are false positives across unrelated bugs.", + "confidence": 0.74, + "canonical_issue_reason": "issue:44479 is the broadest concrete regression in the cluster: it describes the shared Qwen video-input failure across multiple related model families, and it has the clearest duplicate-style overlap with issue:45381.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 44479 is the best overall issue because it describes a concrete, cross-model regression with strong user impact and likely duplicate traction; it is more central than the narrower, model-specific reports here.", + "best_issue_reason": "issue:44479 is the best representative issue because it is a broad, user-facing regression with multiple related variants and strong duplicate potential.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:41720", "right": "issue:43540", "accept": false, - "reason": "Both are Qwen-family failures, but one is auto device mapping/cuda assert and the other is video-input parsing in Qwen3OmniMoe; different code paths and symptoms." + "reason": "Different Qwen failures: auto device-mapping CUDA assert vs. video-input ValueError in a different model path." }, { "left": "issue:44291", "right": "issue:45042", "accept": false, - "reason": "Completely different areas: init_empty_weights/_is_hf_initialized vs PIL image processor torchvision dependency." + "reason": "Unrelated problems: init_empty_weights argument handling vs. PIL backend image-processor dependency behavior." }, { "left": "issue:42915", "right": "issue:43742", "accept": false, - "reason": "Different model families and failure modes: Qwen3Moe FP8 loading vs MobileLLM key error." + "reason": "Different model/loading failures: FineGrainedFP8Config on Qwen3Moe vs. key error loading MobileLLM-125M." }, { "left": "issue:33357", "right": "issue:34567", "accept": false, - "reason": "Unrelated bugs: Mac bus error with CLIP vs TrainerState token counting not updating." - }, - { - "left": "issue:44479", - "right": "issue:45381", - "accept": true, - "reason": "Both report the same underlying 5.3.0 Qwen video-input regression, specifically incorrect video-position handling for qwen2.5-vl." + "reason": "Completely different subsystems: MacOS CLIP bus error vs. TrainerState token accounting." }, { "left": "issue:43317", "right": "issue:45538", "accept": false, - "reason": "Device-map/offload loading failure is unrelated to CLIPTokenizer model_max_length behavior." + "reason": "Different code paths: dequantized device-map/offload loading bug vs. CLIPTokenizer max-length constant." + }, + { + "left": "issue:44479", + "right": "issue:45381", + "accept": true, + "reason": "Same underlying Qwen video-input regression: both point to the wrong video position-id handling in 5.3.0, with 44479 being the broader report." }, { "left": "issue:44987", "right": "issue:45538", "accept": false, - "reason": "Different loading problems: a specific model load regression vs tokenizer max_length handling." + "reason": "Different failures: loading a specific model on transformers>=5.1.0 vs. tokenizer max-length behavior." }, { "left": "issue:43540", "right": "issue:44560", "accept": false, - "reason": "Both involve Qwen video inputs, but the affected models and concrete errors differ enough to treat as separate bugs." + "reason": "Both are video-related, but the model families and failure modes differ; not clearly the same bug." }, { "left": "issue:42915", "right": "issue:44821", "accept": false, - "reason": "No common underlying bug; Qwen3Moe FP8 config vs AutoImageProcessor URL loading." + "reason": "Unrelated model quantization/loading issue vs. AutoImageProcessor URL loading problem." }, { "left": "issue:43540", "right": "issue:44479", "accept": false, - "reason": "Both are video-related, but they affect different Qwen model families and are not clearly the same code-path failure." + "reason": "Both involve video inputs, but one is a Qwen3OmniMoe processing ValueError and the other is a broader Qwen video regression; not the same concrete bug." }, { "left": "issue:43927", "right": "issue:44912", "accept": false, - "reason": "DiaConfig token-ID persistence and MXFP4 loading/fallback are unrelated." + "reason": "Different features: DiaConfig save/load token IDs vs. MXFP4 quantization loading fallback." }, { - "left": "issue:43257", - "right": "issue:43756", + "left": "issue:43577", + "right": "issue:45072", "accept": false, - "reason": "Different model-specific issues: Qwen3 MOE loading with accelerate/deepspeed vs Smollm3 RoPE layer dropping." + "reason": "Both mention dtype, but the affected models and failure mechanisms are different." }, { - "left": "issue:43577", - "right": "issue:45072", + "left": "issue:43257", + "right": "issue:43756", "accept": false, - "reason": "Both mention dtype, but they concern different architectures and distinct dtype-mismatch bugs." + "reason": "Different model-specific issues: Qwen3 MoE weight conversion vs. Smollm3 RoPE-layer dropping." }, { "left": "issue:42915", "right": "issue:45538", "accept": false, - "reason": "No overlap beyond being model-loading/configuration issues." + "reason": "Different areas entirely: Qwen3Moe FP8 loading vs. CLIPTokenizer max-length default." }, { "left": "issue:43257", "right": "issue:43696", "accept": false, - "reason": "Qwen3 MOE conversion/loading bug is unrelated to GPT-oss out-of-memory." + "reason": "Different failure modes and models: MoE weight conversion vs. GPT-oss CUDA OOM." }, { "left": "issue:45137", "right": "issue:45405", "accept": false, - "reason": "ZeRO3 deque error and PEFT version bump are unrelated." + "reason": "No relation: DeepSpeed ZeRO3 deque error vs. PEFT version pinning." }, { "left": "issue:38175", "right": "issue:44898", "accept": false, - "reason": "Both are vision-model correctness issues, but different models and different failure mechanisms." + "reason": "Different models and symptoms: SigLIP2 zero probabilities vs. Perceiver resolution/interpolation failure." }, { "left": "issue:44737", "right": "issue:45588", "accept": false, - "reason": "XLNet CPU device placement bug is unrelated to flash-attention s_aux=None crash." + "reason": "Unrelated implementation bugs: XLNet CPU arange placement vs. flash_attention auxiliary-state crash." }, { "left": "issue:44062", "right": "issue:44991", "accept": false, - "reason": "Both are tokenizer-related, but one is AddedToken argument duplication and the other is a model-specific tokenizer load failure." + "reason": "Both tokenizer-related, but one is AddedToken argument duplication and the other is a model-specific tokenizer loading regression." }, { "left": "issue:43742", "right": "issue:44291", "accept": false, - "reason": "Different subsystems and symptoms: model key lookup vs init_empty_weights argument incompatibility." + "reason": "Different layers: MobileLLM key lookup vs. init_empty_weights/_is_hf_initialized argument handling." }, { "left": "issue:43906", "right": "issue:45381", "accept": false, - "reason": "A reproduction of a different issue is not the same bug as the Qwen2.5-VL video regression." + "reason": "The reproduction issue is too generic and does not match the specific Qwen2.5-VL video-position-id regression tightly enough." }, { "left": "issue:43646", "right": "issue:43749", "accept": false, - "reason": "Custom model init breakage and FSDP CPU RAM efficient loading are separate release regressions." + "reason": "Different core problems: custom model init breakage in v5.0.0 vs. FSDP CPU RAM-efficient loading regression." }, { "left": "issue:43749", "right": "issue:44756", "accept": false, - "reason": "Both relate to memory/loading, but one is FSDP CPU-efficient loading and the other is mmap/OOM on Strix Halo; not the same bug." + "reason": "Both involve loading/memory, but one is FSDP RAM-efficient loading and the other is mmap on Strix Halo; not the same bug." }, { "left": "issue:34567", "right": "issue:36331", "accept": false, - "reason": "TrainerState token accounting and custom compute_loss signature mismatch are different trainer API issues." + "reason": "Training-state token counting vs. custom trainer compute_loss signature mismatch are unrelated." }, { "left": "issue:44062", "right": "issue:45081", "accept": false, - "reason": "Tokenizer construction error and Mistral regex patch crash are distinct failures." + "reason": "Both touch tokenizer internals, but one is AddedToken construction and the other is Mistral regex patching; different code paths." }, { "left": "issue:43643", "right": "issue:45375", "accept": false, - "reason": "Both involve missing config fields, but one is generic trust_remote_code AutoConfig behavior and the other is a specific strict-field drop in Qwen3_5MoeVisionConfig." + "reason": "Similar theme of config-field handling, but they are different bugs and different model/config types." }, { "left": "issue:36010", "right": "issue:42175", "accept": false, - "reason": "GenerationMixin import regression and TensorFlow-backend packaging issue are unrelated." + "reason": "Completely unrelated: GenerationMixin import path vs. TensorFlow not included in the torch extra." }, { "left": "issue:43493", "right": "issue:43994", "accept": true, - "reason": "Both describe the same SigLIP2 model behaving incorrectly in HF compared with expected/original behavior, with matching model family and symptom class." + "reason": "Same SigLIP2 checkpoint family and same symptom class: incorrect model behavior/output compared to the expected implementation, plausibly fixed by one SigLIP2 code-path correction." }, { "left": "issue:43452", "right": "issue:44488", "accept": false, - "reason": "GGUF loading breaks in general and a specific model failing to load are not clearly the same underlying bug." + "reason": "Different loading problems: GGUF tokenizer/model loading vs. inability to load a specific bert model." }, { "left": "issue:42175", "right": "issue:43824", "accept": false, - "reason": "Packaging/backend selection issue vs missing Qwen2_5_VL export import error." + "reason": "Unrelated packaging/back-end inclusion issue vs. a specific model import error." }, { "left": "issue:43526", "right": "issue:45479", "accept": false, - "reason": "Label-reduction bug in BeitImageProcessorFast is unrelated to degenerate zero-loss classification behavior." + "reason": "Different image-processing and classification bugs: BEiT label reduction vs. sequence-classification zero-loss behavior." }, { "left": "issue:44291", "right": "issue:44821", "accept": false, - "reason": "init_empty_weights/_is_hf_initialized and AutoImageProcessor-from-URL are different code paths with no shared concrete failure." + "reason": "No overlap: init_empty_weights argument handling vs. AutoImageProcessor URL loading." } ] }, "evaluator_result": { "accept": true, - "feedback": "The issue selection and pair judgments are conservative and grounded in the packet. The canonical choice of 44479 is well supported by the titles, and the two accepted soft pairs (44479/45381 and 43493/43994) are plausible duplicates without overclaiming. No clear overstatement or mergeability problem is present." + "feedback": "The summary is grounded in the packet: the Qwen video-input issues and the SigLIP2 model-output issues are the only plausible duplicate pockets, and the other pairwise rejections are conservative. I don\u2019t see any overstatement that would warrant rejection." }, "evaluator_used": true, "retried": false, @@ -5576,13 +5495,13 @@ "issue:44792|issue:44964", "issue:28282|issue:34689", "issue:45042|issue:45081", - "issue:43756|issue:45325", "issue:45405|issue:45600", + "issue:43756|issue:45325", "issue:43742|issue:44589", "issue:44484|issue:44704", + "issue:43531|issue:44821", "issue:28282|issue:29127", "issue:43697|issue:45216", - "issue:43531|issue:44821", "issue:44514|issue:45381", "issue:43643|issue:44561", "issue:44625|issue:44877", @@ -5600,190 +5519,190 @@ "issue:24643|issue:29127", "issue:38175|issue:43295" ], - "prepared_review_unit_hash": "0306d7771edab73e928e225dd762e40c73688ee9f4de28cbad6c7680ff76223e", + "prepared_review_unit_hash": "060ab9fc870c1c9512a455a216de5f197431448956c674e4b101bbaa5b1ba504", "component_budget": { "node_count": 387, "item_count": 387, "soft_pair_count": 1455, - "serialized_chars": 749552, - "estimated_input_tokens": 187388, - "estimated_eval_tokens": 375032 + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 }, "budget": { "node_count": 47, "item_count": 47, "soft_pair_count": 26, - "serialized_chars": 35268, - "estimated_input_tokens": 8817, - "estimated_eval_tokens": 17890 + "serialized_chars": 35269, + "estimated_input_tokens": 8818, + "estimated_eval_tokens": 17892 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is heterogeneous: the soft-similarity pairs mostly share a broad subsystem theme (tokenizers, loading, vision, tied weights, configs) but not the same concrete bug or mergeable code-path. I rejected all soft edges as likely false positives.", - "confidence": 0.93, - "canonical_issue_reason": "No single canonical issue cleanly represents the cluster because the items span unrelated bugs; if a placeholder anchor is needed, issue 24643 is the most discussed and self-contained report, but it does not subsume the others.", - "canonical_pr_reason": "No pull requests are present in the cluster.", - "best_issue_reason": "Issue 24643 is the strongest general anchor by discussion activity and a clearly reproducible failure, but it is not a duplicate target for the rest of the set.", - "best_pr_reason": "No pull requests are present in the cluster.", + "summary": "These are mostly superficially similar bug reports across loader, tokenizer, processor, and config paths, but each pair has a distinct trigger or code path. No soft edges should be merged as duplicates.", + "confidence": 0.78, + "canonical_issue_reason": "issue:44704 is the best representative of the batch because it is a clear, cross-cutting `from_pretrained`/loader-plumbing regression with broader API impact than the model-specific reports.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:44704 is the strongest issue candidate overall: it has a precise failure mode, clear repro scope, and broader relevance than the narrowly model-specific symptom reports.", + "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44792", "right": "issue:44964", "accept": false, - "reason": "Both are model-loading failures, but for different models and likely different root causes; not the same bug." + "reason": "Both are model-loading failures, but they affect different models and likely different root causes." }, { "left": "issue:28282", "right": "issue:34689", "accept": false, - "reason": "One is missing PyTorch in the environment, the other is a specific Llama 3.2 Vision loading regression." + "reason": "One is a missing-PyTorch import error; the other is a Llama vision model-loading regression." }, { "left": "issue:45042", "right": "issue:45081", "accept": false, - "reason": "Both involve tokenizer/image-processing internals, but one is a torchvision dependency regression and the other is a Mistral regex crash." + "reason": "These involve tokenizer/image-processor loading, but the reported failures are different: torchvision dependency vs Mistral regex crash." }, { - "left": "issue:43756", - "right": "issue:45325", + "left": "issue:45405", + "right": "issue:45600", "accept": false, - "reason": "Both touch Qwen/vision positional handling, but the affected models and failure modes differ materially." + "reason": "These are unrelated maintenance/config issues: PEFT version pinning versus stale auto-mapping references." }, { - "left": "issue:45405", - "right": "issue:45600", + "left": "issue:43756", + "right": "issue:45325", "accept": false, - "reason": "PEFT version pinning and stale auto_mappings config references are separate maintenance issues." + "reason": "Both mention RoPE/positioning in Qwen-family models, but one is about missing RoPE layers and the other about video position-id scaling." }, { "left": "issue:43742", "right": "issue:44589", "accept": false, - "reason": "Different failures: one is a MobileLLM key lookup error, the other is a Float8 storage lookup/type error." + "reason": "These are different loading/storage failures on different components, with no shared concrete code path." }, { "left": "issue:44484", "right": "issue:44704", "accept": false, - "reason": "save_pretrained shard sizing and AutoProcessor cached_file kwargs are unrelated API concerns." + "reason": "`max_shard_size` behavior and `cached_file` kwargs propagation are separate `save_pretrained`/load-pretrained concerns." + }, + { + "left": "issue:43531", + "right": "issue:44821", + "accept": false, + "reason": "One is a Qwen3-MoE sliding-window bug; the other is inability to load `AutoImageProcessor` from a URL." }, { "left": "issue:28282", "right": "issue:29127", "accept": false, - "reason": "Missing PyTorch import vs. LayoutLMv3 box-validation messaging are unrelated problems." + "reason": "A missing PyTorch dependency error is unrelated to a LayoutLMv3 box-validation error message issue." }, { "left": "issue:43697", "right": "issue:45216", "accept": false, - "reason": "RTDetrV2 inference mismatch and Qwen3.5 checkpoint serialization regressions are different code paths." - }, - { - "left": "issue:43531", - "right": "issue:44821", - "accept": false, - "reason": "Qwen3-MoE sliding window behavior and AutoImageProcessor URL loading are unrelated." + "reason": "One reports different runtime outputs in v5; the other is a regression in saved checkpoint serialization for Qwen3.5." }, { "left": "issue:44514", "right": "issue:45381", "accept": false, - "reason": "Both are Qwen2.5-VL issues, but chat-template batching and video position IDs are distinct defects." + "reason": "Both are Qwen2.5-VL-related, but one crashes in chat-template batching while the other has incorrect video vision position IDs." }, { "left": "issue:43643", "right": "issue:44561", "accept": false, - "reason": "Both mention trust_remote_code, but one is missing fields in AutoConfig and the other is a v5 API removal breakage." + "reason": "These are distinct config/utility regressions: missing fields in `AutoConfig` versus removal of `is_torch_fx_available` breaking remote-code models." }, { "left": "issue:44625", "right": "issue:44877", "accept": false, - "reason": "Qwen3.5 num_labels propagation and granite_speech strict config loading are separate bugs." + "reason": "Qwen3.5 label propagation and strict granite_speech config loading are different configuration bugs." }, { "left": "issue:43582", "right": "issue:44977", "accept": false, - "reason": "An Apple Silicon TypeError in caching_allocator_warmup is unrelated to Qwen3.5 flash-attention generation." + "reason": "Apple Silicon allocator warmup and Qwen3.5 flash-attention generation are unrelated failures." }, { "left": "issue:43452", "right": "issue:43577", "accept": false, - "reason": "gguf_file loading failures and Blip2 dtype propagation are different model-loading issues." + "reason": "Both concern model/tokenizer loading, but one is broken by `gguf_file` handling and the other by dtype propagation in Blip2." }, { "left": "issue:43720", "right": "issue:44898", "accept": false, - "reason": "BitNet packed-weight unpacking and Perceiver positional interpolation are unrelated fixes." + "reason": "These are different model-path bugs: BitNet packed-weight unpacking versus Perceiver interpolation at non-default resolution." }, { "left": "issue:43452", "right": "issue:44843", "accept": false, - "reason": "Both affect loading workflows, but gguf_file handling and offline model_info calls are different bugs." + "reason": "`gguf_file` loading and offline `model_info()` calls in Mistral regex patching are separate failure modes." }, { "left": "issue:44514", "right": "issue:45290", "accept": false, - "reason": "Both involve apply_chat_template, but batched padding behavior and tool-call messages with no content are different edge cases." + "reason": "Both touch chat templates, but the failures are different: batched `padding=False` versus tool-call assistant messages without content." }, { "left": "issue:44466", "right": "issue:45005", "accept": false, - "reason": "Both concern tied weights, but one is device-dependent lm_head serialization and the other is broader translation-model behavior; not clearly the same code-path bug." + "reason": "Both involve tied weights, but one is about serialization depending on device and the other is a translation-model issue in v5." }, { "left": "issue:43819", "right": "issue:44811", "accept": false, - "reason": "DAC latent reconstruction and Whisper batch_decode special-token handling are unrelated." + "reason": "DAC latent decoding mismatch and Whisper `batch_decode` ignoring `skip_special_tokens` are unrelated decoding bugs." }, { "left": "issue:42175", "right": "issue:44488", "accept": false, - "reason": "Backend packaging for TensorFlow extra and loading cjvt/sleng-bert are unrelated issues." + "reason": "A packaging/backend dependency omission is not the same as a model-specific load failure for `cjvt/sleng-bert`." }, { "left": "issue:41762", "right": "issue:44936", "accept": false, - "reason": "DeepSpeed ZeRO-3 Gemma3 loading and trainer.evaluate() failure after train do not share a concrete bug." + "reason": "Gemma3 ZeRO-3 loading and `trainer.evaluate()` after `trainer.train()` are different runtime paths." }, { "left": "issue:43295", "right": "issue:45356", "accept": false, - "reason": "A processor.tokenizer regression and a Kimi-K2.5 tokenizer codec/fix_mistral_regex regression are distinct." + "reason": "These are separate regressions: processor/tokenizer access in one case and Kimi-K2.5 codec/fix-mistral-regex behavior in the other." }, { "left": "issue:43653", "right": "issue:44568", "accept": false, - "reason": "Both are tokenizer special-token problems, but they affect different tokenizers and different underlying behaviors." + "reason": "BigBird special-token registration and mDeBERTa BOS/EOS insertion are different tokenizer behaviors." }, { "left": "issue:24643", "right": "issue:29127", "accept": false, - "reason": "DeepSpeed weight dimensionality failure and LayoutLMv3 box-error messaging are unrelated." + "reason": "DeepSpeed training shape error and LayoutLMv3 box-validation messaging are not the same bug." }, { "left": "issue:38175", "right": "issue:43295", "accept": false, - "reason": "Siglip2 zero probabilities and a processor.tokenizer regression are unrelated model/processor bugs." + "reason": "SigLIP2 zero-probability output and a processor/tokenizer regression are unrelated issues." } ] }, @@ -5820,6 +5739,7 @@ "issue:43295", "issue:43329", "issue:43452", + "issue:43493", "issue:43504", "issue:43525", "issue:43531", @@ -5832,7 +5752,7 @@ "issue:43720", "issue:43742", "issue:43749", - "issue:43756", + "issue:43825", "issue:43906", "issue:43927", "issue:44361", @@ -5845,7 +5765,6 @@ "issue:44821", "issue:44843", "issue:44849", - "issue:44913", "issue:44936", "issue:44977", "issue:44991", @@ -5862,8 +5781,8 @@ "issue:43577|issue:43697", "issue:44464|issue:44743", "issue:44843|issue:44991", - "issue:41720|issue:43906", "issue:44479|issue:44936", + "issue:41720|issue:43906", "issue:44977|issue:45356", "issue:43688|issue:45440", "issue:43531|issue:43742", @@ -5880,242 +5799,242 @@ "issue:28282|issue:35141", "issue:43504|issue:44610", "issue:43452|issue:44451", - "issue:41720|issue:44464", "issue:43742|issue:44821", + "issue:41720|issue:44464", "issue:43329|issue:45325", - "issue:43688|issue:43749", "issue:33666|issue:36010", - "issue:44849|issue:45357", + "issue:43688|issue:43749", "issue:43525|issue:43572", + "issue:44849|issue:45357", "issue:43262|issue:43295", - "issue:43756|issue:44913" + "issue:43493|issue:43825" ], - "prepared_review_unit_hash": "553019f655697dc9e055e0bffad575d07a794bbf7f6d77a02cd366aac4462e0e", + "prepared_review_unit_hash": "31333bb96adb2b6e5edc13f3651c1b8c326606a9cef8289d18b5ec3ec0b34e64", "component_budget": { "node_count": 387, "item_count": 387, "soft_pair_count": 1455, - "serialized_chars": 749552, - "estimated_input_tokens": 187388, - "estimated_eval_tokens": 375032 + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 }, "budget": { "node_count": 47, "item_count": 47, "soft_pair_count": 33, - "serialized_chars": 37159, - "estimated_input_tokens": 9290, - "estimated_eval_tokens": 18836 + "serialized_chars": 37174, + "estimated_input_tokens": 9294, + "estimated_eval_tokens": 18844 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is heterogeneous: most items are unrelated bug reports across tokenizers, trainers, multimodal processors, and model-loading regressions. I did not find any soft pair that looks like the same underlying bug/change; several are only loosely similar by subsystem or version/regression theme.", - "confidence": 0.89, - "canonical_issue_reason": "issue:44843 is the best representative of the set because it is a broad, user-facing tokenizer-loading regression in a core code path and is more likely than the others to attract duplicate reports.", + "summary": "The cluster is mostly a collection of unrelated Transformers bug reports rather than a single duplicate set. I did not find any soft pair that clearly shares the same underlying bug or fix path. The most central issue in the set is the FSDP CPU RAM efficient loading regression.", + "confidence": 0.71, + "canonical_issue_reason": "Issue 43749 has the strongest cluster-centrality signal (highest discussion/inbound refs) and reads like a broader regression report, while the other items are largely distinct model/tokenizer/training bugs.", "canonical_pr_reason": null, - "best_issue_reason": "issue:44843 is the strongest global anchor: it is cross-cutting, actionable, and affects a common API surface rather than a model-specific edge case.", + "best_issue_reason": "Issue 43749 is the best anchor for this cluster because it is well-discussed, widely referenced, and represents a substantive regression rather than a narrowly model-specific edge case.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44361", "right": "issue:44442", "accept": false, - "reason": "Different failures in different tokenizer classes; one is an AttributeError in MLuke tasks, the other is AutoTokenizer failing to load FastSpeech2ConformerTokenizer." + "reason": "Both are tokenizer-loading failures, but they affect different tokenizer classes and different failure modes; not the same underlying bug." }, { "left": "issue:43653", "right": "issue:44361", "accept": false, - "reason": "BigBird mask-token special-token decode bug vs MLuke task AttributeError; same area only at a very high level." + "reason": "Different tokenizer/model paths and different symptoms; no evidence they share a common code-path bug." }, { "left": "issue:30064", "right": "issue:33666", "accept": false, - "reason": "Image processor void-segmentation-map handling is unrelated to Qwen2-VL multi-GPU training." + "reason": "Image processor void-segmentation handling and Qwen2-VL multi-GPU training are unrelated issues." }, { "left": "issue:43577", "right": "issue:43697", "accept": false, - "reason": "Different model families and different bugs: dtype propagation in Blip2 vs RTDetr output drift in v5." + "reason": "One is a dtype propagation problem in Blip2 loading, the other is an output regression in RTDetrV2; not the same fix." }, { "left": "issue:44464", "right": "issue:44743", "accept": false, - "reason": "Both concern generation/cache behavior, but one is compiled-forward chunked generation and the other is recurrent-state reset in modular_qwen3_5.py." + "reason": "Both involve generation/caching behavior, but they describe different mechanisms and code paths." }, { "left": "issue:44843", "right": "issue:44991", "accept": false, - "reason": "Both are tokenizer-loading regressions, but one is offline model_info() patching and the other is a model-specific tokenizer failure." + "reason": "Both are tokenizer-loading regressions, but one is specifically offline model_info patching and the other is a different tokenizer failure." }, { - "left": "issue:41720", - "right": "issue:43906", + "left": "issue:44479", + "right": "issue:44936", "accept": false, - "reason": "43906 is an isolated reproduction of a different issue, not the same A800 cudaErrorAssert failure." + "reason": "Video-input regression in Qwen VL models and trainer evaluate-after-train failure are unrelated." }, { - "left": "issue:44479", - "right": "issue:44936", + "left": "issue:41720", + "right": "issue:43906", "accept": false, - "reason": "Video-input regression in Qwen VL models vs trainer.evaluate() failing after train; no shared code path." + "reason": "The isolated reproduction does not clearly target the same Qwen3 auto-device-map CUDA assert; insufficient evidence of a shared bug." }, { "left": "issue:44977", "right": "issue:45356", "accept": false, - "reason": "Qwen3.5 flash-attention generation bug and Kimi-K2.5 tokenizer codec regression are unrelated." + "reason": "Qwen3.5 flash-attention generation and Kimi-K2.5 tokenizer codec regression are different problems." }, { "left": "issue:43688", "right": "issue:45440", "accept": false, - "reason": "Both are MoE-related, but one is auxiliary-loss normalization and the other is implementation divergence from a remote model." + "reason": "Both concern MoE models, but one is auxiliary-loss normalization and the other is implementation divergence; not the same underlying defect." }, { "left": "issue:43531", "right": "issue:43742", "accept": false, - "reason": "Sliding-window behavior in Qwen3-MoE has nothing to do with loading facebook/MobileLLM-125M." + "reason": "Sliding-window behavior in Qwen3-MoE and a key error loading MobileLLM are unrelated." }, { "left": "issue:43577", "right": "issue:45588", "accept": false, - "reason": "Blip2 dtype loading bug vs flash_attention.py crashing on s_aux=None; different subsystems and symptoms." + "reason": "Dtype propagation in Blip2 loading is unrelated to flash-attention crashing on s_aux=None." }, { "left": "issue:43653", "right": "issue:43927", "accept": false, - "reason": "BigBird special-token decode issue is unrelated to DiaConfig losing custom token IDs on save/load." + "reason": "BigBird tokenizer special-token registration and DiaConfig custom token IDs are different loading/tokenization issues." }, { "left": "issue:44464", "right": "issue:44977", "accept": false, - "reason": "Compiled-forward chunked generation inconsistency is not the same bug as Qwen3.5 flash-attention generation failure." + "reason": "Both are generation regressions, but one is compiled-forward inconsistency and the other is flash-attention-specific generation behavior." }, { "left": "issue:43550", "right": "issue:43927", "accept": false, - "reason": "torch.compile+SDPA failure in Bamba is unrelated to DiaConfig save/load token ID loss." + "reason": "torch.compile + SDPA failure for Bamba and DiaConfig token-ID persistence are unrelated." }, { "left": "issue:43550", "right": "issue:44361", "accept": false, - "reason": "Different model families and failure modes; no shared underlying defect." + "reason": "Different models and failure modes; no shared concrete code-path problem." }, { "left": "issue:32090", "right": "issue:34567", "accept": false, - "reason": "Trainer broadcast NoneType error vs TrainerState num_input_tokens_seen not updating; separate trainer bugs." + "reason": "Trainer broadcast NoneType error and num_input_tokens_seen not updating are unrelated trainer bugs." }, { "left": "issue:43122", "right": "issue:43906", "accept": false, - "reason": "Tokenization differences across versions are not the same as the isolated reproduction of another issue." + "reason": "Tokenization differences across versions and an isolated reproduction of another issue are not enough to conclude the same bug." }, { "left": "issue:43504", "right": "issue:43720", "accept": false, - "reason": "Beit legacy-field loading bug and BitNet packed-weight unpacking bug affect different loaders and models." + "reason": "Legacy-field model loading and BitNet packed-weight unpacking are different loading paths and fixes." }, { "left": "issue:43504", "right": "issue:43550", "accept": false, - "reason": "Different model-loading regressions with different root causes; too broad to merge." + "reason": "Beit preset loading and Bamba torch.compile failure are unrelated." }, { "left": "issue:32090", "right": "issue:33666", "accept": false, - "reason": "Trainer NoneType broadcast error is unrelated to Qwen2-VL multi-GPU training." + "reason": "Trainer NoneType broadcast and Qwen2-VL multi-GPU training are not the same issue." }, { "left": "issue:28282", "right": "issue:35141", "accept": false, - "reason": "PyTorch-missing ImportError and token-embedding reinitialization are unrelated." + "reason": "PyTorch-missing import error and embedding reinitialization after resize are unrelated." }, { "left": "issue:43504", "right": "issue:44610", "accept": false, - "reason": "Beit pretrained preset loading and OmDet-Turbo input-size mismatch are distinct bugs." + "reason": "Model preset loading and processor/model input-size mismatch are different bugs." }, { "left": "issue:43452", "right": "issue:44451", "accept": false, - "reason": "Both are model/tokenizer loading failures, but they affect different APIs and likely different root causes." - }, - { - "left": "issue:41720", - "right": "issue:44464", - "accept": false, - "reason": "A800 cudaErrorAssert in auto device mapping is unrelated to compiled-forward chunked generation inconsistency." + "reason": "gguf_file loading breakage and inability to load a specific model are not clearly the same underlying failure." }, { "left": "issue:43742", "right": "issue:44821", "accept": false, - "reason": "MobileLLM-125M loading key error and AutoImageProcessor-from-URL failure are different load paths." + "reason": "MobileLLM key error and AutoImageProcessor-from-URL loading are separate code paths." }, { - "left": "issue:43329", - "right": "issue:45325", + "left": "issue:41720", + "right": "issue:44464", "accept": false, - "reason": "Undefined helper/merge_size in multimodal token counting is unrelated to Qwen2.5-VL rope-index scaling for still images." + "reason": "CUDA assert in auto device mapping and compiled-forward generation inconsistency are unrelated." }, { - "left": "issue:43688", - "right": "issue:43749", + "left": "issue:43329", + "right": "issue:45325", "accept": false, - "reason": "MoE auxiliary-loss normalization and FSDP CPU RAM efficient loading are different bugs." + "reason": "Video token counting bug and Qwen2.5-VL rope-index scaling bug are different multimodal issues." }, { "left": "issue:33666", "right": "issue:36010", "accept": false, - "reason": "Qwen2-VL multi-GPU training is unrelated to the GenerationMixin import regression." + "reason": "Qwen2-VL multi-GPU training and GenerationMixin import failure are unrelated." }, { - "left": "issue:44849", - "right": "issue:45357", + "left": "issue:43688", + "right": "issue:43749", "accept": false, - "reason": "Qwen3.5 hidden-state bug and save_pretrained visual-encoder key regression are separate issues." + "reason": "Auxiliary-loss normalization and FSDP CPU RAM efficient loading are different bug classes with no shared fix path." }, { "left": "issue:43525", "right": "issue:43572", "accept": false, - "reason": "Both are config attribute problems, but one is Llama4 pad_token_id and the other is StableLm pad_token_idx after v5.0; different models and fields." + "reason": "Both mention missing padding token config fields, but they affect different model configs and are not clearly one bug." + }, + { + "left": "issue:44849", + "right": "issue:45357", + "accept": false, + "reason": "Qwen3.5 hidden-states generation bug and incorrect visual encoder keys on save_pretrained are different regressions." }, { "left": "issue:43262", "right": "issue:43295", "accept": false, - "reason": "Audio processor sampling-rate defaulting and processor.tokenizer/image-passing regression are not the same code-path problem." + "reason": "Audio chat-template sampling-rate default and processor.tokenizer/image passing regression are unrelated processor issues." }, { - "left": "issue:43756", - "right": "issue:44913", + "left": "issue:43493", + "right": "issue:43825", "accept": false, - "reason": "Smollm3 RoPE-layer mismatch and GPTNeoX rotary_pct reload loss are unrelated configuration bugs." + "reason": "SigLIP2 implementation discrepancy and pipeline translation-task messaging are entirely separate concerns." } ] }, @@ -6164,7 +6083,6 @@ "issue:43720", "issue:43756", "issue:43824", - "issue:43825", "issue:43844", "issue:43927", "issue:44291", @@ -6178,6 +6096,7 @@ "issue:44625", "issue:44734", "issue:44843", + "issue:44913", "issue:44964", "issue:44977", "issue:45083", @@ -6189,16 +6108,16 @@ "issue:45704" ], "soft_pairs": [ - "issue:43493|issue:43825", - "issue:43540|issue:43716", + "issue:43756|issue:44913", "issue:43720|issue:43927", + "issue:43540|issue:43716", "issue:43606|issue:43927", "issue:28282|issue:32090", "issue:28282|issue:33357", "issue:43334|issue:43572", "issue:43577|issue:44554", - "issue:43688|issue:45704", "issue:33666|issue:36331", + "issue:43688|issue:45704", "issue:39692|issue:43493", "issue:43720|issue:44361", "issue:44964|issue:45405", @@ -6223,232 +6142,232 @@ "issue:43575|issue:45405", "issue:43504|issue:43653" ], - "prepared_review_unit_hash": "429d314ea55ed81cf4eb310591d0ca20d4d9066a75c17163bbc7ca2fd24f5809", + "prepared_review_unit_hash": "9edd495e375239f7a96a0c944c4e0e9ac29507c9e42d24b4ddbcb1b8e785f857", "component_budget": { "node_count": 387, "item_count": 387, "soft_pair_count": 1455, - "serialized_chars": 749552, - "estimated_input_tokens": 187388, - "estimated_eval_tokens": 375032 + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 33, - "serialized_chars": 37898, - "estimated_input_tokens": 9475, - "estimated_eval_tokens": 19206 + "serialized_chars": 37881, + "estimated_input_tokens": 9471, + "estimated_eval_tokens": 19198 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is highly heterogeneous; the soft-similarity pairs mostly connect unrelated bugs across loading, config, training, tokenizer, and multimodal code paths. I found no true duplicate pairs to merge.", - "confidence": 0.95, - "canonical_issue_reason": "No single canonical issue is representative; the items split into many unrelated subtopics and failure modes.", + "summary": "This is a heterogeneous cluster of mostly unrelated Transformers bug reports. The soft pairs share broad vocabulary like loading, configs, dtype, or Qwen, but they describe different concrete failures and code paths, so they should not be merged as duplicates.", + "confidence": 0.97, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "No global best issue stands out as a faithful representative of the cluster; at most, the items form several separate mini-clusters rather than one shared bug.", + "best_issue_reason": null, "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:43493", - "right": "issue:43825", + "left": "issue:43756", + "right": "issue:44913", "accept": false, - "reason": "SigLIP2 implementation discrepancy vs pipeline translation-task messaging; different subsystems and bugs." + "reason": "Different bugs: RoPE layer selection in Smollm3 vs GPTNeoX rotary_pct not persisting on reload." }, { - "left": "issue:43540", - "right": "issue:43716", + "left": "issue:43720", + "right": "issue:43927", "accept": false, - "reason": "Video processing ValueError in Qwen3OmniMoe vs image-preprocessor dtype mismatch in Mistral-3; unrelated model paths." + "reason": "Different failures: packed BitNet weights during accelerate loading vs DiaConfig losing custom token IDs on save/load." }, { - "left": "issue:43720", - "right": "issue:43927", + "left": "issue:43540", + "right": "issue:43716", "accept": false, - "reason": "Packed-weight loading failure in BitNet vs DiaConfig token-id persistence bug; different code paths and symptoms." + "reason": "Different multimodal issues: Qwen3OmniMoe video processing error vs Mistral-3 image preprocessor/model dtype mismatch." }, { "left": "issue:43606", "right": "issue:43927", "accept": false, - "reason": "CPU offload device mismatch in bark-small vs DiaConfig save/load token IDs; not the same bug." + "reason": "Different code paths: CPU offload device mismatch in bark-small vs config token ID persistence bug in DiaConfig." }, { "left": "issue:28282", "right": "issue:32090", "accept": false, - "reason": "PyTorch missing ImportError vs Trainer _gpu_broadcast_one NoneType error; unrelated failures." + "reason": "Both are runtime errors, but one is missing PyTorch import dependency and the other is Trainer GPU broadcast handling of NoneType." }, { "left": "issue:28282", "right": "issue:33357", "accept": false, - "reason": "PyTorch missing ImportError vs MacOS bus error on CLIP; different environment and runtime issue." + "reason": "Unrelated problems: missing PyTorch dependency vs MacOS bus error loading a CLIP model." }, { "left": "issue:43334", "right": "issue:43572", "accept": false, - "reason": "Missing pad_token_id in Qwen3-VL text config vs missing pad_token_idx in StableLmConfig; similar wording only." + "reason": "Both mention missing config fields, but Qwen3-VL pad_token_id and StableLm pad_token_idx are different model/config regressions." }, { "left": "issue:43577", "right": "issue:44554", "accept": false, - "reason": "Blip2 dtype propagation bug vs MPS attention correctness issue; separate problems." + "reason": "Different dtype issues: BLIP2 components staying float32 vs an MPS attention correctness bug with mismatched head dimensions." }, { - "left": "issue:43688", - "right": "issue:45704", + "left": "issue:33666", + "right": "issue:36331", "accept": false, - "reason": "Auxiliary loss normalization bug vs T5 apex RMSNorm memory leak; no shared underlying defect." + "reason": "Qwen2-VL multi-GPU training support question/problem is not the same as CustomTrainer compute_loss signature breakage." }, { - "left": "issue:33666", - "right": "issue:36331", + "left": "issue:43688", + "right": "issue:45704", "accept": false, - "reason": "Qwen2-VL multi-GPU training request vs custom Trainer compute_loss signature breakage; unrelated." + "reason": "Both involve training behavior, but OLMoE/GPT OSS auxiliary loss normalization is unrelated to T5 using apex FusedRMSNorm memory leak." }, { "left": "issue:39692", "right": "issue:43493", "accept": false, - "reason": "Docs example/model-processor mismatch vs SigLIP2 HF-vs-JAX implementation discrepancy; not the same concrete bug." + "reason": "SigLIP2 docs/example errors are not the same as a deeper HF vs JAX implementation discrepancy." }, { "left": "issue:43720", "right": "issue:44361", "accept": false, - "reason": "BitNet packed-weight loading vs MLukeTokenizer AttributeError; unrelated load/tokenizer issues." + "reason": "Different loading issues: BitNet packed weight unpacking vs MLukeTokenizer task-time AttributeError." }, { "left": "issue:44964", "right": "issue:45405", "accept": false, - "reason": "Model loading regression vs PEFT version bump to unreleased release; no shared bug." + "reason": "Cannot load Phi-4 multimodal model is unrelated to an external minimum PEFT version bump that is unreleased on PyPI." }, { "left": "issue:44479", "right": "issue:44625", "accept": false, - "reason": "Video regression in Qwen models vs num_labels propagation in Qwen3.5 config; different failure modes." + "reason": "Qwen3.5 video input regression is unrelated to num_labels propagation between configs." }, { "left": "issue:43540", "right": "issue:44977", "accept": false, - "reason": "Qwen3OmniMoe video-input ValueError vs Qwen3.5 flash-attention generation bug; not the same path." + "reason": "Both are Qwen3-family issues, but one is video input handling and the other is flash-attention generation correctness." }, { "left": "issue:43493", "right": "issue:43844", "accept": false, - "reason": "SigLIP2 implementation mismatch vs ZeRO-3 gradient blow-up; unrelated training semantics." + "reason": "SigLIP2 implementation discrepancy is not the same as ZeRO-3 gradient inflation with HfDeepSpeedConfig." }, { "left": "issue:44488", "right": "issue:44843", "accept": false, - "reason": "Model load failure for cjvt/sleng-bert vs offline hub regression in AutoTokenizer; different causes." + "reason": "Loading a specific model failing is unrelated to AutoTokenizer making an unconditional hub model_info call in offline mode." }, { "left": "issue:43756", "right": "issue:45083", "accept": false, - "reason": "Smollm3 RoPE layer count bug vs qwen3_omni_moe feature-length helper behavior; unrelated." + "reason": "Smollm3 RoPE layer count issue does not match a helper-length calculation bug in qwen3_omni_moe." }, { "left": "issue:45200", "right": "issue:45229", "accept": false, - "reason": "Gemma4 text-only fine-tuning token-type default vs multi-GPU inference OOM; not a duplicate." + "reason": "Gemma 4 text-only mm_token_type_ids defaulting is a configuration bug, while 31B multi-GPU inference OOM is a capacity/runtime issue." }, { "left": "issue:43278", "right": "issue:43844", "accept": false, - "reason": "Embedding dtype drift across train/eval vs abnormal gradient increase under ZeRO-3; separate numerical issues." + "reason": "Embedding dtype changing between train and eval is unrelated to DeepSpeed ZeRO-3 gradient scaling behavior." }, { "left": "issue:43257", "right": "issue:43299", "accept": false, - "reason": "Qwen3 MoE loading/weight-conversion issue vs Qwen3VL MoE dev-breakage; related model family but different concrete bugs." + "reason": "Qwen3 MOE weight conversion with accelerate+deepspeed is not the same as Qwen3VL loading breakage from a dev-version config/API change." }, { "left": "issue:43646", "right": "issue:44466", "accept": false, - "reason": "Custom model initialization breakage vs inconsistent lm_head.weight serialization; different code paths." + "reason": "Transformers 5 custom model initialization breakage is broader and different from inconsistent lm_head.weight serialization depending on device." }, { "left": "issue:43824", "right": "issue:44589", "accept": false, - "reason": "Missing Qwen2_5_VL import vs missing Float8 storage object; unrelated import/runtime errors." + "reason": "Missing Qwen2_5_VL class import is not the same underlying issue as missing Float8 storage support." }, { "left": "issue:39692", "right": "issue:43295", "accept": false, - "reason": "Docs example problems vs v4.57.5 custom processor/tokenizer regression; not the same underlying bug." + "reason": "Doc/example errors in SigLIP2 are unrelated to a v4.57.5 regression in processor.tokenizer and image tokenization behavior." }, { "left": "issue:43688", "right": "issue:45701", "accept": false, - "reason": "Aux-loss normalization vs tokenization changes across versions; no shared defect." + "reason": "Aux loss normalization and version-dependent tokenization changes are different bugs with different failure modes." }, { "left": "issue:44589", "right": "issue:45520", "accept": false, - "reason": "Float8 storage lookup failure vs flash_attn import_utils KeyError on Python 3.13; different incompatibilities." + "reason": "Both are import/storage related, but Float8 storage lookup and flash_attn KeyError on Python 3.13 are distinct issues." }, { "left": "issue:33357", "right": "issue:33666", "accept": false, - "reason": "MacOS CLIP bus error vs Qwen2-VL multi-GPU training request; unrelated." + "reason": "MacOS bus error loading a CLIP model is unrelated to Qwen2-VL multi-GPU training behavior." }, { "left": "issue:43504", "right": "issue:43606", "accept": false, - "reason": "Legacy field loading failure in Beit vs device mismatch in bark-small CPU offload; different models and bugs." + "reason": "Legacy field preventing Beit semantic segmentation loading is unrelated to bark-small CPU offload device mismatch." }, { "left": "issue:44423", "right": "issue:44734", "accept": false, - "reason": "serve continuous-batching multimodal crash vs KV-cache continuation tensor indexing bug; separate server paths." + "reason": "Continuous-batching crash on multimodal models is a different code path from /v1/responses KV-cache continuation indexing." }, { "left": "issue:43572", "right": "issue:44291", "accept": false, - "reason": "Missing pad_token_idx in StableLmConfig vs init_empty_weights unexpected argument; unrelated config/init issues." + "reason": "Missing pad_token_idx in StableLmConfig is not the same as init_empty_weights passing an unexpected _is_hf_initialized argument." }, { "left": "issue:43927", "right": "issue:44361", "accept": false, - "reason": "DiaConfig token-ID persistence bug vs MLukeTokenizer task AttributeError; different failure mechanisms." + "reason": "DiaConfig token ID persistence and MLukeTokenizer task AttributeError are unrelated." }, { "left": "issue:43575", "right": "issue:45405", "accept": false, - "reason": "TP loading OOM for a model vs version bump of MIN_PEFT_VERSION; no common bug." + "reason": "Tensor-parallel OOM while loading Qwen2-57B is unrelated to a PEFT minimum-version packaging issue." }, { "left": "issue:43504", "right": "issue:43653", "accept": false, - "reason": "Beit legacy field loading issue vs BigBirdTokenizer special-token registration bug; different tokenizer/model surfaces." + "reason": "Beit legacy field loading bug is not the same as BigBirdTokenizer mask token registration/empty decode output." } ] }, @@ -6522,8 +6441,8 @@ ], "soft_pairs": [ "issue:44625|issue:45759", - "issue:43606|issue:44112", "issue:44792|issue:45405", + "issue:43606|issue:44112", "issue:43994|issue:44493", "issue:44521|issue:45290", "issue:44464|issue:45584", @@ -6534,8 +6453,8 @@ "issue:43688|issue:43873", "issue:15354|issue:38175", "issue:44554|issue:45446", - "issue:44568|issue:45701", "issue:43540|issue:45200", + "issue:44568|issue:45701", "issue:43232|issue:44849", "issue:45072|issue:45198", "issue:39401|issue:44568", @@ -6548,191 +6467,200 @@ "issue:43334|issue:43525", "issue:44466|issue:45127" ], - "prepared_review_unit_hash": "52859ebe79c4e8e318376d22a29abbbd232157566af8d7d688f3949c4aecf29e", + "prepared_review_unit_hash": "e5145c0e163026d7898acccc7789cb8c4838488fc10669c1d1fd9c44a109cd1f", "component_budget": { "node_count": 387, "item_count": 387, "soft_pair_count": 1455, - "serialized_chars": 749552, - "estimated_input_tokens": 187388, - "estimated_eval_tokens": 375032 + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 }, "budget": { "node_count": 47, "item_count": 47, "soft_pair_count": 26, - "serialized_chars": 34986, - "estimated_input_tokens": 8747, - "estimated_eval_tokens": 17750 + "serialized_chars": 34990, + "estimated_input_tokens": 8748, + "estimated_eval_tokens": 17752 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This is a heterogeneous cluster of unrelated bug reports; the soft-similarity pairs are all superficial and should not be merged as duplicates.", - "confidence": 0.94, - "canonical_issue_reason": "No single canonical issue is a safe anchor here: the items span unrelated tokenizer, generation, training, multimodal, and CI regressions.", + "summary": "Most items are unrelated bug reports spanning tokenizers, multimodal processors, generation, and training infra. The only plausible duplicate pair is the missing `pad_token_id` config-load failure in Qwen3-VL and Llama4; the rest are distinct issues.", + "confidence": 0.61, + "canonical_issue_reason": "issue:43525 is the best root-style representative because it captures the shared composite-config loading failure around missing `pad_token_id`, which is the closest true duplicate theme in this set.", "canonical_pr_reason": null, - "best_issue_reason": "No global best issue stands out; none of the issues covers the cluster\u2019s disparate failure modes well enough to serve as a representative duplicate target.", + "best_issue_reason": "issue:43525 is the most reusable global representative of the cluster: it is a concrete, actionable loading bug and generalizes better than the model-specific report.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44625", "right": "issue:45759", "accept": false, - "reason": "Both concern Qwen3.5/Qwen3.6 config handling, but one is num_labels propagation and the other is unwrapping text_config for multimodal from_config; different code paths and symptoms." + "reason": "Both involve Qwen 3.5 config behavior, but one is `num_labels` propagation and the other is `from_config` unwrapping `text_config`; different code paths and symptoms." }, { - "left": "issue:43606", - "right": "issue:44112", + "left": "issue:44792", + "right": "issue:45405", "accept": false, - "reason": "One is a CPU offload device-mismatch with quantization, the other is a stale device override CI failure in GraniteSpeech; not the same bug." + "reason": "A failed janus image-generation test and an unmet PEFT version requirement are unrelated." }, { - "left": "issue:44792", - "right": "issue:45405", + "left": "issue:43606", + "right": "issue:44112", "accept": false, - "reason": "A failing janus model test and a MIN_PEFT_VERSION release mismatch are unrelated." + "reason": "Both mention device/CI problems, but one is a real CPU-offload mismatch and the other is a stale test; not the same bug." }, { "left": "issue:43994", "right": "issue:44493", "accept": false, - "reason": "Model/pipeline nonsense outputs and unexpected position-id keys point to different failures and fixes." + "reason": "Different failures: one is bad model outputs for SigLIP2, the other is unexpected position-id keys across many models." }, { "left": "issue:44521", "right": "issue:45290", "accept": false, - "reason": "Both involve apply_chat_template, but one is all-zero assistant masks for multimodal inputs while the other crashes on tool-call assistant messages with no content." + "reason": "Both touch chat templating, but one is all-zero assistant masks and the other is a crash on tool-call assistant messages." }, { "left": "issue:44464", "right": "issue:45584", "accept": false, - "reason": "Chunked generation with compiled forward and Whisper empty-transcription handling are different code paths." + "reason": "Chunked generation with compiled forward and Whisper empty-transcription handling are distinct generation bugs." }, { "left": "issue:29942", "right": "issue:30333", "accept": false, - "reason": "Flash Attention 2 CI test failures are unrelated to MLflow job-status reporting." + "reason": "A FlashAttention CI failure and an MLflow reporting issue are unrelated." }, { "left": "issue:28218", "right": "issue:34567", "accept": false, - "reason": "Tokenizer spacing after added tokens is unrelated to TrainerState.num_input_tokens_seen not updating." + "reason": "Tokenizer spacing after added tokens and TrainerState token counters do not share the same underlying defect." }, { "left": "issue:43844", "right": "issue:44964", "accept": false, - "reason": "ZeRO-3 gradient inflation and loading a multimodal model are different problems." + "reason": "ZeRO-3 gradient growth and loading Phi-4 multimodal are different training vs model-loading issues." }, { "left": "issue:45084", "right": "issue:45520", "accept": false, - "reason": "A compilation error for non-template nodes is not the same as a Python 3.13 flash_attn import KeyError." + "reason": "Template-node compilation and a Python 3.13 import KeyError are separate failures." }, { "left": "issue:43688", "right": "issue:43873", "accept": false, - "reason": "Auxiliary-loss normalization and quantization/offloading behavior are distinct issues." + "reason": "Aux-loss normalization and quantization offload behavior are unrelated." }, { "left": "issue:15354", "right": "issue:38175", "accept": false, - "reason": "torch.jit.script GeneratorExp support is unrelated to unexpected zero probabilities in siglip2." + "reason": "TorchScript generator-expression export failure and SigLIP2 zero probabilities are different problems." }, { "left": "issue:44554", "right": "issue:45446", "accept": false, - "reason": "MPS attention correctness with mismatched head dims is unrelated to a PyTorch version check guarding AuxRequest imports." + "reason": "MPS attention correctness and a flex_attention PyTorch version check are unrelated code paths." }, { - "left": "issue:44568", - "right": "issue:45701", + "left": "issue:43540", + "right": "issue:45200", "accept": false, - "reason": "Both mention tokenization regressions, but one is a specific mdeberta-v3-base special-token bug and the other is a broader version-dependent tokenization change." + "reason": "Qwen3OmniMoe video-processing failure and Gemma 4 token-type defaults are different multimodal bugs." }, { - "left": "issue:43540", - "right": "issue:45200", + "left": "issue:44568", + "right": "issue:45701", "accept": false, - "reason": "Qwen3OmniMoe video-input processing and Gemma 4 mm_token_type_ids defaults are different multimodal issues." + "reason": "One is a specific `add_special_tokens` regression for mDeBERTa; the other is a broad tokenizer-version change report." }, { "left": "issue:43232", "right": "issue:44849", "accept": false, - "reason": "_update_model_kwargs_for_generation after sync_gpus and Qwen3.5 output_hidden_states are separate generation/config bugs." + "reason": "Generation kwargs handling after `sync_gpus` and a Qwen3.5 hidden-states bug are not the same issue." }, { "left": "issue:45072", "right": "issue:45198", "accept": false, - "reason": "bf16 dtype mismatches in inference do not match Wav2Vec2 save_pretrained/tokenization failures." + "reason": "bfloat16 dtype mismatches and Wav2Vec2 save/tokenization failure are unrelated." }, { "left": "issue:39401", "right": "issue:44568", "accept": false, - "reason": "Both are tokenizer bugs, but offset_mapping errors for Qwen3 and BOS/EOS insertion failures for mdeberta-v3-base are not the same underlying defect." + "reason": "Both are tokenizer bugs, but offset mapping and special-token insertion are distinct behaviors." }, { "left": "issue:43278", "right": "issue:43701", "accept": false, - "reason": "Embedding dtype drift in evaluate and resume_from_checkpoint key mismatch are unrelated training-state issues." + "reason": "Embedding dtype drift between train/eval and checkpoint-key mismatch are unrelated." }, { "left": "issue:44315", "right": "issue:44849", "accept": false, - "reason": "Liger Kernel not applied with model_init has no concrete overlap with the Qwen3.5 hidden-states bug." + "reason": "Liger Kernel application with `model_init` and Qwen3.5 hidden-states output are different bugs." + }, + { + "left": "issue:28218", + "right": "issue:33357", + "accept": false, + "reason": "Tokenizer spacing and a MacOS bus error in CLIP are unrelated." }, { "left": "issue:45200", "right": "issue:45325", "accept": false, - "reason": "Gemma 4 mm_token_type_ids defaults and Qwen2.5-VL rope-index temporal scaling are different multimodal configuration bugs." + "reason": "Both are multimodal config issues, but one is missing `mm_token_type_ids` defaults and the other is incorrect temporal position scaling." }, { "left": "issue:43295", "right": "issue:44610", "accept": false, - "reason": "Processor/tokenizer regression when passing images is unrelated to OmDet-Turbo\u2019s 640px-vs-224px preprocessing mismatch." + "reason": "Both are processor/model regressions, but one is about `processor.tokenizer`/image passing and the other is an image-size mismatch." }, { "left": "issue:44554", "right": "issue:45356", "accept": false, - "reason": "An MPS attention correctness issue is unrelated to the Kimi-K2.5 tokenizer codec regression." + "reason": "Attention correctness on MPS and Kimi tokenizer codec handling are unrelated." }, { "left": "issue:43334", "right": "issue:43525", - "accept": false, - "reason": "They share a missing-pad_token_id symptom, but they affect different config classes and model families, so they are not clearly the same bug." + "accept": true, + "reason": "Both report loading composite multimodal configs failing because `pad_token_id` is missing on the text/config object; same underlying config-handling bug." }, { "left": "issue:44466", "right": "issue:45127", "accept": false, - "reason": "Inconsistent serialized tied weights and LoRA merge collapse are different problems in different parts of the stack." + "reason": "Serialization of tied weights and LoRA merge collapse are different failures." } ] }, - "evaluator_result": null, - "evaluator_used": false, + "evaluator_result": { + "accept": true, + "feedback": "The cluster summary is grounded in the packet, and the only accepted soft edge\u2014issue:43334 with issue:43525\u2014matches a clearly shared pad_token_id composite-config loading bug and looks mergeable into one fix. The other pairwise rejections stay conservative and are supported by the titles." + }, + "evaluator_used": true, "retried": false, - "accepted_nontrivial_soft_edge": false, + "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null }, @@ -6820,8 +6748,8 @@ "issue:43827|issue:43901", "issue:43653|issue:44112", "issue:43450|issue:43493", - "issue:44360|issue:44561", "issue:43450|issue:43575", + "issue:44360|issue:44561", "issue:44792|issue:45563", "issue:45127|issue:45245", "issue:43819|issue:45083", @@ -6834,245 +6762,248 @@ "issue:45290|issue:45381", "issue:43278|issue:43825" ], - "prepared_review_unit_hash": "649645b42f2d9c3bf57bfcbe416a86255fc3373d43c37ac16dda9e975322bd00", + "prepared_review_unit_hash": "eff1c7af9ccdd26c0e9141dc09922c3267a15603e72c4ab606a10f2754e5c6d9", "component_budget": { "node_count": 387, "item_count": 387, "soft_pair_count": 1455, - "serialized_chars": 749552, - "estimated_input_tokens": 187388, - "estimated_eval_tokens": 375032 + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 34, - "serialized_chars": 37798, - "estimated_input_tokens": 9450, - "estimated_eval_tokens": 19156 + "serialized_chars": 37791, + "estimated_input_tokens": 9448, + "estimated_eval_tokens": 19152 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The set is heterogeneous and does not contain any clear duplicate issue/PR cluster. All proposed soft pairs are false positives: they share broad Transformers themes but not the same underlying bug or change.", - "confidence": 0.94, - "canonical_issue_reason": "No true canonical issue emerges from this batch because the items cover unrelated regressions across imports, processors, training, tokenizers, and model-specific behavior. If one issue must anchor the set, issue:43824 is the strongest candidate because it is a concrete top-level import regression with many references and a clear failure mode.", + "summary": "This set is mostly heterogeneous and does not form one duplicate cluster. The only plausible duplicate-looking pair is the two SigLIP2 issues; the rest are separate bugs across unrelated models, processors, and training/runtime paths.", + "confidence": 0.82, + "canonical_issue_reason": "No single canonical issue fits the whole set because the items span many unrelated failure modes. If forced to anchor the most coherent duplicate subcluster, issue 43493 is the best representative because it is the broad SigLIP2 implementation-parity report that plausibly covers the zero-probabilities symptom.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43824 is the best global issue candidate: it is a crisp, user-visible regression, has the strongest duplicate-magnet signals (inbound refs/discussion), and describes a narrowly reproducible import failure.", + "best_issue_reason": "Issue 43493 is the strongest issue candidate: it is broader than the specific zero-probabilities report, points at the underlying model-implementation discrepancy, and is the best consolidation target for that subcluster.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44898", "right": "issue:45412", "accept": false, - "reason": "Different bugs: Perceiver non-default resolution handling vs RT-DETR model memory release." + "reason": "Different models and symptoms: Perceiver interpolation failure vs RT-DETR memory not being released." }, { "left": "issue:43493", "right": "issue:43701", "accept": false, - "reason": "Different failure classes: SigLIP2 implementation mismatch vs resume_from_checkpoint key mismatch." + "reason": "SigLIP2 model-parity bug vs checkpoint resume key mismatch; unrelated code paths." }, { "left": "issue:37428", "right": "issue:43957", "accept": false, - "reason": "Both are import/load regressions, but for unrelated code paths and models." + "reason": "ImportError for a flash-attention helper vs meta-device loading breakage; not the same defect." }, { "left": "issue:43572", "right": "issue:44987", "accept": false, - "reason": "Both involve v5 model/config compatibility, but one is a missing config field and the other is a model load failure for a specific repo." + "reason": "Missing config field in StableLm vs loading failure for a different model package." }, { "left": "issue:43873", "right": "issue:44464", "accept": false, - "reason": "Quantization offloading behavior is unrelated to chunked generation with compiled forward." + "reason": "Quantization/offloading behavior vs compiled-forward chunked generation outputs; separate problems." }, { "left": "issue:43572", "right": "issue:43957", "accept": false, - "reason": "Shared v5 breakage theme only; config field omission and meta-device loading are different bugs." + "reason": "Config regression vs meta-device model loading failure; only broad version overlap." }, { "left": "issue:43295", "right": "issue:43450", "accept": false, - "reason": "Processor/tokenizer regression is unrelated to batched video shape handling." + "reason": "Processor/tokenizer regression vs batched video shape bug; different APIs and failure modes." }, { "left": "issue:43824", "right": "issue:43975", "accept": false, - "reason": "Different symptoms and code paths: missing exported class import vs incorrect detokenization." + "reason": "Missing exported class ImportError vs incorrect detokenization in v5; not the same issue." }, { "left": "issue:43844", "right": "issue:44936", "accept": false, - "reason": "Gradient explosion in DeepSpeed ZeRO-3 is unrelated to trainer.evaluate() state after train()." + "reason": "ZeRO-3 gradient explosion vs trainer.evaluate() state issue after training; unrelated training bugs." }, { "left": "issue:43295", "right": "issue:43873", "accept": false, - "reason": "Processor API regression and quantization/offloading behavior are unrelated." + "reason": "Processor regression vs quantization/offloading behavior; no shared root cause." }, { "left": "issue:44514", "right": "issue:44521", "accept": false, - "reason": "Both are chat-template related, but one is a padding=False crash on batched inputs while the other is all-zero assistant masks for multimodal prompts." + "reason": "Both involve apply_chat_template, but one is a batched crash with tool calls and the other is all-zero assistant masks; too different to merge confidently." }, { "left": "issue:38175", "right": "issue:45412", "accept": false, - "reason": "Different models and problems: SigLIP2 output issue vs RT-DETR memory cleanup." + "reason": "SigLIP2 output bug vs RT-DETR memory leak; completely different subsystems." }, { "left": "issue:43278", "right": "issue:43856", "accept": false, - "reason": "Embedding dtype drift in eval is unrelated to Qwen3 MoE training memory usage." + "reason": "Dtype drift in evaluation vs Qwen3 MoE memory inefficiency; no common bug." }, { "left": "issue:43295", "right": "issue:43994", "accept": false, - "reason": "Custom processor/tokenizer regression is not the same as SigLIP2 inference mismatch." + "reason": "Custom processor/tokenizer regression vs SigLIP2 inference quality issue; different models and code paths." }, { "left": "issue:44442", "right": "issue:45198", "accept": false, - "reason": "Different tokenizer families and failures: FastSpeech2Conformer loading vs Wav2Vec2 save/tokenization." + "reason": "Both are tokenizer-related, but for different model families and distinct save/load failures." }, { "left": "issue:37428", "right": "issue:43824", "accept": false, - "reason": "Both are import errors, but they affect different symbols and modules." + "reason": "Both are ImportErrors, but for different missing symbols in unrelated model modules." }, { "left": "issue:38175", "right": "issue:43493", - "accept": false, - "reason": "Both concern SigLIP2, but one is zero probabilities in pipeline/model output and the other is divergence from original JAX implementation." + "accept": true, + "reason": "Both are SigLIP2 correctness reports; the broader parity/discrepancy issue plausibly subsumes the zero-probabilities symptom." }, { "left": "issue:42915", "right": "issue:43824", "accept": false, - "reason": "Qwen3MoE FP8 config failure is unrelated to a missing top-level import symbol." + "reason": "Qwen3Moe FP8 failure vs missing Qwen2.5-VL export; different models and problems." }, { "left": "issue:43827", "right": "issue:43901", "accept": false, - "reason": "Both are docs/pipeline API topics, but one is pipeline removal fallout and the other is a specific TextClassificationPipeline doc mismatch." + "reason": "Both are docs-ish around pipeline behavior, but they cover different tasks and different regressions." }, { "left": "issue:43653", "right": "issue:44112", "accept": false, - "reason": "Tokenizer special-token registration bug is unrelated to the GraniteSpeech CI device-override test." + "reason": "Tokenizer special-token registration bug vs GraniteSpeech CI device override test failure." }, { "left": "issue:43450", "right": "issue:43493", "accept": false, - "reason": "Video batch shape bug is unrelated to SigLIP2 implementation discrepancy." + "reason": "Batched video processor shape issue vs SigLIP2 implementation discrepancy; unrelated areas." }, { - "left": "issue:44360", - "right": "issue:44561", + "left": "issue:43450", + "right": "issue:43575", "accept": false, - "reason": "DSA indexer discussion is unrelated to trust_remote_code breakage from removing is_torch_fx_available." + "reason": "Video batch-shape bug vs tensor-parallel OOM on loading a large model; no shared root cause." }, { - "left": "issue:43450", - "right": "issue:43575", + "left": "issue:44360", + "right": "issue:44561", "accept": false, - "reason": "Video processor shape handling is unrelated to Qwen2-57B-A14B-Instruct OOM under tensor parallel loading." + "reason": "Discussion about a DSA indexer activation vs removal of is_torch_fx_available breaking trust_remote_code models." }, { "left": "issue:44792", "right": "issue:45563", "accept": false, - "reason": "Janus image-generation test failure is unrelated to a stale warning in paged generate()." + "reason": "Janus image-generation test failure vs a stale generate() warning; different severity and code paths." }, { "left": "issue:45127", "right": "issue:45245", "accept": false, - "reason": "LoRA/vocabulary merge collapse is unrelated to the categories upper-bound runtime error." + "reason": "LoRA merge collapse with extended vocab vs category-count runtime limit; unrelated." }, { "left": "issue:43819", "right": "issue:45083", "accept": false, - "reason": "DAC STE mismatch and qwen3_omni_moe feature-length helper behavior are different code paths." + "reason": "DAC latent/STE mismatch vs Qwen3 Omni feature-length helper behavior; different model families." }, { "left": "issue:44206", "right": "issue:45584", "accept": false, - "reason": "Feature extractor center-arg crash and Whisper empty-transcription failure are unrelated processor bugs." + "reason": "Feature extractor center-arg regression vs Whisper empty-transcription failure after token alignment." }, { "left": "issue:43262", "right": "issue:43450", "accept": false, - "reason": "Audio chat-template sampling-rate defaulting is unrelated to batched video processor shapes." + "reason": "Audio apply_chat_template sampling-rate bug vs video processor batched-shape bug; not the same issue." }, { "left": "issue:43819", "right": "issue:44112", "accept": false, - "reason": "Different systems entirely: DAC latent decoding vs GraniteSpeech CI device override." + "reason": "DAC forward/from_latents mismatch vs GraniteSpeech CI device override failure; unrelated." }, { "left": "issue:44610", "right": "issue:44857", "accept": false, - "reason": "OmDet-Turbo image size mismatch is unrelated to LwDetrImageLoss AMP/CUDA crash." + "reason": "OmDet processor size mismatch vs LwDetrImageLoss AMP/CUDA crash; separate defects." }, { "left": "issue:37428", "right": "issue:43425", "accept": false, - "reason": "ImportError for a flash-attention helper is unrelated to Torch 2.10 compatibility." + "reason": "Specific flash-attention import failure vs a broad torch compatibility complaint; not the same bug." }, { "left": "issue:44857", "right": "issue:45593", "accept": false, - "reason": "Both are training-loss bugs, but one is AMP/CUDA crash and the other is missing auxiliary losses in D-FINE." + "reason": "AMP/CUDA loss crash vs missing auxiliary losses when denoising is off." }, { "left": "issue:45290", "right": "issue:45381", "accept": false, - "reason": "Both involve chat/template or multimodal inputs, but one crashes on assistant tool-call messages and the other reports wrong vision_position_ids." + "reason": "Assistant-message tool-call crash in chat templating vs Qwen2.5-VL video vision_position_ids bug." }, { "left": "issue:43278", "right": "issue:43825", "accept": false, - "reason": "Embedding dtype regression is unrelated to a pipeline error message suggesting unsupported translation tasks." + "reason": "Embedding dtype regression vs pipeline docs/error-message mismatch; different layers of the stack." } ] }, - "evaluator_result": null, - "evaluator_used": false, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet and stays conservative: it correctly treats the set as mostly heterogeneous, and the single accepted soft edge (38175 \u2194 43493) is reasonably supported by the two SigLIP2 titles. The rejected pairs are also framed cautiously based on distinct model families, APIs, or failure modes." + }, + "evaluator_used": true, "retried": false, - "accepted_nontrivial_soft_edge": false, + "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null }, @@ -7143,12 +7074,12 @@ "issue:43606|issue:45083", "issue:44448|issue:44743", "issue:43335|issue:43550", - "issue:44561|issue:44779", "issue:37428|issue:44351", + "issue:44561|issue:44779", "issue:44462|issue:44561", "issue:43611|issue:43931", - "issue:43388|issue:43526", "issue:39692|issue:43526", + "issue:43388|issue:43526", "issue:43819|issue:44493", "issue:43335|issue:43653", "issue:29942|issue:32090", @@ -7176,14 +7107,14 @@ "issue:40444|issue:43526", "issue:44857|issue:45072" ], - "prepared_review_unit_hash": "f685b6d7f7ef5f7432b4b46875a6f8ca4a9c36644f642e92a718d4fb58441a68", + "prepared_review_unit_hash": "d115817b0b9cd2460692970ac1afe0012d58e31dfac6ec47e0d70ffcf643f5b7", "component_budget": { "node_count": 387, "item_count": 387, "soft_pair_count": 1455, - "serialized_chars": 749552, - "estimated_input_tokens": 187388, - "estimated_eval_tokens": 375032 + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 }, "budget": { "node_count": 48, @@ -7198,234 +7129,234 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Only one soft duplicate is convincing: the two identical Qwen3.5 MoE from_pretrained error reports. The rest are distinct issues across different models, APIs, or failure modes.", - "confidence": 0.94, - "canonical_issue_reason": "Issue 45310 is the better canonical for the Qwen3.5 MoE from_pretrained failure: it appeared first, has the same exact title as 45478, and looks like the fuller original report; 45478 is a duplicate re-report.", + "summary": "The set is mostly a broad false-positive similarity cluster: most issues share subsystem keywords but describe different bugs. The only clear duplicate in the soft edges is the pair of identical Qwen3.5 MoE from_pretrained reports.", + "confidence": 0.96, + "canonical_issue_reason": "issue:45310 is the earlier, fuller report of the same Qwen3.5 MoE from_pretrained failure; issue:45478 is a near-identical duplicate.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 45310 is the best issue to keep as the representative because it most clearly names the concrete bug and predates the duplicate report.", + "best_issue_reason": "issue:45310 is the strongest representative: it is specific, reproducible, and clearly states the concrete Qwen3.5 MoE loading failure that the duplicate mirrors.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45200", "right": "issue:45588", "accept": false, - "reason": "Different bugs: Gemma 4 text-only fine-tuning token-type defaults vs a flash_attention.py crash on sink-less models." + "reason": "Different failures: Gemma 4 finetuning needs mm_token_type_ids; flash_attention.py crashes on s_aux=None." }, { "left": "issue:43606", "right": "issue:45083", "accept": false, - "reason": "Different components and failures: Bark CPU offload device mismatch vs a qwen3_omni_moe feature-length helper behavior." + "reason": "Different code paths and symptoms: CPU offload device mismatch vs a feature-extractor helper behavior issue." }, { "left": "issue:44448", "right": "issue:44743", "accept": false, - "reason": "Different model behaviors: Pegasus output mismatch across versions vs Qwen3.5. modular recurrent-state reset with cache." + "reason": "Both mention generation behavior, but one is Pegasus output drift and the other is Qwen3.5 recurrent state resetting." }, { "left": "issue:43335", "right": "issue:43550", "accept": false, - "reason": "Different code paths: SwitchTransformers sparse-layer config bug vs Bamba torch.compile + SDPA failure." + "reason": "Unrelated bugs: SwitchTransformers config creates an unwanted sparse layer; Bamba-9B-v2 fails under torch.compile with SDPA." }, { - "left": "issue:44561", - "right": "issue:44779", + "left": "issue:37428", + "right": "issue:44351", "accept": false, - "reason": "Unrelated regressions: trust_remote_code breakage from a removed utility vs DeepSeek tokenizer output changes." + "reason": "Both are import errors, but for different missing symbols and different root causes." }, { - "left": "issue:37428", - "right": "issue:44351", + "left": "issue:44561", + "right": "issue:44779", "accept": false, - "reason": "Both are import errors, but for different symbols and subsystems (_flash_supports_window_size vs HybridCache)." + "reason": "Different regressions: removal of is_torch_fx_available vs DeepSeek tokenizer output changes in v5." }, { "left": "issue:44462", "right": "issue:44561", "accept": false, - "reason": "AutoTokenizer ignoring tokenizer.json is a tokenizer-loading issue, not the v5 trust_remote_code utility removal breakage." + "reason": "AutoTokenizer ignoring tokenizer.json is unrelated to the v5 trust_remote_code import breakage." }, { "left": "issue:43611", "right": "issue:43931", "accept": false, - "reason": "General model-loading breakage vs a specific Qwen3-VL weight-shape mismatch; not the same underlying defect." + "reason": "Different loading problems: base_model_prefix regression vs Qwen3-VL weight-shape mismatch." }, { - "left": "issue:43388", + "left": "issue:39692", "right": "issue:43526", "accept": false, - "reason": "Different tensor/label bugs: gather_for_metrics truncation vs BeitImageProcessorFast reduce_labels returning one label." + "reason": "Documentation/example mistakes are not the same underlying bug as BeitImageProcessorFast reduce_labels returning the wrong labels." }, { - "left": "issue:39692", + "left": "issue:43388", "right": "issue:43526", "accept": false, - "reason": "Documentation/example errors vs a fast image processor label handling bug." + "reason": "Different data-handling bugs: gather_for_metrics dropping tuple labels vs reduce_labels truncating label arrays." }, { "left": "issue:43819", "right": "issue:44493", "accept": false, - "reason": "Different failures: DAC latent/forward mismatch vs unexpected position-id keys across many models." + "reason": "DAC from_latents/STE mismatch is unrelated to position-id key warnings across many models." }, { "left": "issue:43335", "right": "issue:43653", "accept": false, - "reason": "Different model families and symptoms: SwitchTransformers sparse-layer config vs BigBirdTokenizer special-token registration." + "reason": "Different tokenizer/config areas and different concrete failures." }, { "left": "issue:29942", "right": "issue:32090", "accept": false, - "reason": "Unrelated runtime problems: FlashAttention 2 test failures vs Trainer _gpu_broadcast_one NoneType handling." + "reason": "Flash Attention 2 test failures and _gpu_broadcast_one NoneType errors are separate issues." }, { "left": "issue:44855", "right": "issue:45083", "accept": false, - "reason": "Python 3.13 import/parsing issue in DebertaV2Model vs helper output-length logic in qwen3_omni_moe." + "reason": "Python 3.13 import parsing failure in DebertaV2Model is unrelated to the qwen3_omni_moe helper bug." }, { "left": "issue:43404", "right": "issue:45200", "accept": false, - "reason": "Distinct generation/tie-weight bug vs Gemma 4 multimodal token-type defaulting." + "reason": "lm_head tying bug in Mistral3ForConditionalGeneration is distinct from Gemma 4 token-type-id handling." }, { "left": "issue:45310", "right": "issue:45478", "accept": true, - "reason": "Identical title and same concrete Qwen3.5 MoE from_pretrained regression on transformers>=5.4.0; 45478 is a duplicate report." + "reason": "Near-identical title and failure mode: same Qwen3.5 MoE from_pretrained error, with 45478 looking like a duplicate follow-up." }, { "left": "issue:39692", "right": "issue:44493", "accept": false, - "reason": "SigLIP2 documentation/example issues are not the same as broad position-id key warnings." + "reason": "Doc/example errors are not the same as the broad unexpected-key-position-id regression." }, { "left": "issue:43550", "right": "issue:44112", "accept": false, - "reason": "Different CI/runtime failures: Bamba SDPA torch.compile vs GraniteSpeech stale device override test." + "reason": "Bamba compile/SDPA failure is unrelated to a stale GraniteSpeech CI test." }, { "left": "issue:43335", "right": "issue:45083", "accept": false, - "reason": "Different bugs in different models: SwitchTransformers layer creation vs qwen3_omni_moe feature-length computation." + "reason": "SwitchTransformers sparse-layer construction bug does not match the qwen3_omni_moe output-length helper bug." }, { "left": "issue:43335", "right": "issue:43504", "accept": false, - "reason": "SwitchTransformers sparse-layer creation is unrelated to BEiT pretrained preset loading with a legacy field." + "reason": "Sparse-layer creation and BEiT legacy-field loading are different model-specific bugs." }, { "left": "issue:44292", "right": "issue:45520", "accept": false, - "reason": "Different Qwen-related failures with different causes: NVFP4 runtime error vs missing flash_attn import on Python 3.13." + "reason": "Qwen-3-8B-NVFP4 runtime error and flash_attn import KeyError are different failure classes." }, { "left": "issue:44448", "right": "issue:44779", "accept": false, - "reason": "Different symptoms and models: Pegasus output drift vs DeepSeek tokenizer correctness on v5." + "reason": "Pegasus output differences and DeepSeek tokenizer regressions do not share the same code path." }, { "left": "issue:38175", "right": "issue:44493", "accept": false, - "reason": "SigLIP2 zero-probability issue is unrelated to unexpected position-id key warnings." + "reason": "Zero probabilities in SigLIP2 are unrelated to unexpected position-id keys." }, { "left": "issue:28218", "right": "issue:35141", "accept": false, - "reason": "Tokenizer adds extra spacing after added tokens, while the other issue is about output embeddings being reinitialized on resize." + "reason": "Tokenizer spacing after added tokens is unrelated to embedding reinitialization after resize." }, { "left": "issue:43335", "right": "issue:45072", "accept": false, - "reason": "Different bfloat16/ dtype mismatch failures in different models, not the same bug." + "reason": "SwitchTransformers sparse-layer config bug is unrelated to dtype mismatches in bfloat16 inference." }, { "left": "issue:44977", "right": "issue:45229", "accept": false, - "reason": "Different Qwen/Gemma issues: flash-attention generation problems vs multi-GPU CUDA OOM." + "reason": "Qwen3.5 flash-attention generation bug is unrelated to Gemma4 multi-GPU OOM." }, { "left": "issue:43335", "right": "issue:45593", "accept": false, - "reason": "SwitchTransformers sparse-layer config bug is unrelated to D-FINE auxiliary-loss behavior." + "reason": "Sparse-layer creation is unrelated to D-FINE auxiliary-loss behavior when denoising is off." }, { "left": "issue:43720", "right": "issue:44112", "accept": false, - "reason": "BitNet packed-weight loading bug is a different failure mode from GraniteSpeech device-override CI flakiness." + "reason": "Packed-weight loading in BitNet is unrelated to the GraniteSpeech device-override CI failure." }, { "left": "issue:44448", "right": "issue:44977", "accept": false, - "reason": "Different models and root causes: Pegasus v4/v5 output mismatch vs Qwen3.5 flash-attention generation failure." + "reason": "Different models and symptoms: Pegasus output drift vs Qwen3.5 flash-attention generation failure." }, { "left": "issue:45459", "right": "issue:45712", "accept": false, - "reason": "Tokenizer exception handling hiding real errors is unrelated to leftover dummy classes and repo checks." + "reason": "Tokenizer exception masking is unrelated to leftover dummy classes in dummy_pt_objects.py." }, { "left": "issue:44568", "right": "issue:45491", "accept": false, - "reason": "Tokenizer add_special_tokens behavior is unrelated to Gemma3 NaN embeddings with sliding-window attention." + "reason": "Missing BOS/EOS with add_special_tokens and NaN embeddings from sliding-window attention are distinct bugs." }, { "left": "issue:45198", "right": "issue:45803", "accept": false, - "reason": "Wav2Vec2 save/tokenization failure is a different bug than a bare except swallowing errors in FuyuBatchFeature." + "reason": "Wav2Vec2 save/tokenization failure is unrelated to the bare-except FuyuBatchFeature issue." }, { "left": "issue:43606", "right": "issue:44857", "accept": false, - "reason": "Different models and failure contexts: CPU offload device mismatch vs float16 AMP crash in LwDetrImageLoss." + "reason": "CPU offload device mismatch and float16 AMP crashes in LwDetrImageLoss are different problems." }, { "left": "issue:29127", "right": "issue:32090", "accept": false, - "reason": "LayoutLMv3 error-message clarity is unrelated to the Trainer NoneType broadcast exception." + "reason": "A clearer LayoutLMV3 error message request is not the same as the Trainer _gpu_broadcast_one NoneType crash." }, { "left": "issue:40444", "right": "issue:43526", "accept": false, - "reason": "Qwen2.5-VL iterable dataset / multi-image prompt failure is unrelated to BeitImageProcessorFast label reduction." + "reason": "Multi-image IterableDataset finetuning failure is unrelated to BeitImageProcessorFast reduce_labels returning only one label." }, { "left": "issue:44857", "right": "issue:45072", "accept": false, - "reason": "Different numerical stability/runtime issues in different models, not the same underlying bug." + "reason": "AMP/CUDA crash in LwDetrImageLoss is unrelated to SwitchTransformers/TimmWrapperModel bfloat16 dtype mismatches." } ] }, "evaluator_result": { "accept": true, - "feedback": "The output is grounded in the packet and stays conservative overall. The only accepted soft duplicate is the exact-title Qwen3.5 MoE from_pretrained pair, which is well supported by the data. The remaining pair rejections are phrased cautiously and align with the distinct titles/subsystems shown in the packet." + "feedback": "The clustering is conservative and grounded in the packet. The only accepted soft-edge pair, issue:45310 and issue:45478, is strongly supported by identical titles and highly similar metadata. The remaining false verdicts are reasonably cautious and do not overstate sameness." }, "evaluator_used": true, "retried": false, @@ -7507,8 +7438,8 @@ "issue:45245|issue:45563", "issue:43873|issue:44493", "issue:29942|issue:35141", - "issue:45372|issue:45459", "issue:36010|issue:37428", + "issue:45372|issue:45459", "issue:43278|issue:43381", "issue:25251|issue:32090", "issue:44448|issue:44466", @@ -7524,217 +7455,210 @@ "issue:44871|issue:44977", "issue:44265|issue:45198", "issue:43421|issue:45491", - "issue:44206|issue:44479", - "issue:43065|issue:43994" + "issue:44206|issue:44479" ], - "prepared_review_unit_hash": "d9df7bbde41786b7761cc618c806864e8062f8765795bada5a99e6617743060f", + "prepared_review_unit_hash": "245300eca58a1f229b1d94efd95a9bb039d295d99cf8147c4d44154bc8fe0db5", "component_budget": { "node_count": 387, "item_count": 387, "soft_pair_count": 1455, - "serialized_chars": 749552, - "estimated_input_tokens": 187388, - "estimated_eval_tokens": 375032 + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 30, - "serialized_chars": 36679, - "estimated_input_tokens": 9170, - "estimated_eval_tokens": 18596 + "soft_pair_count": 29, + "serialized_chars": 36372, + "estimated_input_tokens": 9093, + "estimated_eval_tokens": 18442 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This set is heterogeneous: tokenizer, loading, export, flash-attention, quantization, and runtime issues mixed together. None of the soft pairs look like true duplicates or mergeable change-equivalents. Issue 43873 is the broadest representative, but the cluster is not a tight duplicate family.", - "confidence": 0.46, - "canonical_issue_reason": "Issue 43873 is the broadest, most active open report here and works best as a representative anchor, though it does not actually duplicate the other items.", + "summary": "This cluster is noisy: the soft edges connect clearly unrelated issues across tokenization, generation, export, quantization, memory, and model-loading behavior. No soft pair looks like the same underlying bug; issue 45588 is only the most central representative, not a true duplicate target.", + "confidence": 0.81, + "canonical_issue_reason": "Issue 45588 is the most central and activity-rich node in this otherwise heterogeneous cluster, with multiple nearby soft links around the same flash-attention crash theme.", "canonical_pr_reason": null, - "best_issue_reason": "43873 has the widest scope and strongest discussion/inbound-ref signal among the issues, making it the best single issue representative of this mixed set.", + "best_issue_reason": "Issue 45588 is the best representative issue here because it is the densest hub in the cluster and describes a concrete runtime failure, even though the cluster does not form a real duplicate set.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:34689", "right": "issue:35141", "accept": false, - "reason": "Model-loading regression vs embedding reinit after resize; different code paths and symptoms." + "reason": "Model-loading regression for Llama 3.2 Vision vs embedding reinit on resize; unrelated failures." }, { "left": "issue:43701", "right": "issue:45083", "accept": false, - "reason": "Checkpoint key mismatch vs a qwen3_omni_moe helper-function behavior issue; unrelated bugs." + "reason": "Checkpoint key mismatch vs helper output-length behavior in qwen3_omni_moe; different bug surfaces." }, { "left": "issue:45372", "right": "issue:45588", "accept": false, - "reason": "Missing ReasoningEffort import vs flash_attention s_aux=None crash; not the same failure." + "reason": "Gemma 4 processor import failure due to mistral_common vs flash_attention s_aux=None crash; different code paths." }, { "left": "issue:43927", "right": "issue:45698", "accept": false, - "reason": "Save/load config field loss vs wrong custom module resolution after save_pretrained; related lifecycle, but distinct bugs." + "reason": "Saved custom token IDs lost on load vs wrong custom module loaded after save_pretrained; related area but distinct bugs." }, { "left": "issue:44977", "right": "issue:45563", "accept": false, - "reason": "Flash-attention generation regression vs stale warning in paged generate; different behavior." + "reason": "Qwen3.5 flash-attention generation regression vs stale num_return_sequences warning in paged generate(); not the same issue." }, { "left": "issue:43550", "right": "issue:45083", "accept": false, - "reason": "torch.compile/SDPA failure vs _get_feat_extract_output_lengths behavior; unrelated." + "reason": "Bamba torch.compile + SDPA failure vs qwen3_omni_moe feature-length helper behavior; unrelated subsystems." }, { "left": "issue:43412", "right": "issue:45588", "accept": false, - "reason": "ExecuTorch export guard/erfinv support issue vs flash_attention attribute crash; different code paths." + "reason": "Gemma3 executorch export guard/erfinv support issue vs sink-less flash_attention AttributeError; different failures." }, { "left": "issue:43066", "right": "issue:44462", "accept": false, - "reason": "Tokenizer decoder type mismatch vs AutoTokenizer ignoring tokenizer.json; same area, but different bug." + "reason": "Tokenizer decoder type mismatch in v5 vs AutoTokenizer ignoring tokenizer.json; both tokenizer-related but different bugs." }, { "left": "issue:45245", "right": "issue:45563", "accept": false, - "reason": "Category-limit RuntimeError vs stale warning text; unrelated." + "reason": "Category-count runtime error vs paged generate warning; not the same underlying problem." }, { "left": "issue:43873", "right": "issue:44493", "accept": false, - "reason": "Quantization/offloading problem vs unexpected position-id key warning; not the same issue." + "reason": "Quantization offloading issue vs position-id key warning across many models; different behavior." }, { "left": "issue:29942", "right": "issue:35141", "accept": false, - "reason": "Flash Attention 2 test failures vs embedding reinitialization after resize; no shared bug." + "reason": "Flash Attention 2 test failures vs embedding reinitialization after resize; different issues." }, { - "left": "issue:45372", - "right": "issue:45459", + "left": "issue:36010", + "right": "issue:37428", "accept": false, - "reason": "ImportError for ReasoningEffort vs protobuf-error masking in tokenizer code; distinct import/error-handling bugs." + "reason": "GenerationMixin import error vs _flash_supports_window_size import error; both imports but unrelated symbols and modules." }, { - "left": "issue:36010", - "right": "issue:37428", + "left": "issue:45372", + "right": "issue:45459", "accept": false, - "reason": "Both are import errors, but they involve different missing symbols and different modules/code paths." + "reason": "Gemma 4 processor import dependency issue vs protobuf exception masking tokenizer errors; distinct import/error handling bugs." }, { "left": "issue:43278", "right": "issue:43381", "accept": false, - "reason": "Eval-time dtype drift vs gradient-checkpointing being unsupported in eval; different problems." + "reason": "Embedding dtype drift in eval vs gradient checkpointing disallowed in eval mode; separate training/eval regressions." }, { "left": "issue:25251", "right": "issue:32090", "accept": false, - "reason": "Pipeline top_k nesting bug vs Trainer NoneType broadcast error; unrelated." + "reason": "top_k pipeline output nesting change vs Trainer _gpu_broadcast_one NoneType crash; unrelated." }, { "left": "issue:44448", "right": "issue:44466", "accept": false, - "reason": "Different output between v4 and v5 vs serialization of tied weights depending on device; not the same concrete bug." + "reason": "Pegasus output mismatch across v4/v5 vs lm_head.weight serialization inconsistency; different model behavior." }, { "left": "issue:44857", "right": "issue:45588", "accept": false, - "reason": "AMP/CUDA crash in image loss vs flash_attention s_aux=None AttributeError; unrelated." + "reason": "LwDetrImageLoss AMP/CUDA crash vs flash_attention s_aux=None AttributeError; different components." }, { "left": "issue:43065", "right": "issue:45412", "accept": false, - "reason": "Dummy Conv2d placeholder issue vs memory not released on delete; different symptoms and fixes." + "reason": "Dummy Conv2d in Sam3PixelDecoder vs RT-DETR memory not released; unrelated model internals." }, { "left": "issue:43653", "right": "issue:44743", "accept": false, - "reason": "Tokenizer special-token registration bug vs recurrent state reset with cache; different subsystems." + "reason": "BigBirdTokenizer special-token registration bug vs Qwen3.5 recurrent state reset with cache; different tokenizer/model logic." }, { "left": "issue:44451", "right": "issue:45520", "accept": false, - "reason": "Model loading failure vs Python 3.13 flash_attn import KeyError; unrelated." + "reason": "ScandiBERT loading failure vs Python 3.13 flash_attn import KeyError; unrelated import/load paths." }, { "left": "issue:44186", "right": "issue:45198", "accept": false, - "reason": "LayoutLMv2 tokenizer NER/padding crash vs Wav2Vec2 save_pretrained/tokenization failure; both tokenizer-related but not the same bug." - }, - { - "left": "issue:44265", - "right": "issue:45198", - "accept": false, - "reason": "torch.export check failure vs save_pretrained/tokenization failure; unrelated code paths." + "reason": "LayoutLMv2Tokenizer NER padding/truncation crash vs Wav2Vec2 save_pretrained/tokenization failure; different tokenizers and symptoms." }, { "left": "issue:43976", "right": "issue:45440", "accept": false, - "reason": "Python-version compatibility bug vs model divergence from remote implementation; unrelated." + "reason": "Python 3.9/3.10 compatibility issue vs DeepseekV3MoE divergence from remote implementation; unrelated." }, { "left": "issue:43994", "right": "issue:45412", "accept": false, - "reason": "AutoModel/pipeline nonsensical output vs RT-DETR memory release issue; different failures." + "reason": "SigLIP2 nonsensical outputs with AutoModel/pipeline vs RT-DETR memory leak; different model families and bugs." }, { "left": "issue:44361", "right": "issue:45198", "accept": false, - "reason": "MLukeTokenizer AttributeError vs Wav2Vec2 save_pretrained/tokenization failure; not the same issue." + "reason": "MLukeTokenizer AttributeError vs Wav2Vec2 save/tokenization failure; both tokenizer-related but not the same defect." }, { "left": "issue:42915", "right": "issue:42994", "accept": false, - "reason": "Quantized inference/config failure vs quantized model saving failure; same broad domain, but different concrete bugs." + "reason": "Qwen3Moe FineGrainedFP8Config failure vs quantized model saving failure; quantization-adjacent but separate problems." }, { "left": "issue:44871", "right": "issue:44977", "accept": false, - "reason": "eos_token_id config mismatch vs flash-attention generation regression; unrelated." + "reason": "Gemma-3 eos_token_id mismatch vs Qwen3.5 flash-attention generation regression; different model/config issues." + }, + { + "left": "issue:44265", + "right": "issue:45198", + "accept": false, + "reason": "torch.export.export failure with torch_compilable_check vs Wav2Vec2 save/tokenization failure; unrelated export vs tokenizer bug." }, { "left": "issue:43421", "right": "issue:45491", "accept": false, - "reason": "Runtime tokenizer post-processor update feature request vs Gemma3 NaN embeddings on GPU; different topics." + "reason": "Runtime special-token post-processor update feature request vs Gemma3 NaN embeddings with sliding-window attention; different concerns." }, { "left": "issue:44206", "right": "issue:44479", "accept": false, - "reason": "Unsupported feature-extractor arg vs video-input regression in Qwen VL models; unrelated." - }, - { - "left": "issue:43065", - "right": "issue:43994", - "accept": false, - "reason": "Dummy Conv2d placeholder vs model-output mismatch; different bugs." + "reason": "LasrFeatureExtractor unsupported center arg vs Qwen v5.3.0 video input regression; different input-processing bugs." } ] }, @@ -7768,6 +7692,7 @@ "issue:39692", "issue:41720", "issue:42915", + "issue:43065", "issue:43262", "issue:43278", "issue:43329", @@ -7785,6 +7710,7 @@ "issue:43824", "issue:43873", "issue:43931", + "issue:43994", "issue:44186", "issue:44206", "issue:44279", @@ -7792,8 +7718,6 @@ "issue:44554", "issue:44610", "issue:44734", - "issue:44743", - "issue:44811", "issue:44857", "issue:44977", "issue:45072", @@ -7809,6 +7733,7 @@ ], "soft_pairs": [ "issue:29942|issue:30064", + "issue:43065|issue:43994", "issue:44734|issue:45406", "issue:33357|issue:39692", "issue:29127|issue:34567", @@ -7831,229 +7756,228 @@ "issue:44977|issue:45245", "issue:43824|issue:44279", "issue:43526|issue:44857", - "issue:42915|issue:43931", "issue:43526|issue:44610", + "issue:42915|issue:43931", "issue:43262|issue:44206", - "issue:43441|issue:45588", "issue:43452|issue:45588", + "issue:43441|issue:45588", "issue:44442|issue:45072", - "issue:44743|issue:44811", "issue:43575|issue:43701" ], - "prepared_review_unit_hash": "f1a8f439ca48e3be6c6d42e91cc82008e5468976ad82d5009ee3916cc2376062", + "prepared_review_unit_hash": "743184a7cab918dd86b88f093dbc0b190c2a59726c01fa8742b30715f83056cd", "component_budget": { "node_count": 387, "item_count": 387, "soft_pair_count": 1455, - "serialized_chars": 749552, - "estimated_input_tokens": 187388, - "estimated_eval_tokens": 375032 + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 31, - "serialized_chars": 37028, - "estimated_input_tokens": 9257, - "estimated_eval_tokens": 18770 + "serialized_chars": 36979, + "estimated_input_tokens": 9245, + "estimated_eval_tokens": 18746 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These issues are mostly unrelated reports across transformers subsystems; I don\u2019t see a true duplicate cluster. All soft edges look like broad thematic similarity rather than the same underlying bug.", - "confidence": 0.91, - "canonical_issue_reason": "If a representative issue must be chosen, issue 45372 is the clearest and most actionable report: it has a concrete import failure, version scope, and downstream processor-loading impact.", + "summary": "These items are a loose similarity cluster, not a true duplicate set: they span many unrelated bugs across tokenizers, processors, training, serving, quantization, and model-specific inference issues. I don\u2019t see a single canonical issue or PR for the whole cluster.", + "confidence": 0.96, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "Issue 45372 is the strongest standalone issue in this set because it is specific, reproducible, and clearly explains a user-facing breakage path.", + "best_issue_reason": null, "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:29942", "right": "issue:30064", "accept": false, - "reason": "Different failures in unrelated areas: Flash Attention 2 test instability vs image processor void segmentation handling." + "reason": "Unrelated failures: FlashAttention 2 tests vs void segmentation-map processing." + }, + { + "left": "issue:43065", + "right": "issue:43994", + "accept": false, + "reason": "Different vision-model bugs: dummy Conv2d in Sam3PixelDecoder vs bad SigLIP2 outputs." }, { "left": "issue:44734", "right": "issue:45406", "accept": false, - "reason": "Both mention transformers serve, but one is KV-cache tensor indexing and the other is a missing Gemma4Processor attribute." + "reason": "Both touch serve/processor code, but one is KV-cache tensor indexing and the other is missing `_tokenizer` on Gemma4Processor." }, { "left": "issue:33357", "right": "issue:39692", "accept": false, - "reason": "Different model families and bugs: MacOS bus error on CLIP vs SigLIP2 doc/example errors." + "reason": "Different problems: MacOS bus error for community CLIP vs documentation example errors for SigLIP2." }, { "left": "issue:29127", "right": "issue:34567", "accept": false, - "reason": "LayoutLMv3 error messaging vs TrainerState token counting update are unrelated." + "reason": "LayoutLMv3 error-message clarity is unrelated to TrainerState token-count updating." }, { "left": "issue:44610", "right": "issue:45412", "accept": false, - "reason": "Input resolution mismatch for OmDet-Turbo is unrelated to RT-DETR memory release behavior." + "reason": "Different object-detection issues: wrong processor image size vs RT-DETR memory not being released." }, { "left": "issue:45103", "right": "issue:45372", "accept": false, - "reason": "auto_docstring annotation handling and Gemma4 processor import failure are different code paths and symptoms." + "reason": "auto_docstring future-annotations crash is unrelated to Gemma4 loading failure from missing ReasoningEffort." }, { "left": "issue:43873", "right": "issue:45229", "accept": false, - "reason": "Both touch memory/quantization, but one is offloading behavior and the other is multi-GPU OOM for a specific model." + "reason": "Both involve memory/quantization broadly, but offloading behavior and Gemma4 multi-GPU OOM are not the same bug." }, { "left": "issue:43756", "right": "issue:45440", "accept": false, - "reason": "Both are model-implementation divergence reports, but they concern different models and different defects." + "reason": "Different model-implementation divergences in unrelated architectures." }, { "left": "issue:45372", "right": "issue:45406", "accept": false, - "reason": "Same broader Gemma4 area, but one is a mistral_common import dependency issue and the other is a missing _tokenizer attribute in serve." + "reason": "Both mention Gemma4, but one is an import/dependency issue and the other is a serve-time `_tokenizer` attribute crash." }, { "left": "issue:43723", "right": "issue:44554", "accept": false, - "reason": "Tokenizer loading failure and MPS attention correctness are unrelated." + "reason": "Tokenizer loading in v5 is unrelated to the MPS attention correctness issue." }, { "left": "issue:43278", "right": "issue:43575", "accept": false, - "reason": "BF16-to-FP32 dtype drift during eval is unrelated to tensor-parallel OOM when loading Qwen2-57B." + "reason": "BF16/FP32 eval dtype change is unrelated to Qwen2 TP loading OOM." }, { "left": "issue:43381", "right": "issue:43701", "accept": false, - "reason": "Eval-mode gradient checkpointing restriction and resume_from_checkpoint key mismatch are distinct training issues." + "reason": "Gradient checkpointing in eval mode is unrelated to checkpoint resume key mismatches." }, { "left": "issue:43278", "right": "issue:43329", "accept": false, - "reason": "Different bug classes: dtype change in eval vs undefined variables in multimodal token counting video branch." + "reason": "Different bugs: embedding dtype regression vs multimodal token counting video-branch NameError." }, { "left": "issue:43653", "right": "issue:44442", "accept": false, - "reason": "Tokenizer special-token registration bug is unrelated to AutoTokenizer failing to load a different tokenizer class." + "reason": "Tokenizer special-token registration bug is unrelated to FastSpeech2Conformer AutoTokenizer loading failure." }, { "left": "issue:41720", "right": "issue:45229", "accept": false, - "reason": "Different model-specific runtime failures: cudaErrorAssert on Qwen3 vs OOM on Gemma4 31B." + "reason": "Different model/loading failures: Qwen3 auto device mapping assert vs Gemma4 OOM." }, { "left": "issue:45440", "right": "issue:45593", "accept": false, - "reason": "Native implementation divergence is unrelated to missing auxiliary losses when denoising is off." + "reason": "DeepSeekV3 implementation divergence is unrelated to D-FINE auxiliary-loss handling." }, { "left": "issue:29942", "right": "issue:36010", "accept": false, - "reason": "Flash Attention test failures are unrelated to GenerationMixin import errors." + "reason": "FlashAttention test failures are unrelated to GenerationMixin import errors." }, { "left": "issue:32090", "right": "issue:33357", "accept": false, - "reason": "Trainer broadcast NoneType error and MacOS bus error are unrelated." + "reason": "Trainer GPU broadcast NoneType crash is unrelated to MacOS CLIP bus error." }, { "left": "issue:43334", "right": "issue:43531", "accept": false, - "reason": "Qwen3-VL pad_token_id loading issue and Qwen3-MoE sliding_window behavior are different defects." + "reason": "Different Qwen3-family bugs: missing pad_token_id in VL config vs sliding-window MoE behavior." }, { "left": "issue:43526", "right": "issue:44186", "accept": false, - "reason": "BeitImageProcessorFast reduce_labels bug and LayoutLMv2Tokenizer NER/padding crash are unrelated preprocessing issues." + "reason": "BeitImageProcessorFast reduce_labels bug is unrelated to LayoutLMv2Tokenizer NER padding/truncation crashes." }, { "left": "issue:44977", "right": "issue:45245", "accept": false, - "reason": "Flash-attention generation bug and category-count limit error are unrelated." + "reason": "Flash-attention generation failure is unrelated to category-count limits." }, { "left": "issue:43824", "right": "issue:44279", "accept": false, - "reason": "Specific Qwen2.5-VL import error vs a generic dependency issue are not the same bug." + "reason": "Specific missing Qwen2.5-VL export is not the same as a generic dependency issue." }, { "left": "issue:43526", "right": "issue:44857", "accept": false, - "reason": "Label reduction bug and float16/CUDA loss crash are different components and failure modes." + "reason": "Different failures: Beit label reduction vs LwDetrImageLoss AMP CUDA crash." }, { - "left": "issue:42915", - "right": "issue:43931", + "left": "issue:43526", + "right": "issue:44610", "accept": false, - "reason": "FineGrainedFP8Config failure and Qwen3-VL weight-shape mismatch are distinct model-loading problems." + "reason": "Both are image-processor related, but they concern different models and different bugs." }, { - "left": "issue:43526", - "right": "issue:44610", + "left": "issue:42915", + "right": "issue:43931", "accept": false, - "reason": "Beit label reduction and OmDet-Turbo input-size mismatch are unrelated." + "reason": "Qwen3MoE FP8 config failure is different from Qwen3-VL weight-shape mismatch." }, { "left": "issue:43262", "right": "issue:44206", "accept": false, - "reason": "Audio chat-template sampling-rate default and unsupported center argument in LasrFeatureExtractor are different processor bugs." + "reason": "Audio chat-template sampling-rate default is unrelated to unsupported `center` argument in feature extraction." }, { - "left": "issue:43441", + "left": "issue:43452", "right": "issue:45588", "accept": false, - "reason": "Both involve flash attention, but one is a Ministral-3 regression and the other is null-handling for sink-less models." + "reason": "Tokenizer/model loading via gguf_file is unrelated to flash_attention sink-less `s_aux=None` crash." }, { - "left": "issue:43452", + "left": "issue:43441", "right": "issue:45588", "accept": false, - "reason": "GGUF loading via AutoTokenizer/AutoModelForCausalLM is unrelated to flash-attention s_aux null handling." + "reason": "Both mention flash attention, but the concrete failing paths and fixes are different." }, { "left": "issue:44442", "right": "issue:45072", "accept": false, - "reason": "Tokenizer loading and bfloat16 dtype mismatch are unrelated issues." - }, - { - "left": "issue:44743", - "right": "issue:44811", - "accept": false, - "reason": "Recurrent state reset in Qwen3.5 modular code and Whisper skip_special_tokens behavior are different problems." + "reason": "Different issues: tokenizer loading vs bfloat16 dtype mismatches in inference." }, { "left": "issue:43575", "right": "issue:43701", "accept": false, - "reason": "Tensor-parallel OOM and checkpoint resume key mismatch are unrelated." + "reason": "TP OOM when loading Qwen2 is unrelated to resume-from-checkpoint key mismatch." } ] }, @@ -8088,7 +8012,6 @@ "issue:38175", "issue:39692", "issue:42907", - "issue:43295", "issue:43316", "issue:43381", "issue:43493", @@ -8111,6 +8034,7 @@ "issue:44464", "issue:44479", "issue:44743", + "issue:44811", "issue:44898", "issue:44964", "issue:44977", @@ -8127,6 +8051,7 @@ "issue:45803" ], "soft_pairs": [ + "issue:44743|issue:44811", "issue:44464|issue:45491", "issue:44442|issue:45803", "issue:29127|issue:35141", @@ -8137,13 +8062,13 @@ "issue:33357|issue:38175", "issue:45588|issue:45600", "issue:44479|issue:45290", + "issue:43316|issue:43525", "issue:44964|issue:45440", "issue:43761|issue:44079", - "issue:43316|issue:43525", "issue:28218|issue:29127", "issue:43931|issue:45216", - "issue:42907|issue:45698", "issue:28218|issue:32090", + "issue:42907|issue:45698", "issue:33290|issue:34567", "issue:34689|issue:43611", "issue:43761|issue:44188", @@ -8152,182 +8077,202 @@ "issue:43381|issue:43873", "issue:43493|issue:43827", "issue:43550|issue:44743", - "issue:43994|issue:44206", - "issue:43295|issue:45584" + "issue:43994|issue:44206" ], - "prepared_review_unit_hash": "6e4375c362c714d1eacc015622fe8d19b3868b46e75c7e6035b9801b4d25adf7", + "prepared_review_unit_hash": "b85d752dbb3e516cfe35822bfb0fadff3400ec348d67e28ec4ae1e2bef30c3b9", "component_budget": { "node_count": 387, "item_count": 387, "soft_pair_count": 1455, - "serialized_chars": 749552, - "estimated_input_tokens": 187388, - "estimated_eval_tokens": 375032 + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 27, - "serialized_chars": 35874, - "estimated_input_tokens": 8969, - "estimated_eval_tokens": 18194 + "serialized_chars": 35848, + "estimated_input_tokens": 8962, + "estimated_eval_tokens": 18180 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is highly heterogeneous: most pairs share only a broad theme (loading, OOM, regressions, serialization) but not the same concrete bug or change, so no soft edges should be merged.", - "confidence": 0.94, - "canonical_issue_reason": "issue:45491 is the clearest standalone bug report here: a specific, reproducible GPU NaN regression tied to sliding-window attention and mixed-length batching.", + "summary": "The set is mostly heterogeneous; nearly all candidate pairs are separate bugs. The only strong duplicate-like pair is the generic `ModelOutput` None-key assignment issue and the CLIPVision hidden_states regression it can explain.", + "confidence": 0.84, + "canonical_issue_reason": "issue:44079 is the best canonical issue because it describes the broader root-cause bug in `ModelOutput` key assignment; `issue:43761` looks like a model-specific symptom of that same problem.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45491 is the best representative issue because it has a focused failure mode, clear environment constraints, and a plausible code-path root cause.", + "best_issue_reason": "issue:44079 is the strongest representative of the only plausible duplicate subcluster and has the most general, reusable fix scope.", "best_pr_reason": null, "soft_edge_verdicts": [ + { + "left": "issue:44743", + "right": "issue:44811", + "accept": false, + "reason": "Different subsystems and failure modes: cache/recurrent-state reset vs Whisper decoding flags." + }, { "left": "issue:44464", "right": "issue:45491", "accept": false, - "reason": "Different failures: compiled-forward inconsistency vs. NaNs from sliding-window attention on mixed-length batches." + "reason": "Both involve generation quality, but the reported bugs are distinct: compiled chunked generation vs Gemma3 NaN embeddings with mixed-length windows." }, { "left": "issue:44442", "right": "issue:45803", "accept": false, - "reason": "Tokenizer loading failure and bare-except error swallowing are unrelated bugs in different code paths." + "reason": "Unrelated tokenizer-loading failure vs a tensor conversion exception swallowing interrupts." }, { "left": "issue:29127", "right": "issue:35141", "accept": false, - "reason": "Clarity of an error message in LayoutLMv3 is unrelated to embedding reinitialization after resizing tokens." + "reason": "LayoutLMv3 box validation/error messaging is unrelated to embedding reinitialization after resize." }, { "left": "issue:43646", "right": "issue:44077", "accept": false, - "reason": "Custom model initialization regression and patchtsmixer post_init policy are not the same concrete bug." + "reason": "Related theme, but not the same concrete bug: generic custom-model init breakage vs a specific `patchtsmixer` `post_init` compatibility issue." }, { "left": "issue:43575", "right": "issue:44387", "accept": false, - "reason": "Both mention OOM, but one is tensor-parallel loading and the other is int4 quantization memory growth." + "reason": "Both are OOM reports, but with different triggers and code paths: tensor parallel loading vs int4 quantization memory growth." }, { "left": "issue:45072", "right": "issue:45229", "accept": false, - "reason": "Bfloat16 dtype mismatch in inference and multi-GPU CUDA OOM are different failure modes." + "reason": "Different problems: dtype mismatch in inference vs multi-GPU CUDA OOM." }, { "left": "issue:44977", "right": "issue:45584", "accept": false, - "reason": "Flash-attention generation trouble and empty-transcription handling after align_special_tokens are separate code paths." + "reason": "Unrelated model-generation regressions in different models and code paths." }, { "left": "issue:33357", "right": "issue:38175", "accept": false, - "reason": "MacOS bus error on CLIP and zero probabilities on SigLIP2 are different model/runtime bugs." + "reason": "Different models and symptoms: MacOS bus error on CLIP vs zero-probability outputs on SigLIP2." }, { "left": "issue:45588", "right": "issue:45600", "accept": false, - "reason": "Crash on s_aux=None in flash_attention is unrelated to stale auto_mappings references in CI." + "reason": "Different code paths: flash-attention crash on `s_aux=None` vs stale auto-mapping references breaking CI." }, { "left": "issue:44479", "right": "issue:45290", "accept": false, - "reason": "Video-input regression for Qwen VL models and chat-template tool-call crashes are distinct issues." + "reason": "Video-input regression and chat-template/tool-call crash are separate bugs." + }, + { + "left": "issue:43316", + "right": "issue:43525", + "accept": false, + "reason": "Both are config/API issues, but they concern different missing/uneven attributes in different model configs." }, { "left": "issue:44964", "right": "issue:45440", "accept": false, - "reason": "Phi-4 multimodal loading failure and DeepseekV3MoE implementation divergence do not share the same root cause." + "reason": "Different model families and failure modes: loading Phi-4 multimodal vs DeepSeekV3 implementation divergence." }, { "left": "issue:43761", - "right": "issue:44188", + "right": "issue:44079", + "accept": true, + "reason": "`issue:44079` describes the generic `ModelOutput` None-key assignment bug that plausibly causes `issue:43761`'s missing `hidden_states` symptom." + }, + { + "left": "issue:28218", + "right": "issue:32090", "accept": false, - "reason": "Missing hidden_states in CLIPVisionModel and attention-kernel divergence under torch.compile are different bugs." + "reason": "Tokenizer spacing bug and Trainer/NCCL broadcast TypeError are unrelated." }, { - "left": "issue:44442", - "right": "issue:44898", + "left": "issue:43931", + "right": "issue:45216", "accept": false, - "reason": "AutoTokenizer loading failure and Perceiver image-classification resolution handling are unrelated." + "reason": "Both mention Qwen and loading/saving, but one is a shape-mismatch load error and the other is an incorrect checkpoint save regression." }, { - "left": "issue:39692", - "right": "issue:45491", + "left": "issue:42907", + "right": "issue:45698", "accept": false, - "reason": "Documentation example/model mismatch and GPU NaNs in mixed-length batching are not the same underlying problem." + "reason": "Dequantized-model saving failure is not the same as loading the wrong custom module after `save_pretrained`." }, { - "left": "issue:43381", - "right": "issue:43873", + "left": "issue:33290", + "right": "issue:34567", "accept": false, - "reason": "Eval-mode gradient checkpointing and quantized offloading behavior are separate concerns." + "reason": "OOM with Adafactor/DeepSpeed is unrelated to the trainer state token counter not updating." }, { - "left": "issue:43493", - "right": "issue:43827", + "left": "issue:34689", + "right": "issue:43611", "accept": false, - "reason": "SigLIP2 implementation mismatch and docs still referencing pipeline() are unrelated." + "reason": "Both are loading regressions, but they appear to stem from different version changes and different affected mechanics." }, { - "left": "issue:43550", - "right": "issue:44743", + "left": "issue:43761", + "right": "issue:44188", "accept": false, - "reason": "SDPA + torch.compile failure in Bamba and recurrent-state reset in Qwen3.5 are different code-path bugs." + "reason": "`hidden_states` output handling is unrelated to attention-kernel divergence under `torch.compile`." }, { - "left": "issue:43994", - "right": "issue:44206", + "left": "issue:44442", + "right": "issue:44898", "accept": false, - "reason": "SigLIP2 output corruption and LASR feature extractor center-arg crash are unrelated." + "reason": "Different vision-tokenizer/model failures with no shared concrete bug." }, { - "left": "issue:43295", - "right": "issue:45584", + "left": "issue:39692", + "right": "issue:45491", "accept": false, - "reason": "Processor/tokenizer regression and Whisper empty-transcription handling are different APIs and symptoms." + "reason": "Documentation example errors are unrelated to the Gemma3 mixed-length NaN embedding regression." }, { - "left": "issue:28218", - "right": "issue:32090", + "left": "issue:43381", + "right": "issue:43873", "accept": false, - "reason": "Added-token spacing bug and Trainer GPU broadcast NoneType error are unrelated." + "reason": "Gradient checkpointing in eval mode and quantization offloading are separate features/issues." }, { - "left": "issue:33290", - "right": "issue:34567", + "left": "issue:43493", + "right": "issue:43827", "accept": false, - "reason": "AdaFactor/deepspeed OOM and TrainerState token counter not updating do not share a concrete root cause." + "reason": "SigLIP2 implementation drift is unrelated to docs still referencing removed `pipeline()` APIs." }, { - "left": "issue:34689", - "right": "issue:43611", + "left": "issue:43550", + "right": "issue:44743", "accept": false, - "reason": "These are both loading regressions, but they involve different versions and mechanisms (Llama vision model load vs base_model_prefix handling)." + "reason": "torch.compile+SDPA failure and cache/recurrent-state reset are distinct bugs." }, { - "left": "issue:42907", - "right": "issue:45698", + "left": "issue:43994", + "right": "issue:44206", "accept": false, - "reason": "Both touch save/load flows, but one is dequantized-model saving and the other is loading the wrong custom module after save_pretrained." + "reason": "SigLIP2 wrong outputs and a LASR feature-extractor crash are unrelated." } ] }, - "evaluator_result": null, - "evaluator_used": false, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet and stays mostly conservative. The single accepted pair (`issue:43761` / `issue:44079`) is defensible from the titles alone: a generic `ModelOutput` key-assignment bug plausibly explains a model-specific `hidden_states=None` regression. The other soft-edge rejections are also cautious and do not overclaim shared root causes." + }, + "evaluator_used": true, "retried": false, - "accepted_nontrivial_soft_edge": false, + "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null }, @@ -8394,12 +8339,13 @@ "issue:45759" ], "soft_pairs": [ + "issue:43295|issue:45584", "issue:43316|issue:45759", "issue:44568|issue:45479", - "issue:32090|issue:33290", "issue:41628|issue:45372", - "issue:43643|issue:44913", "issue:43720|issue:44442", + "issue:43643|issue:44913", + "issue:32090|issue:33290", "issue:44190|issue:44496", "issue:43873|issue:44871", "issue:34689|issue:36010", @@ -8408,8 +8354,8 @@ "issue:44855|issue:45137", "issue:33290|issue:45656", "issue:29127|issue:33357", - "issue:43540|issue:44514", "issue:43381|issue:43742", + "issue:43540|issue:44514", "issue:43316|issue:44877", "issue:33290|issue:34689", "issue:42175|issue:43316", @@ -8423,202 +8369,208 @@ "issue:43550|issue:43701", "issue:43606|issue:43701" ], - "prepared_review_unit_hash": "87eb556f1627daebf5afebd07dbd9c3e9e9d827d48b543adecbc7f05373a9217", + "prepared_review_unit_hash": "126822630cb276da560052c51516b2b065a9f843e3bba95bd01ce8e006c826eb", "component_budget": { "node_count": 387, "item_count": 387, "soft_pair_count": 1455, - "serialized_chars": 749552, - "estimated_input_tokens": 187388, - "estimated_eval_tokens": 375032 + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 28, - "serialized_chars": 36309, - "estimated_input_tokens": 9078, - "estimated_eval_tokens": 18412 + "soft_pair_count": 29, + "serialized_chars": 36610, + "estimated_input_tokens": 9153, + "estimated_eval_tokens": 18562 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is not a true duplicate set: the soft links mostly reflect broad topical similarity (config loading, multimodal processors, deepspeed, tokenizers) rather than the same underlying bug. I would not merge any of the proposed soft pairs.", - "confidence": 0.89, - "canonical_issue_reason": "No strong canonical issue: the items span many unrelated regressions, and the soft links are mostly superficial thematic matches rather than one shared root cause.", + "summary": "No convincing duplicates here: the soft pairs span unrelated issues across config loading, generation, multimodal processors, DeepSpeed, tokenizers, and model-specific regressions. I would not merge any of the suggested pairs, and there is no safe canonical representative for the set.", + "confidence": 0.93, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "No global best issue for deduplication; issue 43316 is the closest thematic anchor for the config/API subset, but it does not represent the rest of the cluster.", + "best_issue_reason": "No issue in this set is a safe canonical representative; the items are largely unrelated one-off bugs rather than variants of the same underlying problem.", "best_pr_reason": null, "soft_edge_verdicts": [ + { + "left": "issue:43295", + "right": "issue:45584", + "accept": false, + "reason": "Different bugs: processor.tokenizer/images-to-tokenizer regression vs Whisper empty-transcription handling after align_special_tokens." + }, { "left": "issue:43316", "right": "issue:45759", "accept": false, - "reason": "Both are config-related, but they hit different model families and different code paths; not the same bug." + "reason": "Both involve config/model loading, but the concrete bugs differ: Gemma3TextConfig API mismatch vs Qwen composite config unwrap logic." }, { "left": "issue:44568", "right": "issue:45479", "accept": false, - "reason": "Tokenizer special-token handling and zero-loss classification are unrelated failures." + "reason": "Unrelated: tokenizer special-token insertion regression vs zero-loss classification bug for num_labels=1." }, { - "left": "issue:32090", - "right": "issue:33290", + "left": "issue:41628", + "right": "issue:45372", "accept": false, - "reason": "Both involve trainer/deepspeed, but one is a NoneType broadcast error and the other is an Adafactor OOM." + "reason": "Both are import errors, but they stem from different missing symbols and different load paths." }, { - "left": "issue:41628", - "right": "issue:45372", + "left": "issue:43720", + "right": "issue:44442", "accept": false, - "reason": "Both are import/loading errors, but for different symbols and different processor stacks." + "reason": "Packed-weight loading bug vs AutoTokenizer failing for a specific tokenizer class; no shared code path." }, { "left": "issue:43643", "right": "issue:44913", "accept": false, - "reason": "Both concern config persistence, but one is missing fields with trust_remote_code and the other resets a hyperparameter on reload." + "reason": "Different config persistence issues: trust_remote_code field loss vs rotary_pct resetting on reload." }, { - "left": "issue:43720", - "right": "issue:44442", + "left": "issue:32090", + "right": "issue:33290", "accept": false, - "reason": "Packed-weight loading and tokenizer loading are different failures with no shared concrete code path." + "reason": "Trainer broadcast NoneType error and DeepSpeed+Adafactor OOM are distinct failures." }, { "left": "issue:44190", "right": "issue:44496", "accept": false, - "reason": "Example-script dataset loading and unrecognized model config errors are unrelated." + "reason": "Local dataset loading in an example script vs unrecognized model/config metadata loading." }, { "left": "issue:43873", "right": "issue:44871", "accept": false, - "reason": "Quantized offloading behavior and eos_token_id inconsistency are distinct issues." + "reason": "Quantization/offloading behavior vs Gemma eos_token_id config mismatch are unrelated." }, { "left": "issue:34689", "right": "issue:36010", "accept": false, - "reason": "Model-loading regression and GenerationMixin import failure are not the same bug." + "reason": "Model loading regression for Llama 3.2 Vision vs GenerationMixin import failure are different issues." }, { "left": "issue:43381", "right": "issue:43856", "accept": false, - "reason": "Gradient checkpointing in eval mode and MoE memory usage are different problems." + "reason": "Gradient checkpointing in eval mode and Qwen3 MoE memory inefficiency are separate training concerns." }, { "left": "issue:44077", "right": "issue:44479", "accept": false, - "reason": "A post_init policy change and a video input regression do not share the same underlying defect." + "reason": "PatchTSMixer post_init validation vs Qwen video-input regression are unrelated." }, { "left": "issue:44855", "right": "issue:45137", "accept": false, - "reason": "Python 3.13 import parsing and DeepSpeed deque underflow are unrelated." + "reason": "Python 3.13 torch.jit parsing error vs DeepSpeed ZeRO3 deque underflow are different failures." }, { "left": "issue:33290", "right": "issue:45656", "accept": false, - "reason": "Both mention deepspeed/optimizer, but OOM from Adafactor and double optimizer-step calls are different bugs." + "reason": "Both mention DeepSpeed/optimizer, but one is OOM during Adafactor use and the other is double optimizer stepping." }, { "left": "issue:29127", "right": "issue:33357", "accept": false, - "reason": "LayoutLMv3 box validation and a CLIP bus error on macOS are not the same issue." + "reason": "LayoutLMv3 error-message clarity vs MacOS bus error on CLIP are not the same bug." }, { - "left": "issue:43540", - "right": "issue:44514", + "left": "issue:43381", + "right": "issue:43742", "accept": false, - "reason": "Both are multimodal/video-related, but the failures occur in different processor APIs and different inputs." + "reason": "Gradient checkpointing eval-mode issue and MobileLLM key error are unrelated." }, { - "left": "issue:43381", - "right": "issue:43742", + "left": "issue:43540", + "right": "issue:44514", "accept": false, - "reason": "Gradient checkpointing eval behavior and MobileLLM key-loading are unrelated." + "reason": "Different Qwen multimodal failures: video processing ValueError vs batched chat-template crash with padding=False." }, { "left": "issue:43316", "right": "issue:44877", "accept": false, - "reason": "Config API discrepancy and strict granite_speech config loading are different model-config problems." + "reason": "Gemma3TextConfig API discrepancy is not the same as granite_speech being rejected by strict config loading." }, { "left": "issue:33290", "right": "issue:34689", "accept": false, - "reason": "Deepspeed OOM and Llama 3.2 model-loading breakage do not share a code path." + "reason": "Adafactor/DeepSpeed OOM and Llama 3.2 Vision loading breakage are unrelated." }, { "left": "issue:42175", "right": "issue:43316", "accept": false, - "reason": "Package/backend installation and model config API mismatch are unrelated." + "reason": "Package/backend dependency issue vs Gemma3 config API mismatch; no shared underlying defect." }, { "left": "issue:25251", "right": "issue:28282", "accept": false, - "reason": "Pipeline output shape regression and missing PyTorch import are unrelated." + "reason": "Pipeline nesting behavior and missing PyTorch ImportError are completely different." }, { "left": "issue:44162", "right": "issue:44792", "accept": false, - "reason": "ESM2 breakage and janus test failure are distinct model-specific issues." + "reason": "ESM2 breakage and janus image-generation test failure are different model-specific issues." }, { "left": "issue:43404", "right": "issue:43441", "accept": false, - "reason": "A tied-weight bug and a FlashAttention failure are different model/runtime problems." + "reason": "Mistral3 weight tying bug and Ministral FlashAttention failure are distinct code paths." }, { "left": "issue:43232", "right": "issue:43388", "accept": false, - "reason": "Generation kwargs handling and gather_for_metrics label dropping are unrelated trainer issues." + "reason": "Generation kwargs update bug vs metric-gathering label truncation are unrelated." }, { "left": "issue:43295", "right": "issue:43825", "accept": false, - "reason": "Processor.tokenizer/image regression and pipeline error-message wording are not the same underlying bug." + "reason": "Processor/tokenizer regression vs pipeline error-message wording are different problems." }, { "left": "issue:41950", "right": "issue:43475", "accept": false, - "reason": "Pipeline image-processor lookup and a missing SAM3 video output attribute are different failures." + "reason": "Video-classification pipeline choosing image processors is unrelated to SAM3 output attribute missing." }, { "left": "issue:44871", "right": "issue:45584", "accept": false, - "reason": "EOS-token config mismatch and Whisper empty transcription after alignment are unrelated." + "reason": "Gemma eos_token_id mismatch and Whisper empty transcription failure are different bugs." }, { "left": "issue:43550", "right": "issue:43701", "accept": false, - "reason": "torch.compile SDPA failure and resume_from_checkpoint key mismatch are not the same bug." + "reason": "torch.compile/SDPA issue in Bamba is unrelated to resume_from_checkpoint key mismatch." }, { "left": "issue:43606", "right": "issue:43701", "accept": false, - "reason": "CPU offload device mismatch and checkpoint key mismatch are unrelated." + "reason": "CPU offload device mismatch and checkpoint-resume key mismatch are separate issues." } ] }, @@ -8694,10 +8646,10 @@ "soft_pairs": [ "issue:45127|issue:45200", "issue:43295|issue:44871", - "issue:43761|issue:44479", "issue:33290|issue:35141", - "issue:33290|issue:36331", + "issue:43761|issue:44479", "issue:43065|issue:44617", + "issue:33290|issue:36331", "issue:43575|issue:43742", "issue:43452|issue:44496", "issue:42617|issue:43575", @@ -8708,8 +8660,8 @@ "issue:41720|issue:43575", "issue:44077|issue:44534", "issue:44479|issue:44734", - "issue:43931|issue:44060", "issue:44464|issue:44871", + "issue:43931|issue:44060", "issue:44617|issue:45584", "issue:43906|issue:45290", "issue:43065|issue:44610", @@ -8718,219 +8670,219 @@ "issue:44534|issue:44913", "issue:41720|issue:45245", "issue:44190|issue:44964", - "issue:44737|issue:45803", "issue:44792|issue:45600", + "issue:44737|issue:45803", "issue:44855|issue:45588", "issue:44521|issue:45381" ], - "prepared_review_unit_hash": "c232b5dad56d77fe7deea3f9702d1619dfbe9ddeeaeada2f2aaa8ffe4b7bca1b", + "prepared_review_unit_hash": "33acde81c7497f34ccecb505c4e460001b31d1cfa31fcdb580d22b85fd578674", "component_budget": { "node_count": 387, "item_count": 387, "soft_pair_count": 1455, - "serialized_chars": 749552, - "estimated_input_tokens": 187388, - "estimated_eval_tokens": 375032 + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 30, - "serialized_chars": 36829, - "estimated_input_tokens": 9208, - "estimated_eval_tokens": 18672 + "serialized_chars": 36824, + "estimated_input_tokens": 9206, + "estimated_eval_tokens": 18668 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is highly heterogeneous: the issues span unrelated areas such as multimodal regressions, model loading, trainer API changes, OOMs, cache behavior, docs, and CI/test breakages. I do not see any soft pair that clearly describes the same underlying bug or change.", - "confidence": 0.96, - "canonical_issue_reason": "Issue 44521 is the most active and externally referenced item here, so it is the best representative if one issue must be chosen, but the cluster does not look like a true duplicate set.", + "summary": "No soft-edge pair looks like a true duplicate; the issues span unrelated regressions, model-specific bugs, and API changes. The most generally representative issue is the open, highly-referenced multimodal chat-template bug, but it is not a good duplicate target for the rest.", + "confidence": 0.95, + "canonical_issue_reason": "issue:44521 is the clearest representative entry: it is open, has the most inbound references, and describes a concrete, user-facing bug in a widely used API. That said, the cluster does not contain a single underlying duplicate problem.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 44521 has the strongest signal for prominence (open, more inbound references, and higher discussion) compared with the rest of this mixed cluster.", + "best_issue_reason": "issue:44521 is the best issue to anchor this set because it is active, broadly applicable, and has clear reproduction/context; the others are too diverse to serve as one duplicate target.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45127", "right": "issue:45200", "accept": false, - "reason": "Different bugs: LoRA/extended-vocab collapse on tied embeddings vs Gemma 4 multimodal token-type IDs defaulting behavior." + "reason": "Different bugs: LoRA/extended-vocab collapse on tied embeddings vs Gemma 4 multimodal token-type IDs defaulting to zeros." }, { "left": "issue:43295", "right": "issue:44871", "accept": false, - "reason": "Different regression classes: custom model code accessing processor.tokenizer/images vs eos_token_id config inconsistency." - }, - { - "left": "issue:43761", - "right": "issue:44479", - "accept": false, - "reason": "CLIPVisionModel hidden_states regression and Qwen video-input vision_position_ids bug are unrelated." + "reason": "Both are regression-style config issues, but one is about processor/tokenizer API breakage and the other is an eos_token_id mismatch in Gemma-3; not the same failure path." }, { "left": "issue:33290", "right": "issue:35141", "accept": false, - "reason": "OOM with Adafactor+Deepspeed is unrelated to embedding reinitialization during post_init." + "reason": "OOM in Adafactor/DeepSpeed training is unrelated to token-embedding resizing and post_init reinitialization." }, { - "left": "issue:33290", - "right": "issue:36331", + "left": "issue:43761", + "right": "issue:44479", "accept": false, - "reason": "Optimizer memory blowup and Trainer compute_loss signature mismatch are separate problems." + "reason": "Different model families and symptoms: CLIPVisionModel missing hidden_states vs Qwen multimodal video-input regression." }, { "left": "issue:43065", "right": "issue:44617", "accept": false, - "reason": "Both are SAM3-related, but one is a dummy Conv2d implementation issue and the other is a video-model OOM." + "reason": "Sam3PixelDecoder dummy Conv2d is a modeling/implementation issue, while Sam3Video OOM is a memory/performance problem." + }, + { + "left": "issue:33290", + "right": "issue:36331", + "accept": false, + "reason": "Training OOM with Adafactor/DeepSpeed is unrelated to a Trainer API signature mismatch in compute_loss()." }, { "left": "issue:43575", "right": "issue:43742", "accept": false, - "reason": "Large-model tensor-parallel OOM vs MobileLLM loading KeyError are not the same failure." + "reason": "Both concern loading large models, but one is a TP/OOM issue and the other is a key/config loading error for a specific model." }, { "left": "issue:43452", "right": "issue:44496", "accept": false, - "reason": "GGUF loading breakage and missing model_type config are distinct model-loading paths." + "reason": "AutoTokenizer/AutoModel gguf_file breakage is a loading-path issue; the other is an unrecognized model/config.json issue for a different checkpoint." }, { "left": "issue:42617", "right": "issue:43575", "accept": false, - "reason": "3d_parallel.py failure is unrelated to Qwen2 TP loading OOM." + "reason": "Failure running a 3d_parallel example is not the same as OOM when loading Qwen2-57B with tensor parallelism." }, { "left": "issue:43646", "right": "issue:44913", "accept": false, - "reason": "Custom model initialization breakage differs from GPTNeoX rotary_pct not persisting on reload." + "reason": "Transformers v5 custom model initialization breakage is broader API behavior, while GPTNeoX rotary_pct reset is a specific config reload bug." }, { "left": "issue:43673", "right": "issue:44743", "accept": false, - "reason": "General cache missing in v5.0.0 during chunked prefill is not the same as Qwen3.5 recurrent-state reset in modular code." + "reason": "GenerationMixin cache missing during chunked_prefill is a v5 cache API issue; recurrent state reset in modular_qwen3_5.py is a separate cache/state bug." }, { "left": "issue:44936", "right": "issue:45698", "accept": false, - "reason": "trainer.evaluate() after train and wrong custom module loading after save_pretrained are different lifecycle bugs." + "reason": "trainer.evaluate() after train() is a Trainer lifecycle bug; wrong custom module after save_pretrained concerns model class resolution on load." }, { "left": "issue:43761", "right": "issue:44242", "accept": false, - "reason": "Hidden_states being None and load-balancing loss omission are unrelated model behaviors." + "reason": "CLIPVision hidden_states regression and missing load-balancing loss when output_router_logits=False are unrelated model/subsystem bugs." }, { "left": "issue:41720", "right": "issue:43575", "accept": false, - "reason": "cudaErrorAssert on auto device mapping is a different failure mode from tensor-parallel OOM." + "reason": "Qwen3 auto device mapping cudaErrorAssert is a runtime placement/assertion bug, not the same as TP OOM when loading Qwen2-57B." }, { "left": "issue:44077", "right": "issue:44534", "accept": false, - "reason": "Optional post_init policy and junk-filled non-persistent buffers are separate v5 issues." + "reason": "PatchTST post_init allowance and non-persistent buffers filled with junk are different initialization-policy regressions." }, { "left": "issue:44479", "right": "issue:44734", "accept": false, - "reason": "Qwen video-input regression and serve KV-cache indexing crash affect different code paths." + "reason": "Qwen video-input vision_position_ids bug is unrelated to transformers serve KV-cache continuation tensor indexing." }, { - "left": "issue:43931", - "right": "issue:44060", + "left": "issue:44464", + "right": "issue:44871", "accept": false, - "reason": "Weight-shape mismatch on model load is not the same as an incorrect tied-weights warning." + "reason": "Chunked generation inconsistency under compiled forward is a generation/runtime bug; Gemma-3 eos_token_id mismatch is a config mismatch." }, { - "left": "issue:44464", - "right": "issue:44871", + "left": "issue:43931", + "right": "issue:44060", "accept": false, - "reason": "Chunked generation inconsistency with compiled forward is unrelated to Gemma eos_token_id config mismatch." + "reason": "Qwen3-VL shape mismatch on load and incorrect tied-weights warning in Qwen3-Next are separate model-specific loading/config issues." }, { "left": "issue:44617", "right": "issue:45584", "accept": false, - "reason": "SAM3Video OOM and Whisper empty-transcription generation failure are unrelated." + "reason": "Sam3Video CUDA OOM and Whisper empty-transcription failure after align_special_tokens are unrelated bugs." }, { "left": "issue:43906", "right": "issue:45290", "accept": false, - "reason": "An isolated reproduction of another issue is too vague to match the specific chat-template tool-call crash." + "reason": "An isolated reproduction of a prior issue is not the same bug as apply_chat_template crashing on assistant tool-call messages with no content." }, { "left": "issue:43065", "right": "issue:44610", "accept": false, - "reason": "Both touch vision models, but dummy Conv2d in Sam3PixelDecoder and OmDet-Turbo input-size mismatch are different bugs." + "reason": "Sam3PixelDecoder dummy layer issue is unrelated to OmDet-Turbo processor/model input-size mismatch." }, { "left": "issue:43827", "right": "issue:43994", "accept": false, - "reason": "Docs still referencing pipeline() is unrelated to a broken Siglip2 pipeline/AutoModel result." + "reason": "Docs still referencing pipeline() after removals is documentation cleanup; siglip2 producing nonsensical outputs is a model/pipeline behavior bug." }, { "left": "issue:44792", "right": "issue:44977", "accept": false, - "reason": "Janus image-generation test failure and Qwen3.5 flash-attention generation issue are distinct." + "reason": "janus image-generation test failure and Qwen3.5 flash-attention generation bug are different models and failure modes." }, { "left": "issue:44534", "right": "issue:44913", "accept": false, - "reason": "Non-persistent buffer initialization and GPTNeoX rotary_pct reload behavior are unrelated." + "reason": "Non-persistent buffer junk in v5 is an initialization serialization issue, not the same as GPTNeoX rotary_pct reset on reload." }, { "left": "issue:41720", "right": "issue:45245", "accept": false, - "reason": "Auto device mapping cuda assert and category-count limit runtime error have no substantive overlap." + "reason": "Both mention runtime failures, but one is a Qwen3 device-mapping assert and the other is a categorical-cardinality limit error; not the same bug." }, { "left": "issue:44190", "right": "issue:44964", "accept": false, - "reason": "Local dataset loading in an example script is unrelated to loading Phi-4 multimodal-instruct." - }, - { - "left": "issue:44737", - "right": "issue:45803", - "accept": false, - "reason": "XLNet CPU tensor creation and a bare-except swallowing interrupts in FuyuBatchFeature are different defects." + "reason": "Local dataset loading failure in an example script is unrelated to loading microsoft/Phi-4-multimodal-instruct with latest transformers." }, { "left": "issue:44792", "right": "issue:45600", "accept": false, - "reason": "Janus test failure and removed Sam3LiteText auto-mapping references are unrelated CI issues." + "reason": "A failed janus image-generation test is not the same as auto_mappings.py still referencing removed Sam3LiteText configs." + }, + { + "left": "issue:44737", + "right": "issue:45803", + "accept": false, + "reason": "XLNet CPU allocation bug from missing device= in torch.arange is unrelated to a bare-except swallowing interrupts/errors in FuyuBatchFeature.convert_to_tensors()." }, { "left": "issue:44855", "right": "issue:45588", "accept": false, - "reason": "Python 3.13 TorchScript parsing failure is unrelated to flash_attention crashing on s_aux=None." + "reason": "Python 3.13 IndentationError in DebertaV2Model import is unrelated to flash_attention.py crashing on s_aux=None for sink-less models." }, { "left": "issue:44521", "right": "issue:45381", "accept": false, - "reason": "All-zero assistant masks in apply_chat_template and wrong video vision_position_ids are both multimodal, but they are different bugs in different stages." + "reason": "Both concern multimodal chat/template behavior, but one is assistant_masks being all zero and the other is video input vision_position_ids being wrong in Qwen2.5-VL; different code paths." } ] }, @@ -9018,14 +8970,14 @@ "issue:44112|issue:45600", "issue:43650|issue:44220", "issue:43116|issue:44190", - "issue:43335|issue:44190", "issue:43295|issue:45276", + "issue:43335|issue:44190", "issue:45245|issue:45584", "issue:43404|issue:45216", "issue:43479|issue:45375", "issue:36246|issue:41628", - "issue:44186|issue:44734", "issue:44933|issue:45362", + "issue:44186|issue:44734", "issue:44464|issue:45563", "issue:38175|issue:43065", "issue:44683|issue:45446", @@ -9034,208 +8986,214 @@ "issue:44734|issue:45632", "issue:44246|issue:45003" ], - "prepared_review_unit_hash": "06d41d23e579231d0f319c72e2677a15550333abdda0a6c9d188cdaa99ee5109", + "prepared_review_unit_hash": "5c64d0d7b2c7a9e2bfe187a9cdcff32575e313c8fd746b4e9ec4729efaa0d2ac", "component_budget": { "node_count": 387, "item_count": 387, "soft_pair_count": 1455, - "serialized_chars": 749552, - "estimated_input_tokens": 187388, - "estimated_eval_tokens": 375032 + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 29, - "serialized_chars": 36479, - "estimated_input_tokens": 9120, - "estimated_eval_tokens": 18496 + "serialized_chars": 36481, + "estimated_input_tokens": 9121, + "estimated_eval_tokens": 18498 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Loose collection of mostly unrelated Transformers bug reports; no pull requests. The soft pairs mostly share superficial subsystem words but not the same concrete bug, so I reject all of them and use the broad v5 compatibility regression as the best representative issue.", - "confidence": 0.38, - "canonical_issue_reason": "issue:44561 is the clearest broad regression in the set: a specific API removal in v5 breaking multiple trust_remote_code models, with explicit failure mode and broader impact than the narrower model-specific reports.", + "summary": "The cluster is highly heterogeneous: these are mostly unrelated single-bug issues, with no soft-edge pair close enough to merge as duplicates. I found no same-root-cause issue pairs under a strict duplicate standard.", + "confidence": 0.95, + "canonical_issue_reason": "issue:41628 is the cleanest, most concrete, and actionable report in the set: a specific import regression with a clear failing symbol and module path.", "canonical_pr_reason": null, - "best_issue_reason": "issue:44561 is the strongest standalone anchor because it has a concrete root cause, cross-cutting impact, and looks more representative of the cluster\u2019s regression/compatibility theme than the one-off model bugs.", + "best_issue_reason": "issue:41628 is the best representative issue because it is narrowly scoped, externally visible, and likely to map to a precise code change or export fix.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45127", "right": "issue:45357", "accept": false, - "reason": "Both involve Qwen3.5 tied-weight/serialization problems, but one is LoRA merge collapse and the other is incorrect visual encoder keys; different codepaths and failures." + "reason": "Different bugs: LoRA merge/tied embeddings vs. incorrect visual encoder keys in save_pretrained for Qwen3.5." }, { "left": "issue:43335", "right": "issue:44855", "accept": false, - "reason": "Config construction bug vs Python 3.13 JIT parsing/IndentationError; unrelated issues." + "reason": "Unrelated failures: SwitchTransformers sparse-layer config creation vs. Python 3.13 import/torch.jit parsing error." }, { "left": "issue:44077", "right": "issue:44877", "accept": false, - "reason": "Both touch config strictness, but they concern different models and different fields (`post_init` vs granite_speech loading)." + "reason": "Different config-validation problems affecting different models and fields; not the same bug." }, { "left": "issue:43644", "right": "issue:44913", "accept": false, - "reason": "Non-persistent buffer junk on v5 vs GPTNeoX rotary_pct not preserved on reload; distinct state/serialization bugs." + "reason": "Non-persistent buffer junk initialization is unrelated to rotary_pct not persisting on config reload." }, { "left": "issue:44938", "right": "issue:45542", "accept": false, - "reason": "Python 3.14 import/load failure vs tensorboard-without-TensorFlow backend error; not the same bug." + "reason": "Python 3.14 import/load failure is a different issue from tensorboard-only backend resolution without TensorFlow." }, { "left": "issue:43976", "right": "issue:44855", "accept": false, - "reason": "Python version support issue vs decorator/comment parsing error; unrelated." + "reason": "Different Python-version compatibility problems with different failure modes." }, { "left": "issue:43550", "right": "issue:44190", "accept": false, - "reason": "torch.compile + SDPA failure in Bamba vs local dataset loading in an example script; unrelated." + "reason": "Torch.compile/SDPA model failure is unrelated to the image-classification script dataset loading bug." }, { "left": "issue:43540", "right": "issue:44734", "accept": false, - "reason": "Qwen3OmniMoe video-input validation vs serve KV-cache indexing bug; different codepaths." + "reason": "Different Qwen serving/multimodal code paths: video-input validation vs. KV-cache continuation indexing." }, { "left": "issue:43761", "right": "issue:44743", "accept": false, - "reason": "CLIPVision hidden_states regression vs recurrent state reset in modular_qwen3_5; unrelated." + "reason": "CLIPVision hidden-states regression and Qwen recurrent-state reset are separate model-specific bugs." }, { "left": "issue:44496", "right": "issue:45803", "accept": false, - "reason": "Model/config loading failure vs bare except swallowing errors; no shared underlying bug." + "reason": "Model auto-loading/config resolution failure is unrelated to FuyuBatchFeature swallowing exceptions." }, { "left": "issue:45276", "right": "issue:45584", "accept": false, - "reason": "Embedding resize/tied-weights issue vs Whisper empty transcription after token alignment; different behavior and model path." + "reason": "Token-embedding resize semantics are unrelated to Whisper empty-transcription generation failure." }, { "left": "issue:44112", "right": "issue:45600", "accept": false, - "reason": "CI stale device override test vs removed Sam3LiteText auto-mapping references; unrelated." + "reason": "Both touch CI around GraniteSpeech/Sam3LiteText, but the causes are different: stale device override test vs removed auto-mapping config references." }, { "left": "issue:43650", "right": "issue:44220", "accept": false, - "reason": "Placeholder/add-data issue vs fbank feature extraction bug; not duplicates." + "reason": "Placeholder issue text and fbank feature extraction error are not the same underlying bug." }, { "left": "issue:43116", "right": "issue:44190", "accept": false, - "reason": "Multi-label classification empty results vs local dataset loading; separate example-script issues." + "reason": "Multi-label classification script behavior is unrelated to local dataset loading in a different training example." }, { - "left": "issue:43335", - "right": "issue:44190", + "left": "issue:43295", + "right": "issue:45276", "accept": false, - "reason": "SwitchTransformers config bug vs example dataset loading failure; unrelated." + "reason": "Processor/tokenizer regression is unrelated to resize_token_embeddings not updating tied embeddings." }, { - "left": "issue:43295", - "right": "issue:45276", + "left": "issue:43335", + "right": "issue:44190", "accept": false, - "reason": "Both are regression-style issues, but one is processor/tokenizer API breakage and the other is tied-embedding resize behavior; not the same bug." + "reason": "SwitchTransformers config construction and example-script dataset loading are unrelated." }, { "left": "issue:45245", "right": "issue:45584", "accept": false, - "reason": "Category-count limit error vs Whisper empty transcription; unrelated." + "reason": "Category-count runtime limit is unrelated to Whisper empty-transcription handling." }, { "left": "issue:43404", "right": "issue:45216", "accept": false, - "reason": "Mistral3 lm_head tying bug vs Qwen3.5 save_pretrained checkpoint regression; different models and symptoms." + "reason": "Different tied-weight/save bugs in different multimodal model families." }, { "left": "issue:43479", "right": "issue:45375", "accept": false, - "reason": "Both are Qwen3.5 multimodal config issues, but one is default initialization from None and the other is a missing strict field; distinct bugs." + "reason": "Both are multimodal config issues, but one is default-init behavior and the other is a missing field dropped by strict config parsing." }, { "left": "issue:36246", "right": "issue:41628", "accept": false, - "reason": "Both are import/export failures, but they involve different symbols (`Qwen2_5_VLImageProcessor` vs `AutoImageProcessor`) and likely different fixes." + "reason": "Both are import errors, but they affect different symbols and paths; not the same export bug." }, { - "left": "issue:44186", - "right": "issue:44734", + "left": "issue:44933", + "right": "issue:45362", "accept": false, - "reason": "LayoutLMv2 tokenization crash vs serve KV-cache continuation bug; unrelated." + "reason": "Missing image_utils import is unrelated to Qwen3.5 chat crash behavior." }, { - "left": "issue:44933", - "right": "issue:45362", + "left": "issue:44186", + "right": "issue:44734", "accept": false, - "reason": "Missing import from image_utils vs Qwen3.5 chat crash; not the same underlying issue." + "reason": "Tokenizer NER/padding bug is unrelated to serve-time KV-cache continuation indexing." }, { "left": "issue:44464", "right": "issue:45563", "accept": false, - "reason": "Compiled-forward chunked generation inconsistency vs stale warning for num_return_sequences; different generation bugs." + "reason": "Compiled-forward chunked generation inconsistency is unrelated to a stale warning in paged generate." }, { "left": "issue:38175", "right": "issue:43065", "accept": false, - "reason": "Zero probabilities in SigLIP2 vs dummy Conv2d in Sam3PixelDecoder; unrelated model bugs." + "reason": "Siglip2 zero-probabilities and Sam3PixelDecoder dummy Conv2d are different model-specific issues." }, { "left": "issue:44683", "right": "issue:45446", "accept": false, - "reason": "Both mention flex_attention/version gating, but one is a torch>=2.9 compile failure and the other is an incorrect import version check; not the same concrete bug." + "reason": "Torch>=2.9 flex_attention failure and a PyTorch version-check import bug are related only by subsystem, not the same defect." }, { "left": "issue:44466", "right": "issue:44913", "accept": false, - "reason": "Serialization of lm_head.weight vs GPTNeoX rotary_pct reload; distinct issues." + "reason": "Serialization of tied lm_head weights is unrelated to rotary_pct not round-tripping." }, { "left": "issue:44561", "right": "issue:45632", "accept": false, - "reason": "API removal breaking trust_remote_code models vs local cache-path collision; unrelated." + "reason": "Removal of is_torch_fx_available affecting trust_remote_code is unrelated to cache-path collisions for local models." }, { "left": "issue:44734", "right": "issue:45632", "accept": false, - "reason": "KV-cache indexing crash vs trust_remote_code cache path collision; different subsystems and failures." + "reason": "Serve API KV-cache indexing crash is unrelated to trust_remote_code cache path collisions." }, { "left": "issue:44246", "right": "issue:45003", "accept": false, - "reason": "Import taking long sometimes vs unsafe sys.modules access; not duplicates." + "reason": "Import latency and unsafe sys.modules access are not the same concrete bug." + }, + { + "left": "issue:44220", + "right": "issue:45003", + "accept": false, + "reason": "Audio fbank feature extraction failure is unrelated to sys.modules access in modeling_utils." } ] }, @@ -9318,9 +9276,9 @@ "issue:43388|issue:43688", "issue:42907|issue:43441", "issue:43295|issue:45335", - "issue:43673|issue:44336", "issue:43854|issue:44964", "issue:44496|issue:45042", + "issue:43673|issue:44336", "issue:44016|issue:44220", "issue:43299|issue:45216", "issue:43531|issue:45600", @@ -9336,18 +9294,18 @@ "issue:43577|issue:45216", "issue:41950|issue:45020", "issue:44016|issue:44062", - "issue:43976|issue:45405", "issue:44351|issue:45278", + "issue:43976|issue:45405", "issue:43388|issue:43827" ], - "prepared_review_unit_hash": "794dcb2847e00911526ddaa418dc79c0eb266be5fcd6679cb6125af3e54595eb", + "prepared_review_unit_hash": "535e704e6f0e5035a72733ac00d63afc2fab04f43281938267198e55eaa145aa", "component_budget": { "node_count": 387, "item_count": 387, "soft_pair_count": 1455, - "serialized_chars": 749552, - "estimated_input_tokens": 187388, - "estimated_eval_tokens": 375032 + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 }, "budget": { "node_count": 47, @@ -9362,208 +9320,205 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a grab bag of distinct Transformers bugs, regressions, and docs/CI issues rather than one tight duplicate set. The only plausible duplicate edge is the generic import-regression report and the specific HybridCache import error.", - "confidence": 0.62, - "canonical_issue_reason": "issue:45278 is the best umbrella issue here: it describes the broader post-upgrade import-regression pattern, which can naturally subsume the specific HybridCache import failure. Most other items are narrower and unrelated to this class of failure.", + "summary": "These are mostly distinct bug reports across model loading, generation, processor behavior, docs, and backend/dependency regressions. The only broad umbrella is the open import-error regression report, but the soft pairs are not close enough to merge as duplicates.", + "confidence": 0.73, + "canonical_issue_reason": "issue:45278 is the broadest representative of the cluster: a post-upgrade import-regression umbrella. The other issues are narrower, model- or subsystem-specific failures rather than the same bug.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45278 is the most representative and actionable issue for the cluster because it captures the shared import-breakage regression at a higher level, while the other reports are largely unrelated one-off bugs.", + "best_issue_reason": "issue:45278 is the best single issue to anchor triage because it captures the widest cross-version import breakage theme and is already phrased as a general regression report.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:36246", "right": "issue:38617", "accept": false, - "reason": "Both are import errors, but they involve different missing symbols and different modules; no clear shared root cause." + "reason": "Both are ImportErrors, but they concern different missing symbols in different modules; not the same underlying regression." }, { "left": "issue:43643", "right": "issue:44496", "accept": false, - "reason": "Both relate to model/config loading, but one is a trust_remote_code config-fields issue and the other is a model_type recognition failure; too different." + "reason": "One is missing fields from trust_remote_code config loading, the other is an unrecognized model/config problem; different failure modes." }, { "left": "issue:45406", "right": "issue:45632", "accept": false, - "reason": "One is a Gemma4Processor serve-time tokenizer attribute error; the other is a trust_remote_code cache-path collision. Different bugs." + "reason": "Gemma4Processor tokenizer attribute loss and trust_remote_code cache-path collisions are unrelated bugs." }, { "left": "issue:45563", "right": "issue:45584", "accept": false, - "reason": "Different generation behaviors and different failure modes; only loosely related by generation code." + "reason": "A stale warning in paged generate() and Whisper failing on empty transcription are separate generation-path issues." }, { "left": "issue:44964", "right": "issue:45600", "accept": false, - "reason": "Phi-4 multimodal load failure and stale auto_mappings references to removed configs are separate problems." + "reason": "Phi-4 multimodal loading and auto_mappings referencing removed Sam3LiteText configs are different code paths." }, { "left": "issue:41950", "right": "issue:44016", "accept": false, - "reason": "A pipeline-image-processor bug and a notebook syntax error are unrelated." + "reason": "Pipeline image-processor lookup bug versus a notebook syntax error; not the same issue." }, { "left": "issue:44112", "right": "issue:44964", "accept": false, - "reason": "CI test flakiness in GraniteSpeech and Phi-4 multimodal loading are different code paths and symptoms." + "reason": "CI stale device-override test failure is unrelated to Phi-4 model loading." }, { "left": "issue:43388", "right": "issue:43688", "accept": false, - "reason": "Both concern training/eval utilities, but one is metric gathering and the other is auxiliary-loss normalization; not the same bug." + "reason": "Metric-gathering label truncation and auxiliary-loss normalization are distinct training/distributed bugs." }, { "left": "issue:42907", "right": "issue:43441", "accept": false, - "reason": "Saving dequantized Ministral/Devstral models and FlashAttention loading issues are distinct model/runtime failures." + "reason": "Different model-specific failures: saving dequantized models versus FlashAttention loading for Ministral-3." }, { "left": "issue:43295", "right": "issue:45335", "accept": false, - "reason": "Processor/tokenizer regression and t5gemma embedding resize behavior are unrelated." - }, - { - "left": "issue:43673", - "right": "issue:44336", - "accept": false, - "reason": "Generation cache regression and ANSI escape output in loading reports have no shared underlying change." + "reason": "Processor/tokenizer regression and t5gemma embedding resize behavior do not share the same root cause." }, { "left": "issue:43854", "right": "issue:44964", "accept": false, - "reason": "Different model families and different load failures; not the same underlying issue." + "reason": "Both are model-loading reports, but for different models and different breakages." }, { "left": "issue:44496", "right": "issue:45042", "accept": false, - "reason": "Model config recognition and PIL backend torchvision dependency are unrelated." + "reason": "Unrecognized config/model-type loading error is unrelated to PIL backend requiring torchvision." + }, + { + "left": "issue:43673", + "right": "issue:44336", + "accept": false, + "reason": "Generation cache regression and ANSI output formatting in loading_report are unrelated." }, { "left": "issue:44016", "right": "issue:44220", "accept": false, - "reason": "Notebook syntax error and feature-extraction bug are unrelated." + "reason": "Notebook syntax error and audio feature extraction failure are unrelated bugs." }, { "left": "issue:43299", "right": "issue:45216", "accept": false, - "reason": "Qwen3VL MoE loading regression and Qwen3.5 save_pretrained checkpoint regression are different concrete bugs." + "reason": "Qwen3VL MoE loading failure and Qwen3.5 save_pretrained checkpoint corruption are different regression classes." }, { "left": "issue:43531", "right": "issue:45600", "accept": false, - "reason": "Sliding-window behavior for Qwen3-MoE and stale auto_mappings references are unrelated." + "reason": "Sliding-window behavior in Qwen3-MoE and auto_mappings references to removed configs are unrelated." }, { "left": "issue:43526", "right": "issue:43901", "accept": false, - "reason": "BeitImageProcessorFast label reduction and text-classification docs wording differ completely." + "reason": "BeitImageProcessorFast label reduction bug versus TextClassificationPipeline docs mismatch; not the same issue." }, { "left": "issue:42175", "right": "issue:44279", "accept": false, - "reason": "A missing TensorFlow backend in an extra install and a generic dependency issue are too vague and not clearly the same bug." + "reason": "A missing optional backend in an extra install and a generic dependency issue are too vague and not clearly the same bug." }, { "left": "issue:45335", "right": "issue:45584", "accept": false, - "reason": "Embedding resize behavior and Whisper empty-transcription generation failure are unrelated." + "reason": "Embedding resize behavior in t5gemma and Whisper empty-transcription generation failure are unrelated." }, { "left": "issue:44075", "right": "issue:44368", "accept": false, - "reason": "SGD optimizer args not used and a tie_word_embeddings warning are different issues." + "reason": "SGD optimizer args being ignored and a tie_word_embeddings warning are different configuration issues." }, { "left": "issue:43381", "right": "issue:45593", "accept": false, - "reason": "Gradient checkpointing in eval mode and D-FINE auxiliary-loss handling are unrelated." + "reason": "Gradient checkpointing in eval mode and D-FINE auxiliary-loss handling are separate training logic bugs." }, { "left": "issue:43299", "right": "issue:45357", "accept": false, - "reason": "Both involve Qwen3.5/Qwen3VL save/load regressions, but they affect different artifacts and serialization paths; not clearly one bug." + "reason": "Both involve Qwen3.5/Qwen3VL regressions, but one is load-time and the other is save_pretrained key serialization; not the same bug." }, { "left": "issue:44496", "right": "issue:45759", "accept": false, - "reason": "Generic unrecognized-model loading and composite Qwen config unwrapping are related only broadly, not the same concrete failure." + "reason": "Unrecognized model/config loading and from_config text_config unwrapping are different AutoModel problems." }, { "left": "issue:44242", "right": "issue:45200", "accept": false, - "reason": "Load-balancing loss gating and Gemma 4 mm_token_type_ids defaults are separate modeling issues." + "reason": "MoE load-balancing loss omission and Gemma4 multimodal token-type defaults are unrelated." }, { "left": "issue:43901", "right": "issue:44496", "accept": false, - "reason": "Documentation mismatch for return_all_scores and a model config loading error are unrelated." + "reason": "Documentation mismatch and model-config loading failure do not describe the same underlying problem." }, { "left": "issue:43577", "right": "issue:45216", "accept": false, - "reason": "Blip2 dtype propagation and Qwen3.5 save_pretrained regression are different bugs." + "reason": "BLIP2 dtype propagation and Qwen3.5 checkpoint saving are different regressions." }, { "left": "issue:41950", "right": "issue:45020", "accept": false, - "reason": "Video-classification pipeline image-processor lookup and broad remote_code breakage are not the same underlying problem." + "reason": "Video-classification pipeline bug and broad remote_code breakage are related only at a high level, not the same concrete bug." }, { "left": "issue:44016", "right": "issue:44062", "accept": false, - "reason": "Notebook syntax error and AddedToken keyword-argument error are unrelated." + "reason": "Notebook syntax error and AddedToken keyword conflict are unrelated." }, { - "left": "issue:43976", - "right": "issue:45405", + "left": "issue:44351", + "right": "issue:45278", "accept": false, - "reason": "Python version support breakage and PEFT version pinning are different dependency issues." + "reason": "HybridCache import failure may be one symptom of the broader import-regression report, but these are not clearly the same underlying bug." }, { - "left": "issue:44351", - "right": "issue:45278", - "accept": true, - "reason": "Both describe import failures after a Transformers upgrade; 45278 is the umbrella regression and 44351 is a concrete instance of the same import-breakage class." + "left": "issue:43976", + "right": "issue:45405", + "accept": false, + "reason": "Python version support breakage and a PEFT minimum-version bump are dependency/versioning issues, but not the same specific problem." }, { "left": "issue:43388", "right": "issue:43827", "accept": false, - "reason": "Metric-gathering tuple handling and outdated docs referencing pipeline() are unrelated." + "reason": "Distributed label gathering bug and docs still referencing pipeline() are unrelated." } ] }, - "evaluator_result": { - "accept": true, - "feedback": "The summary is grounded in the packet: issue:45278 is a reasonable umbrella for the specific HybridCache import failure in issue:44351, and the other soft-edge rejections are conservative given only the titles/body summaries available. No clear overstatement or non-mergeable PR-pair problem appears." - }, - "evaluator_used": true, + "evaluator_result": null, + "evaluator_used": false, "retried": false, - "accepted_nontrivial_soft_edge": true, + "accepted_nontrivial_soft_edge": false, "error_kind": null, "error_message": null }, @@ -9648,10 +9603,10 @@ "issue:43976|issue:44987", "issue:45600|issue:45712", "issue:43450|issue:44279", - "issue:44261|issue:44263", - "issue:42994|issue:43317", "issue:44466|issue:45276", + "issue:42994|issue:43317", "issue:43065|issue:44898", + "issue:44261|issue:44263", "issue:43650|issue:45290", "issue:43441|issue:44683", "issue:44623|issue:45397", @@ -9660,208 +9615,208 @@ "issue:44671|issue:44779", "issue:43653|issue:45593" ], - "prepared_review_unit_hash": "6efb4abbc3a2632959b0a9fffa926e2ff470b205eeffa306bd6d463616d2466b", + "prepared_review_unit_hash": "429213a116068e49d2f66d55f2e8c6f27fd098a6edf497b3db6a54c241d03c92", "component_budget": { "node_count": 387, "item_count": 387, "soft_pair_count": 1455, - "serialized_chars": 749552, - "estimated_input_tokens": 187388, - "estimated_eval_tokens": 375032 + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 29, - "serialized_chars": 36344, - "estimated_input_tokens": 9086, - "estimated_eval_tokens": 18428 + "serialized_chars": 36345, + "estimated_input_tokens": 9087, + "estimated_eval_tokens": 18430 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a false-positive mix of unrelated Transformers issues. The pairwise titles only show loose thematic overlap (tokenizers, multimodal models, quantization, pipeline docs), but not the same underlying bug or mergeable change, so no soft-edge duplicates are accepted.", - "confidence": 0.94, - "canonical_issue_reason": "No single true canonical issue emerges because the cluster is heterogeneous; if a representative is needed, issue 44462 is the broadest tokenizer-loading regression and closest to the cluster\u2019s center of gravity.", + "summary": "This cluster is highly heterogeneous: most items are unrelated model-specific bugs, docs issues, or dependency/regression reports. The soft-edge pairs mostly share only broad subsystem vocabulary (tokenizers, multimodal, quantization, attention) and should not be merged as duplicates.", + "confidence": 0.89, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "Issue 44462 is the best representative because it is a concrete, user-facing regression in core loading behavior and overlaps superficially with several other tokenizer-related reports, while still being specific enough to anchor the cluster.", + "best_issue_reason": "Issue 44462 is the broadest representative of the tokenizer/loading regression subset, but even that theme is only a loose neighborhood here; the cluster is too mixed to support a single true canonical duplicate.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43479", "right": "issue:43525", "accept": false, - "reason": "Different bugs: one is multimodal config defaults for missing vision/audio configs; the other is Llama4Config missing pad_token_id. Both are initialization issues, but not the same code path." + "reason": "Both are config/initialization bugs, but for different model families and different failures (`None` defaults vs missing `pad_token_id`)." }, { "left": "issue:43824", "right": "issue:44933", "accept": false, - "reason": "Both are import-related, but one is a missing Qwen2.5-VL export and the other is a nonexistent image_utils import. Different symbols, different failure modes." + "reason": "Both are import-related, but one is a missing model export and the other is a non-existent utility import; different APIs and root causes." }, { "left": "issue:44462", "right": "issue:45081", "accept": false, - "reason": "Both involve tokenizer loading, but one is AutoTokenizer ignoring tokenizer.json while the other is a Mistral regex patch crash. Separate defects in different code paths." + "reason": "Both touch tokenizer loading, but one ignores `tokenizer.json` while the other crashes in a Mistral regex patch path; not the same bug." }, { "left": "issue:44683", "right": "issue:45542", "accept": false, - "reason": "Unrelated runtime issues: flex_attention/torch compatibility versus a tensorboard-without-TensorFlow backend error. No shared underlying fix." + "reason": "Both involve dependencies, but one is a torch/attention backend failure and the other is an optional TensorFlow backend issue." }, { "left": "issue:32090", "right": "issue:35141", "accept": false, - "reason": "The first is a Trainer GPU broadcast TypeError; the second is embedding reinitialization during post_init. Both mention model initialization, but they are not the same bug." + "reason": "Trainer GPU broadcast `NoneType` and embedding reinit after resize are unrelated code paths." }, { "left": "issue:44279", "right": "issue:45405", "accept": false, - "reason": "A vague dependency issue versus an incorrect PEFT version bump. Not the same failure and not a single code-path fix." + "reason": "A generic dependency problem and an invalid PEFT version bump are different issues." }, { "left": "issue:43299", "right": "issue:43854", "accept": false, - "reason": "Both are model-loading regressions, but for different models and different root causes (Qwen3VL MoE loading vs GLM-4.7-Flash test loading)." + "reason": "Different model families and different failure modes; both are loading bugs only at a very high level." }, { "left": "issue:39692", "right": "issue:43901", "accept": false, - "reason": "SigLIP2 docs/model mismatch and quantization failure is unrelated to TextClassificationPipeline docs mentioning return_all_scores. Different documentation problems." + "reason": "Both are docs-related, but one is a SigLIP2 example/model mismatch and the other is a pipeline parameter docs mismatch." }, { "left": "issue:38175", "right": "issue:44617", "accept": false, - "reason": "Both are performance/model-runtime issues around vision models, but one reports zero probabilities in SigLIP2 and the other CUDA OOM in Sam3Video. Not the same bug." + "reason": "Zero probabilities in SigLIP2 and Sam3Video OOM are unrelated model/runtime failures." }, { "left": "issue:43295", "right": "issue:44779", "accept": false, - "reason": "Both are v5 regressions involving tokenizers, but one is processor.tokenizer / image passing behavior and the other is DeepSeek tokenizer output correctness. Different APIs and symptoms." + "reason": "Both are tokenizer regressions, but one is processor/tokenizer access in multimodal code and the other is DeepSeek tokenization correctness; different paths." }, { "left": "issue:44297", "right": "issue:44462", "accept": false, - "reason": "Both concern tokenizer serialization/loading, but one is save_pretrained writing the wrong tokenizer_class and the other is AutoTokenizer ignoring tokenizer.json. Distinct issues." + "reason": "Both concern tokenizer serialization/loading, but one is wrong `tokenizer_class` on save and the other is repository `tokenizer.json` being ignored." }, { "left": "issue:41950", "right": "issue:43650", "accept": false, - "reason": "Video-classification pipeline processor lookup and a generic 'ADD THE DATA' issue are unrelated; no meaningful duplication." + "reason": "Completely unrelated: pipeline processor selection vs an unclear data-related issue." }, { "left": "issue:44462", "right": "issue:44779", "accept": false, - "reason": "Both are tokenizer regressions, but one is repository tokenizer.json selection and the other is DeepSeek tokenization correctness. Too broad to merge." + "reason": "Repository tokenizer loading and DeepSeek tokenizer output correctness are different bugs." }, { "left": "issue:44521", "right": "issue:45325", "accept": false, - "reason": "Both touch multimodal chat/template behavior, but one is all-zero assistant masks and the other is Qwen2.5-VL rope index temporal scaling. Different functions and outcomes." + "reason": "Both are multimodal, but assistant mask generation and Qwen2.5-VL temporal position scaling are separate code paths." }, { "left": "issue:43976", "right": "issue:44279", "accept": false, - "reason": "Python version compatibility for Transformers 5.1.0 is unrelated to a generic dependency issue." + "reason": "Python version compatibility and a dependency issue are not the same underlying problem." }, { "left": "issue:43976", "right": "issue:44987", "accept": false, - "reason": "Both mention version breakage, but one is Python 3.9+/3.10+ support and the other is loading physical-intelligence/fast. No shared code-path bug." + "reason": "Different regression domains: interpreter-version support vs a specific model load failure." }, { "left": "issue:45600", "right": "issue:45712", "accept": false, - "reason": "Both are repo hygiene/CI issues, but one is stale auto_mappings references and the other is leftover dummy classes leaking into dir(transformers). Different cleanup targets." + "reason": "Both are repository-maintenance style issues, but one is stale auto-mappings and the other is leftover dummy classes." }, { "left": "issue:43450", "right": "issue:44279", "accept": false, - "reason": "Video processor batch-shape bug versus a vague dependency issue. No same underlying defect." + "reason": "Video processor batching shape bug and a dependency issue are unrelated." }, { - "left": "issue:44261", - "right": "issue:44263", + "left": "issue:44466", + "right": "issue:45276", "accept": false, - "reason": "Both are ML/GLM-related precision or indexing concerns, but one is an MLA layernorm config epsilon issue and the other is torch.split return values. Not the same bug." + "reason": "Both involve weight/embedding handling, but one is serialization of tied weights and the other is resize_token_embeddings not updating Gemma-4 outputs." }, { "left": "issue:42994", "right": "issue:43317", "accept": false, - "reason": "Quantized model saving failure and device_map=auto dequantized offload failure are both around quantization, but they affect different stages and different code paths." + "reason": "Quantized-model saving and dequantized model offload loading are different phases of the pipeline." }, { - "left": "issue:44466", - "right": "issue:45276", + "left": "issue:43065", + "right": "issue:44898", "accept": false, - "reason": "Both concern embedding/weight resizing and serialization, but one is inconsistent lm_head.weight saving with tied weights and the other is Gemma4 resize_token_embeddings not updating per-layer/output embeddings. Related area, different bug." + "reason": "A dummy Conv2d implementation and a Perceiver non-default resolution failure are unrelated." }, { - "left": "issue:43065", - "right": "issue:44898", + "left": "issue:44261", + "right": "issue:44263", "accept": false, - "reason": "A dummy Conv2d in Sam3PixelDecoder is unrelated to Perceiver non-default resolution classification failing despite interpolate_pos_encoding=True." + "reason": "Both mention GLM internals, but missing `rms_norm_eps` and `torch.split()` return handling are different bugs." }, { "left": "issue:43650", "right": "issue:45290", "accept": false, - "reason": "Generic issue title versus apply_chat_template crashing on assistant tool-call messages; no overlap in underlying functionality." + "reason": "No meaningful overlap; one is an undefined data issue, the other is a chat-template/tool-call crash." }, { "left": "issue:43441", "right": "issue:44683", "accept": false, - "reason": "Ministral FlashAttention failure in v5 RC and compiled flex_attention failing on torch >= 2.9 are both attention-related, but one is model-specific and the other is a torch-version compatibility break." + "reason": "Both touch attention backends, but they are different models and different backend failures." }, { "left": "issue:44623", "right": "issue:45397", "accept": false, - "reason": "Processor.save_pretrained missing files and gemma-4 zero3 from_pretrained are different save/load problems with no shared bug." + "reason": "Processor save-file omissions and Gemma-4 Zero3 loading are unrelated." }, { "left": "issue:43723", "right": "issue:44162", "accept": false, - "reason": "AutoTokenizer.from_pretrained loading trouble and ESM2 broken workflows are too broad and refer to different subsystems/models." + "reason": "AutoTokenizer loading in v5 and ESM2 breakage are different model/loading issues." }, { "left": "issue:44623", "right": "issue:44749", "accept": false, - "reason": "Processor save/export file omissions versus a slowdown after upgrading Transformers. No common root cause." + "reason": "Missing files on save and a performance regression after upgrade are not the same bug." }, { "left": "issue:44671", "right": "issue:44779", "accept": false, - "reason": "CamemBERT masked LM prediction regression and DeepSeek tokenizer regression are both model correctness issues, but they concern different models and code paths." + "reason": "Different models and different symptoms: masked LM prediction error vs tokenizer output regression." }, { "left": "issue:43653", "right": "issue:45593", "accept": false, - "reason": "BigBirdTokenizer mask token registration and D-FINE auxiliary loss behavior are unrelated; one is tokenizer special-token handling, the other is model loss computation." + "reason": "Tokenizer special-token registration and D-FINE auxiliary-loss behavior are unrelated." } ] }, @@ -9950,10 +9905,10 @@ "issue:43525|issue:43881", "issue:43450|issue:44617", "issue:44589|issue:45362", - "issue:42491|issue:43299", "issue:43901|issue:44246", - "issue:44336|issue:44373", + "issue:42491|issue:43299", "issue:41628|issue:43976", + "issue:44336|issue:44373", "issue:16998|issue:22355", "issue:44242|issue:45593", "issue:43352|issue:45250", @@ -9965,208 +9920,208 @@ "issue:36246|issue:42757", "issue:36246|issue:44351" ], - "prepared_review_unit_hash": "0b4cca80c424b7d0235e7b07108182cba02a031c073cadd692636455c95ce958", + "prepared_review_unit_hash": "6ae55f6d6a9dae10ab2afe5b6d212c1b1c6dccc7fa309d327ae4da72561f8d2c", "component_budget": { "node_count": 387, "item_count": 387, "soft_pair_count": 1455, - "serialized_chars": 749552, - "estimated_input_tokens": 187388, - "estimated_eval_tokens": 375032 + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 29, - "serialized_chars": 36097, - "estimated_input_tokens": 9025, - "estimated_eval_tokens": 18306 + "serialized_chars": 36092, + "estimated_input_tokens": 9023, + "estimated_eval_tokens": 18302 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is highly heterogeneous: import/export regressions, tokenizer/model-loading bugs, processor shape issues, attention-kernel behavior, and a few docs/perf reports. I do not see any soft pair that is clearly the same underlying bug/change.", + "summary": "This cluster is mostly a set of unrelated bug reports that only share broad surface similarity (imports, loading, tokenizer, multimodal, or model-regression themes). None of the soft pairs look like the same underlying bug or change, so all soft edges should be rejected.", "confidence": 0.94, - "canonical_issue_reason": "No single issue cleanly represents a duplicate set; if one must be chosen, issue:44521 is the most active and broadly framed, but the cluster is mostly unrelated.", + "canonical_issue_reason": "issue:41628 is the broadest and most central-looking core-library import regression in the set, and it is the closest thematic representative of the import-name failures here, even though the cluster is not truly duplicate-heavy.", "canonical_pr_reason": null, - "best_issue_reason": "issue:44521 is the most active/open and has the broadest multimodal symptom description, so it is the least arbitrary representative of this mixed set.", + "best_issue_reason": "issue:41628 is the best representative because it is a clear, user-facing core import failure with a well-scoped symptom; the others are either narrower model-specific regressions, docs issues, or unrelated bugs.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43784", "right": "issue:45542", "accept": false, - "reason": "Both are import/runtime errors, but they involve different missing symbols and unrelated backend setups." + "reason": "Both involve import/backend problems, but one is a missing `nn` import on sentence-transformers integration and the other is a TensorBoard/TensorFlow backend detection issue; different code paths." }, { "left": "issue:43901", "right": "issue:43994", "accept": false, - "reason": "One is a docs/behavior mismatch; the other is a model/pipeline output correctness bug." + "reason": "Docs mismatch for `return_all_scores` and incorrect SigLIP2 outputs are unrelated problems in different areas of the library." }, { "left": "issue:43122", "right": "issue:43650", "accept": false, - "reason": "Tokenizer regression vs an unrelated low-signal issue title; no shared code path." + "reason": "Tokenization changes across versions and a nonspecific 'ADD THE DATA' issue do not describe the same bug." }, { "left": "issue:44361", "right": "issue:45593", "accept": false, - "reason": "Different failures in different components: MLukeTokenizer AttributeError vs D-FINE auxiliary-loss logic." + "reason": "`MLukeTokenizer` task AttributeError and missing auxiliary losses in D-FINE are separate model-specific bugs." }, { "left": "issue:44987", "right": "issue:45542", "accept": false, - "reason": "Model loading failure and TensorFlow/tensorboard backend import issue are unrelated." + "reason": "A model-loading failure for `physical-intelligence/fast` is not the same as a TensorBoard-without-TF backend error." }, { "left": "issue:44016", "right": "issue:44291", "accept": false, - "reason": "Notebook syntax error is unrelated to the init_empty_weights TypeError." + "reason": "Notebook syntax error and `init_empty_weights` TypeError are unrelated failures with different triggers and fix paths." }, { "left": "issue:43262", "right": "issue:44521", "accept": false, - "reason": "Audio sampling-rate defaulting and multimodal assistant-mask generation are different bugs." + "reason": "These are both multimodal/chat-template related, but one is an audio sampling-rate default bug and the other is zero assistant masks for multimodal inputs." }, { "left": "issue:44297", "right": "issue:45397", "accept": false, - "reason": "Tokenizer save metadata mismatch is unrelated to the gemma-4 zero3 loading failure." + "reason": "Tokenizer metadata mismatch on save and Gemma-4 zero3 `from_pretrained` failure are different behaviors and code paths." }, { "left": "issue:44188", "right": "issue:44336", "accept": false, - "reason": "Attention-kernel branching under torch.compile is unrelated to ANSI codes in loading_report." + "reason": "Attention-kernel divergence under `torch.compile` and ANSI codes in loading report are completely unrelated." }, { "left": "issue:43673", "right": "issue:43761", "accept": false, - "reason": "Chunked-prefill cache regression and CLIP hidden_states regression are separate code paths." + "reason": "Missing cache in chunked prefill and CLIPVisionModel hidden_states regression affect different components and outputs." }, { "left": "issue:43475", "right": "issue:44016", "accept": false, - "reason": "SAM3 missing attribute bug vs notebook syntax error; no overlap." + "reason": "SAM 3 video missing attribute and notebook syntax error are unrelated." }, { "left": "issue:43525", "right": "issue:45759", "accept": false, - "reason": "Missing pad_token_id and text_config unwrapping are distinct config-handling bugs." + "reason": "Llama4 `pad_token_id` attribute error and Qwen multimodal `text_config` unwrapping are different model/config bugs." }, { "left": "issue:43525", "right": "issue:43881", "accept": false, - "reason": "These are different model-loading problems affecting different models and failure modes." + "reason": "Llama4 `pad_token_id` and glm-4v-9b loading failure are separate model-loading issues with no shared code-path evidence." }, { "left": "issue:43450", "right": "issue:44617", "accept": false, - "reason": "Incorrect batched video-processor shape is not the same as a CUDA OOM." + "reason": "Batched video processor shape bug and SAM 3 CUDA OOM are not the same underlying issue." }, { "left": "issue:44589", "right": "issue:45362", "accept": false, - "reason": "Float8 storage import failure and Qwen3.5 chat crash are unrelated." - }, - { - "left": "issue:42491", - "right": "issue:43299", - "accept": false, - "reason": "Both mention Qwen3 MoE, but they target different model families and incompatibility scenarios; too broad to merge." + "reason": "Float8 storage lookup error and Qwen3.5 chat crash are different runtime failures." }, { "left": "issue:43901", "right": "issue:44246", "accept": false, - "reason": "Docs mismatch vs intermittent import slowness are not the same bug." + "reason": "A documentation inconsistency and a slow-import performance issue are not duplicates." }, { - "left": "issue:44336", - "right": "issue:44373", + "left": "issue:42491", + "right": "issue:43299", "accept": false, - "reason": "ANSI code emission in loading_report and a wrong docstring are unrelated." + "reason": "Both mention Qwen3 MoE loading/version regressions, but they target different version transitions and symptoms; not enough to be the same concrete bug." }, { "left": "issue:41628", "right": "issue:43976", "accept": false, - "reason": "Missing AutoImageProcessor export and Python-version incompatibility are different release issues." + "reason": "Missing `AutoImageProcessor` import and a Python version compatibility problem are different issues." + }, + { + "left": "issue:44336", + "right": "issue:44373", + "accept": false, + "reason": "ANSI escape handling in loading report and a wrong docstring for `position_ids` are unrelated." }, { "left": "issue:16998", "right": "issue:22355", "accept": false, - "reason": "A model_max_length question and a missing transformers.onnx module are unrelated." + "reason": "Model max length behavior for DeBERTa-V3 and missing `transformers.onnx` are unrelated." }, { "left": "issue:44242", "right": "issue:45593", "accept": false, - "reason": "Load-balancing loss gating and D-FINE auxiliary losses concern different models and loss paths." + "reason": "Load-balancing loss routing bug and D-FINE auxiliary-loss behavior are separate model-training issues." }, { "left": "issue:43352", "right": "issue:45250", "accept": false, - "reason": "Model-specific Flash Attention support gap vs a generic Flash Attention topic is too broad to treat as duplicate." + "reason": "A specific Nemotron Flash Attention 2 limitation and a generic Flash Attention 2 issue are too different to merge as one bug." }, { "left": "issue:44295", "right": "issue:45397", "accept": false, - "reason": "Reading position_ids after buffer registration is unrelated to the gemma-4 from_pretrained failure." + "reason": "Reading `position_ids` after buffer registration and Gemma-4 zero3 loading failure are unrelated." }, { "left": "issue:44246", "right": "issue:44496", "accept": false, - "reason": "Intermittent import slowness and unrecognized model config are different problems." + "reason": "Slow import and unrecognized OLMo-Hybrid model config are not the same underlying problem." }, { "left": "issue:41950", "right": "issue:44220", "accept": false, - "reason": "Video-classification processor selection and _torch_extract_fbank_features are separate preprocessing bugs." + "reason": "Video-classification pipeline selecting image processors and `_torch_extract_fbank_features()` are different multimodal/audio bugs." }, { "left": "issue:44462", "right": "issue:45356", "accept": false, - "reason": "Tokenizer.json precedence and Kimi-K2.5 codec regression are different tokenizer issues." + "reason": "AutoTokenizer ignoring repository tokenizer.json and Kimi-K2.5 codec/regex regression are separate tokenizer problems." }, { "left": "issue:44336", "right": "issue:44485", "accept": false, - "reason": "Loading-report ANSI formatting and GLM-5 RoPE implementation are unrelated." + "reason": "Loading-report ANSI codes and GLM-5 RoPE implementation are unrelated." }, { "left": "issue:36246", "right": "issue:42757", "accept": false, - "reason": "Different import errors for different symbols: Qwen2_5_VLImageProcessor vs is_offline_mode." + "reason": "Both are import errors, but they involve different symbols and different dependency layers (`transformers.models...` vs `huggingface_hub`)." }, { "left": "issue:36246", "right": "issue:44351", "accept": false, - "reason": "Both are missing-import reports, but for unrelated exports and code paths." + "reason": "Different missing imports (`Qwen2_5_VLImageProcessor` vs `HybridCache`) indicate distinct breakages and fixes." } ] }, @@ -10250,16 +10205,16 @@ "issue:44295|issue:44749", "issue:44779|issue:45292", "issue:43441|issue:43540", - "issue:44877|issue:45375", "issue:45292|issue:45584", - "issue:44987|issue:45459", + "issue:44877|issue:45375", "issue:45292|issue:45701", + "issue:44987|issue:45459", "issue:43901|issue:45003", "issue:44779|issue:45701", - "issue:43232|issue:44336", - "issue:45561|issue:45563", "issue:36246|issue:43596", "issue:41950|issue:44451", + "issue:45561|issue:45563", + "issue:43232|issue:44336", "issue:45375|issue:45759", "issue:43526|issue:45003", "issue:43257|issue:45137", @@ -10271,20 +10226,20 @@ "issue:42886|issue:45229", "issue:44466|issue:45335" ], - "prepared_review_unit_hash": "21ae3e1cdf9abdaf51aaa1f1efd4b4aab5bb075e5c39800e509b3e7a5ccae321", + "prepared_review_unit_hash": "0a773b8fe9546dcbd45068479c7271f958cf87dd43996ec14f0575662a151d26", "component_budget": { "node_count": 387, "item_count": 387, "soft_pair_count": 1455, - "serialized_chars": 749552, - "estimated_input_tokens": 187388, - "estimated_eval_tokens": 375032 + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 30, - "serialized_chars": 36731, + "serialized_chars": 36729, "estimated_input_tokens": 9183, "estimated_eval_tokens": 18622 }, @@ -10293,192 +10248,192 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The items are mostly unrelated transformer bugs across imports, tokenizers, config loading, generation, deepspeed, and model-specific regressions. None of the soft pairs look like the same underlying defect, so all soft edges should be rejected.", - "confidence": 0.91, - "canonical_issue_reason": "issue:43824 is the strongest representative: it is a broad, user-facing import failure from the top-level transformers API and has multiple inbound references.", + "summary": "This cluster is mostly a loose similarity bucket of unrelated issues. The only plausible duplicate pair is the resize_token_embeddings embedding-propagation bug between the generic and T5Gemma-specific reports.", + "confidence": 0.68, + "canonical_issue_reason": "issue:45292 is the broadest, most general report of the shared resize_token_embeddings / output embedding bug; issue:45335 looks like a model-specific manifestation.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43824 is the best overall issue exemplar in this set because it is high-impact, broadly applicable, and more central than the narrowly scoped model- or backend-specific reports.", + "best_issue_reason": "issue:45292 is the best representative issue for tracking the underlying embedding-resize bug across models.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43673", "right": "issue:44561", "accept": false, - "reason": "Different problems: GenerationMixin cache handling during chunked prefill vs removal of a utility breaking trust_remote_code imports." + "reason": "Different failures: generation cache/chunked prefill vs trust_remote_code breakage from a removed helper." }, { "left": "issue:43824", "right": "issue:43976", "accept": false, - "reason": "Import failure for a specific Qwen2.5 VL class is unrelated to a Python-version compatibility issue for the whole package." + "reason": "Both are import/load problems, but for different symbols and different root causes." }, { "left": "issue:43976", "right": "issue:44964", "accept": false, - "reason": "Python 3.9+/3.10+ support regression is not the same as failing to load one specific model due to config/model mapping." + "reason": "One is a Python-version compatibility issue; the other is a model-loading/config regression." }, { "left": "issue:44246", "right": "issue:44871", "accept": false, - "reason": "Import-time slowdown is unrelated to an eos_token_id configuration mismatch in Gemma-3." + "reason": "Completely unrelated: import-time slowness versus Gemma eos_token_id config mismatch." }, { "left": "issue:44242", "right": "issue:44857", "accept": false, - "reason": "MOE load-balancing loss logic and a CUDA AMP crash in LwDetrImageLoss are different code paths and defects." + "reason": "Different model paths and symptoms; load-balancing loss is not the same as an AMP crash." }, { "left": "issue:44162", "right": "issue:45200", "accept": false, - "reason": "ESM2 breakage is unrelated to Gemma 4 text-only fine-tuning needing default mm_token_type_ids." + "reason": "Distinct model families and bugs; ESM2 breakage is unrelated to Gemma 4 token type IDs." }, { "left": "issue:45362", "right": "issue:45375", "accept": false, - "reason": "A chat/runtime crash for Qwen3.5-35B is not the same bug as a config field being dropped by strict loading." + "reason": "Both mention Qwen, but one is chat/runtime failure and the other is a config field being dropped." }, { "left": "issue:44295", "right": "issue:44749", "accept": false, - "reason": "position_ids buffer access and a post-upgrade filtering slowdown are unrelated symptoms and subsystems." + "reason": "Unrelated issues: position_ids buffer handling vs a performance regression." }, { "left": "issue:44779", "right": "issue:45292", "accept": false, - "reason": "Tokenizer correctness in Deepseek is unrelated to resize_token_embeddings not updating output embeddings." + "reason": "Tokenizer correctness in Deepseek is not the same bug as resize_token_embeddings not updating embeddings." }, { "left": "issue:43441", "right": "issue:43540", "accept": false, - "reason": "Ministral FlashAttention failure and Qwen3OmniMoe video-input validation are different model/runtime issues." + "reason": "Different code paths and model families: FlashAttention failure vs video-input validation." + }, + { + "left": "issue:45292", + "right": "issue:45584", + "accept": false, + "reason": "Resize-token-embeddings bug is unrelated to Whisper empty-transcription generation." }, { "left": "issue:44877", "right": "issue:45375", "accept": false, - "reason": "Both involve config strictness, but one blocks granite_speech loading while the other silently drops a Qwen vision field; different concrete bugs." + "reason": "Different config-loading issues with different models; not the same defect." }, { "left": "issue:45292", - "right": "issue:45584", + "right": "issue:45701", "accept": false, - "reason": "Embedding resize propagation is unrelated to Whisper empty-transcription generation after align_special_tokens." + "reason": "Token embedding resizing and tokenization changes are different problems." }, { "left": "issue:44987", "right": "issue:45459", "accept": false, - "reason": "Loading a specific model on transformers>=5.1.0 is unrelated to tokenizer error masking when protobuf is missing." - }, - { - "left": "issue:45292", - "right": "issue:45701", - "accept": false, - "reason": "A resize-embeddings bug is not the same as a version-dependent tokenization change." + "reason": "Loading a specific model and protobuf error masking are unrelated." }, { "left": "issue:43901", "right": "issue:45003", "accept": false, - "reason": "Documentation mismatch for TextClassificationPipeline is unrelated to unsafe sys.modules access in modeling_utils." + "reason": "Docs mismatch in pipelines vs unsafe sys.modules access are unrelated." }, { "left": "issue:44779", "right": "issue:45701", "accept": false, - "reason": "Deepseek tokenizer regression and a generic version-changes-tokenization report are too broad and not the same concrete bug." + "reason": "Both mention tokenizer behavior, but one is a Deepseek regression and the other is a version-wide tokenization change claim." }, { - "left": "issue:43232", - "right": "issue:44336", + "left": "issue:36246", + "right": "issue:43596", "accept": false, - "reason": "Generation kwargs sync behavior is unrelated to ANSI codes emitted in loading_report." + "reason": "Different failures: Qwen2.5-VL import error vs DeepSpeed/BertModel index error." }, { - "left": "issue:45561", - "right": "issue:45563", + "left": "issue:41950", + "right": "issue:44451", "accept": false, - "reason": "A pytest-xdist race in testing utils is unrelated to a stale warning in paged generate()." + "reason": "Video-classification pipeline processor lookup and ScandiBERT load failure are not the same bug." }, { - "left": "issue:36246", - "right": "issue:43596", + "left": "issue:45561", + "right": "issue:45563", "accept": false, - "reason": "Qwen2.5 VL import failure and a DeepSpeed ZeRO3 BertModel IndexError are unrelated." + "reason": "A pytest-xdist race and a stale warning in generate() are unrelated." }, { - "left": "issue:41950", - "right": "issue:44451", + "left": "issue:43232", + "right": "issue:44336", "accept": false, - "reason": "A video-classification pipeline processor lookup bug is unrelated to failure loading a specific BERT checkpoint." + "reason": "Generation kwargs handling and ANSI output formatting are unrelated." }, { "left": "issue:45375", "right": "issue:45759", "accept": false, - "reason": "Missing a Qwen vision config field is not the same as failing to unwrap text_config in AutoModelForCausalLM.from_config." + "reason": "Both involve Qwen multimodal configs, but the concrete bugs differ: missing field vs text_config unwrapping." }, { "left": "issue:43526", "right": "issue:45003", "accept": false, - "reason": "BeitImageProcessorFast label reduction and sys.modules access in modeling_utils are unrelated bugs." + "reason": "Image processor label reduction bug is unrelated to sys.modules access." }, { "left": "issue:43257", "right": "issue:45137", "accept": false, - "reason": "Qwen3 MOE weight conversion under accelerate+deepspeed is different from a ZeRO3 deque underflow." + "reason": "Different DeepSpeed-related failures with different symptoms and code paths." }, { "left": "issue:43065", "right": "issue:43408", "accept": false, - "reason": "A dummy Conv2d in Sam3PixelDecoder is not the same as a sam3_video/sam3_tracker model-type mismatch." + "reason": "Both are SAM3-related, but one is a dummy layer concern and the other is a model-type warning." }, { "left": "issue:44995", "right": "issue:45200", "accept": false, - "reason": "A stale indexer cache on second forward pass is unrelated to Gemma 4 missing default mm_token_type_ids." + "reason": "Stale indexer cache in GlmMoeDsa is unrelated to Gemma 4 token type IDs." }, { "left": "issue:44496", "right": "issue:45003", "accept": false, - "reason": "Unrecognized model/config loading and unsafe sys.modules access are separate import-time issues." + "reason": "Unrecognized model/config loading versus unsafe sys.modules access are distinct bugs." }, { "left": "issue:43262", "right": "issue:44246", "accept": false, - "reason": "Audio chat-template sampling-rate defaults are unrelated to import-time performance complaints." + "reason": "Audio chat-template sampling-rate bug is unrelated to import-time slowness." }, { "left": "issue:43475", "right": "issue:43540", "accept": false, - "reason": "A missing FPN position embedding attribute in Sam3 video output is unrelated to Qwen3OmniMoe video-input validation." + "reason": "Different models and failures: missing vision-encoder attribute vs video input validation error." }, { "left": "issue:42886", "right": "issue:45229", "accept": false, - "reason": "Offline cache tokenizer loading and Gemma4 multi-GPU OOM are entirely different failure modes." + "reason": "Tokenizer cache/offline loading and Gemma4 multi-GPU OOM are unrelated." }, { "left": "issue:44466", "right": "issue:45335", "accept": false, - "reason": "Device-dependent lm_head serialization is not the same bug as resize_token_embeddings failing to update decoder.embed_tokens." + "reason": "Both mention embeddings, but one is tied-weight serialization and the other is resize_token_embeddings not propagating." } ] }, @@ -10567,38 +10522,38 @@ "issue:44297|issue:44623", "issue:43646|issue:44162", "issue:44361|issue:44749", - "issue:43673|issue:43827", - "issue:43232|issue:43673", "issue:43452|issue:45003", + "issue:43673|issue:43827", "issue:44623|issue:45656", + "issue:43232|issue:43673", "issue:28282|issue:30990", "issue:42994|issue:43482", "issue:44749|issue:45656", "issue:44749|issue:45310", "issue:43475|issue:43643", "issue:45230|issue:45397", - "issue:42371|issue:45092", "issue:44038|issue:44623", + "issue:42371|issue:45092", "issue:43475|issue:45381", "issue:45081|issue:45103", "issue:41628|issue:44933", - "issue:42673|issue:43704", - "issue:42994|issue:43525" + "issue:42994|issue:43525", + "issue:42673|issue:43704" ], - "prepared_review_unit_hash": "3928fbb0f702cd7bab7cc91f82817a16bedbf0eb8c3f61ddc4424d730c2e602c", + "prepared_review_unit_hash": "1b407101868bab82fd63f6e404e9e1b1c928d669094039fd6d901cec34c004b2", "component_budget": { "node_count": 387, "item_count": 387, "soft_pair_count": 1455, - "serialized_chars": 749552, - "estimated_input_tokens": 187388, - "estimated_eval_tokens": 375032 + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 32, - "serialized_chars": 37257, + "serialized_chars": 37259, "estimated_input_tokens": 9315, "estimated_eval_tokens": 18886 }, @@ -10607,204 +10562,216 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a grab bag of unrelated bug reports. The only clear duplicate is the repeated Qwen3ForCausalLM VRAM leak report; the rest differ by model, API surface, or failure mode and should stay separate.", - "confidence": 0.93, - "canonical_issue_reason": "issue:42673 is the best canonical issue: it is the earlier, more-discussed report of the same Qwen3ForCausalLM multi-dataloader-thread VRAM leak also reported in issue:43704.", + "summary": "This cluster is mostly heterogeneous and does not form a single duplicate set. The only clear duplicate pair is the repeated Qwen3ForCausalLM VRAM leak report; everything else appears to be separate bugs or feature regressions.", + "confidence": 0.84, + "canonical_issue_reason": "issue:42673 is the strongest canonical issue because it is an exact title duplicate of issue:43704, is earlier, and has more discussion around the same Qwen3ForCausalLM VRAM leak in multi-dataloader-thread usage.", "canonical_pr_reason": null, - "best_issue_reason": "issue:42673 is the strongest global representative because it is a true duplicate target with the clearest, earliest description of the underlying bug.", + "best_issue_reason": "issue:42673 is the best representative issue in this set: it is specific, actionable, and has the clearest duplicate signal among the listed issues.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44995", "right": "issue:45588", "accept": false, - "reason": "Different failures: stale indexer cache on second forward pass vs flash-attention crash on sink-less models." + "reason": "Different failures: stale indexer cache on second forward pass vs flash-attention crash on s_aux=None." }, { "left": "issue:43653", "right": "issue:44162", "accept": false, - "reason": "Different models and bugs: BigBirdTokenizer special-token registration vs ESM2 breakage." + "reason": "Both involve model/tokenizer breakage, but the concrete bugs are unrelated: BigBird tokenizer special-token registration vs ESM2 loading/runtime failure." }, { "left": "issue:45397", "right": "issue:45656", "accept": false, - "reason": "Unrelated DeepSpeed/ZERO3 from_pretrained issue vs optimizer step being called twice." + "reason": "Different deepspeed-related problems: gemma-4 zero3 from_pretrained vs optimizer.step being called twice." }, { "left": "issue:44368", "right": "issue:44977", "accept": false, - "reason": "Different concrete problems: tie_word_embeddings warning vs flash-attention generation failure." + "reason": "Both mention Qwen3.5, but one is a tied-weights warning and the other is a generation failure with flash-attention; not the same bug." }, { "left": "issue:44368", "right": "issue:45563", "accept": false, - "reason": "Warning about tied embeddings is not the same as stale num_return_sequences warning in paged generate()." + "reason": "A repeated warning about tie_word_embeddings is not the same as a stale num_return_sequences warning in paged generate()." }, { "left": "issue:44060", "right": "issue:44077", "accept": false, - "reason": "Different code paths and symptoms: tied-weights warning vs optional post_init restriction." + "reason": "Both are warning-related, but one is a false tied-weights warning and the other is a patchtsmixer post_init policy change." }, { "left": "issue:45137", "right": "issue:45600", "accept": false, - "reason": "DeepSpeed ZeRO3 deque error is unrelated to removed Sam3LiteText config references." + "reason": "DeepSpeed ZeRO3 deque error and auto_mappings CI breakage are unrelated code paths." }, { "left": "issue:43673", "right": "issue:44734", "accept": false, - "reason": "Both involve generation/cache behavior, but the failing code paths are different enough that they should not be merged." + "reason": "Both are generation/cache bugs, but chunked_prefill cache handling and /v1/responses tensor indexing are different concrete failures." }, { "left": "issue:45003", "right": "issue:45561", "accept": false, - "reason": "Unsafe sys.modules access is unrelated to pytest-xdist file races in testing utils." + "reason": "sys.modules access safety and pytest-xdist file capture races are unrelated." }, { "left": "issue:43931", "right": "issue:44368", "accept": false, - "reason": "Qwen3-VL weight-shape mismatch is unrelated to a tied-embeddings warning." + "reason": "Qwen3-VL weight-shape mismatch and Qwen3.5 LoRA warning are different model-loading vs warning issues." }, { "left": "issue:44933", "right": "issue:45459", "accept": false, - "reason": "Different tokenizer/import problems: missing image_utils import vs protobuf decode-error masking." + "reason": "Nonexistent image_utils import and tokenizer protobuf error handling are unrelated." }, { "left": "issue:44869", "right": "issue:45356", "accept": false, - "reason": "Both touch tokenization/codec handling, but they describe different models and failure modes." + "reason": "Whisper timestamp decode crash and Kimi-K2.5 tokenizer regression are different bugs." }, { "left": "issue:44297", "right": "issue:44623", "accept": false, - "reason": "Tokenizer save metadata mismatch is different from processor.save_pretrained missing files." + "reason": "Tokenizer save metadata mismatch and processor.save_pretrained missing files are separate serialization issues." }, { "left": "issue:43646", "right": "issue:44162", "accept": false, - "reason": "Broad v5 breakage, but one is custom model init and the other is ESM2-specific failure." + "reason": "Custom model initialization breakage and ESM2 breakage are not the same underlying bug." }, { "left": "issue:44361", "right": "issue:44749", "accept": false, - "reason": "MLukeTokenizer AttributeError is unrelated to the reported data-filtering slowdown after upgrade." + "reason": "MLukeTokenizer task AttributeError and a performance regression after upgrading Transformers are unrelated." }, { - "left": "issue:43673", - "right": "issue:43827", + "left": "issue:43452", + "right": "issue:45003", "accept": false, - "reason": "Generation cache failure is not the same as docs still referencing removed pipeline() APIs." + "reason": "gguf_file load-path breakage and unsafe sys.modules access are distinct loading issues." }, { - "left": "issue:43232", - "right": "issue:43673", + "left": "issue:43673", + "right": "issue:43827", "accept": false, - "reason": "Different generation bugs: sync_gpus update handling vs missing GenerationMixin cache in chunked_prefill." + "reason": "Missing GenerationMixin cache in chunked_prefill and docs still referencing pipeline() are unrelated." }, { - "left": "issue:43452", - "right": "issue:45003", + "left": "issue:44623", + "right": "issue:45656", "accept": false, - "reason": "gguf_file/from_pretrained breakage is unrelated to sys.modules access in modeling_utils." + "reason": "Processor save_pretrained missing files and duplicate optimizer steps under deepspeed are different behaviors." }, { - "left": "issue:44623", - "right": "issue:45656", + "left": "issue:43232", + "right": "issue:43673", "accept": false, - "reason": "Missing saved processor files is unrelated to a deepspeed double-step optimizer bug." + "reason": "sync_gpus generation kwargs update and chunked_prefill cache missing are different generation bugs." }, { "left": "issue:28282", "right": "issue:30990", "accept": false, - "reason": "PyTorch missing ImportError is not the same as Sentence Transformers loading hang." + "reason": "PyTorch ImportError and Sentence Transformers hanging on load are unrelated." }, { "left": "issue:42994", "right": "issue:43482", "accept": false, - "reason": "Quantized model saving failure and Qwen2.5-GGUF loading failure are different operations and bugs." + "reason": "Quantized model saving failure and Qwen2.5-GGUF loading failure are opposite-direction, different bugs." }, { "left": "issue:44749", "right": "issue:45656", "accept": false, - "reason": "Performance regression after upgrade is unrelated to optimizer step being called twice." + "reason": "Filtering slowdown after upgrade and duplicate optimizer-step behavior are unrelated." }, { "left": "issue:44749", "right": "issue:45310", "accept": false, - "reason": "Slow filtering after upgrade is unrelated to Qwen3.5 Moe from_pretrained errors." + "reason": "Performance regression in data filtering and Qwen3.5-Moe from_pretrained error are different issues." }, { "left": "issue:43475", "right": "issue:43643", "accept": false, - "reason": "SAM 3 vision encoder missing attribute is unrelated to missing fields from trust_remote_code AutoConfig." + "reason": "SAM3 vision encoder missing attribute and trust_remote_code missing config fields are different missing-field problems." }, { "left": "issue:45230", "right": "issue:45397", "accept": false, - "reason": "Generic bug report does not match the specific Gemma-4 zero3 loading issue." + "reason": "Generic bug report and gemma-4 zero3 from_pretrained are not the same bug." + }, + { + "left": "issue:44038", + "right": "issue:44623", + "accept": false, + "reason": "Qwen3-VL-Moe v5 breakage and processor.save_pretrained missing files are unrelated." }, { "left": "issue:42371", "right": "issue:45092", "accept": false, - "reason": "TF32 settings guidance is unrelated to old InternVL2 meta-init incompatibility." + "reason": "TF32 API settings guidance and InternVL2 meta-init incompatibility are unrelated." }, { - "left": "issue:44038", - "right": "issue:44623", + "left": "issue:43475", + "right": "issue:45381", "accept": false, - "reason": "Qwen3-VL-Moe/v5 issue is unrelated to processor.save_pretrained missing files." + "reason": "SAM 3 video missing attribute and Qwen2.5-VL wrong vision_position_ids are different multimodal bugs." }, { "left": "issue:45081", "right": "issue:45103", "accept": false, - "reason": "Different AttributeErrors in different subsystems: tokenizer patching vs auto-docstring processing." + "reason": "Mistral regex patch crash and auto_docstring future-annotations crash are unrelated." }, { "left": "issue:41628", "right": "issue:44933", "accept": false, - "reason": "Both are import-related, but they point to different missing symbols and likely different fixes." + "reason": "AutoImageProcessor import failure and a different nonexistent image_utils import are not the same concrete issue." + }, + { + "left": "issue:42994", + "right": "issue:43525", + "accept": false, + "reason": "Quantized saving failure and Llama4Config missing pad_token_id are different model/config problems." }, { "left": "issue:42673", "right": "issue:43704", "accept": true, - "reason": "Same underlying bug and effectively the same report: Qwen3ForCausalLM VRAM leak when used across multiple dataloader threads." + "reason": "Exact same title and same underlying bug: Qwen3ForCausalLM VRAM leak when used in multiple dataloader threads." }, { - "left": "issue:42994", - "right": "issue:43525", + "left": "issue:43673", + "right": "issue:43827", "accept": false, - "reason": "Quantized saving failure is unrelated to Llama4Config missing pad_token_id." + "reason": "Repeated pair in the input set would still be unrelated generation cache bug vs docs cleanup." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is grounded in the packet and stays conservative on the soft-edge pairs. The only accepted duplicate, issue:42673 vs issue:43704, is strongly supported by the identical titles and matching VRAM-leak description, with the earlier report reasonably chosen as canonical." + "feedback": "Grounded overall. The summary is conservative and matches the packet: the only clearly duplicate pair is issue:42673 / issue:43704, which have identical titles, earlier/later timestamps, and the same VRAM-leak description. The soft-edge rejections are appropriately cautious and do not overclaim duplication." }, "evaluator_used": true, "retried": false, @@ -10887,8 +10854,8 @@ "issue:36010|issue:41628", "issue:42175|issue:45712", "issue:44038|issue:45310", - "issue:44496|issue:45561", "issue:44964|issue:45137", + "issue:44496|issue:45561", "issue:45561|issue:45600", "issue:42898|issue:44568", "issue:30990|issue:35141", @@ -10904,18 +10871,18 @@ "issue:43723|issue:44568", "issue:42898|issue:45356", "issue:44060|issue:45125", - "issue:43874|issue:45092", "issue:42898|issue:44779", + "issue:43874|issue:45092", "issue:45405|issue:45712" ], - "prepared_review_unit_hash": "fc306137d7ccbf523ae0488d79f0d399c21bce33f3b29cc87e59a9d308a4a856", + "prepared_review_unit_hash": "8bb6a302683a11ad6074122fc6fd21451325736939ef9299c6e2a24e8086e16c", "component_budget": { "node_count": 387, "item_count": 387, "soft_pair_count": 1455, - "serialized_chars": 749552, - "estimated_input_tokens": 187388, - "estimated_eval_tokens": 375032 + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 }, "budget": { "node_count": 48, @@ -10930,204 +10897,204 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These items are mostly unrelated issue reports. The only obvious near-duplicate theme is the repeated Qwen3.5 MoE from_pretrained failure, while the listed soft pairs are not the same underlying bug/change.", - "confidence": 0.86, - "canonical_issue_reason": "issue:45310 is the best canonical anchor because it is the clearest concrete bug report in the set and has an exact later duplicate-style counterpart in issue:45478.", + "summary": "The set is mostly a collection of unrelated bug reports; none of the soft pairs look like true duplicates of the same concrete issue, so all soft edges are rejected. If one issue must represent the cluster, issue:43824 is the strongest candidate due to the highest activity and a clear reproducible import failure.", + "confidence": 0.66, + "canonical_issue_reason": "issue:43824 has the strongest support in this set (high discussion and inbound references) and is a concrete, well-scoped bug report, making it the best representative issue.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45310 is the most suitable representative issue: specific, actionable, and part of the only clear duplicate family here.", + "best_issue_reason": "issue:43824 is the most central and well-documented issue among otherwise unrelated reports, so it is the best issue choice overall.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43475", "right": "issue:44661", "accept": false, - "reason": "Different models and failures: SAM3 vision encoder attribute error vs tokenizer-mapping import path bug." + "reason": "Different failures: a SAM 3 VisionEncoderOutput attribute error vs. a tokenizer-mapping add-new-model-like failure." }, { "left": "issue:45375", "right": "issue:45478", "accept": false, - "reason": "Both mention Qwen3.5 MoE, but one is a missing config field and the other is a from_pretrained regression." + "reason": "Same model family, but the bugs are different: missing vision config field vs. a from_pretrained error in Qwen3.5 MoE." }, { "left": "issue:45310", "right": "issue:45375", "accept": false, - "reason": "Different concrete problems: model-loading error vs missing deepstack_visual_indexes config field." + "reason": "Both involve Qwen3.5 MoE, but one is a from_pretrained failure and the other is a missing config field in the vision config." }, { "left": "issue:42831", "right": "issue:44222", "accept": false, - "reason": "FP8 accuracy drift is not the same as FP8 save_pretrained failure." + "reason": "Both mention FP8, but one is an accuracy regression and the other is a save_pretrained failure." }, { "left": "issue:42222", "right": "issue:45200", "accept": false, - "reason": "Unrelated models and symptoms; vitpose CI breakage is not the Gemma 4 token-id default issue." + "reason": "Unrelated model families and symptoms: broken vitpose models vs. Gemma 4 token-type defaults for text-only fine-tuning." }, { "left": "issue:41628", "right": "issue:44351", "accept": false, - "reason": "Separate import errors for different symbols with no shared code-path bug shown." + "reason": "Both are import errors, but for different symbols and likely different code paths." }, { "left": "issue:44222", "right": "issue:45310", "accept": false, - "reason": "Save-pretrained FP8/moe bug is distinct from Qwen3.5 MoE from_pretrained loading failure." + "reason": "Different concrete bugs: FP8 MoE save_pretrained vs. Qwen3.5 MoE from_pretrained loading." }, { "left": "issue:42994", "right": "issue:43824", "accept": false, - "reason": "Quantized save failure and missing Qwen2_5_VL import are unrelated." + "reason": "A quantized model save failure is not the same as an import error for a specific Qwen2.5 VL class." }, { "left": "issue:43377", "right": "issue:44242", "accept": false, - "reason": "One is batched-vs-single output mismatch from padding mask support; the other is MoE load-balancing loss logic." + "reason": "Different subsystems and failures: MIMI batching/padding-mask behavior vs. MoE load-balancing loss computation." }, { "left": "issue:36010", "right": "issue:41628", "accept": false, - "reason": "Both are import errors, but for different APIs and different failure causes." + "reason": "Different missing imports: GenerationMixin vs. AutoImageProcessor." }, { "left": "issue:42175", "right": "issue:45712", "accept": false, - "reason": "Packaging backend dependency issue is unrelated to leftover dummy class cleanup." + "reason": "Both touch packaging/availability, but one is missing tensorflow in torch extras and the other is leftover dummy classes in the no-torch path." }, { "left": "issue:44038", "right": "issue:45310", "accept": false, - "reason": "Different Qwen model variants and different failure modes." + "reason": "Related Qwen3 model family, but the reports describe different failures and different model variants." }, { - "left": "issue:44496", - "right": "issue:45561", + "left": "issue:44964", + "right": "issue:45137", "accept": false, - "reason": "Model config loading error and xdist file-race bug are unrelated." + "reason": "Completely different bugs: a multimodal model loading failure vs. a DeepSpeed ZeRO3 deque error." }, { - "left": "issue:44964", - "right": "issue:45137", + "left": "issue:44496", + "right": "issue:45561", "accept": false, - "reason": "Multimodal model load failure is not the same as the DeepSpeed ZeRO3 deque error." + "reason": "Unrelated: model config/model_type resolution vs. a pytest-xdist file race." }, { "left": "issue:45561", "right": "issue:45600", "accept": false, - "reason": "Test-worker file race and auto_mappings CI breakage are separate issues." + "reason": "Different issues: test-worker file contention vs. stale auto_mappings references breaking CI." }, { "left": "issue:42898", "right": "issue:44568", "accept": false, - "reason": "Both are tokenizer regressions in v5, but they affect different behavior and different tokenizers." + "reason": "Both are tokenizer behavior regressions in v5, but they affect different behaviors and tokenizers." }, { "left": "issue:30990", "right": "issue:35141", "accept": false, - "reason": "Loading hang and embedding reinitialization are different bugs." + "reason": "Sentence-transformer loading hang and token-embedding resize reinit are unrelated." }, { "left": "issue:43901", "right": "issue:45561", "accept": false, - "reason": "Docstring mismatch is unrelated to a pytest-xdist race." + "reason": "Documentation mismatch for TextClassificationPipeline is unrelated to the xdist captured_info.txt race." }, { "left": "issue:43232", "right": "issue:44792", "accept": false, - "reason": "Generation kwargs bug and janus image-generation test failure are not the same underlying defect." + "reason": "Different code paths: generation kwargs update after sync_gpus vs. a janus image-generation test failure." }, { "left": "issue:43232", "right": "issue:44188", "accept": false, - "reason": "Different code paths: generation state update vs torch.compile attention-kernel branching." + "reason": "Generation bookkeeping bug and attention-kernel divergence under torch.compile are not the same defect." }, { "left": "issue:44188", "right": "issue:44373", "accept": false, - "reason": "Attention-kernel divergence and a wrong docstring are unrelated." + "reason": "A kernel branching bug is unrelated to a docstring typo for position_ids." }, { "left": "issue:43278", "right": "issue:45137", "accept": false, - "reason": "Embedding dtype regression is unrelated to the DeepSpeed ZeRO3 deque underflow." + "reason": "Embedding dtype drift during evaluation is unrelated to the DeepSpeed ZeRO3 deque error." }, { "left": "issue:43784", "right": "issue:43824", "accept": false, - "reason": "Sentence-transformers import NameError and Qwen2_5_VL import error are separate import regressions." + "reason": "Both are import-related, but they concern different missing symbols and different package interactions." }, { "left": "issue:44623", "right": "issue:45310", "accept": false, - "reason": "Missing processor save files is not the same as Qwen3.5 MoE from_pretrained failure." + "reason": "Processor save_pretrained missing files is a serialization bug, not the Qwen3.5 MoE from_pretrained error." }, { "left": "issue:30333", "right": "issue:33290", "accept": false, - "reason": "MLflow reporting after failed training and DeepSpeed OOM with Adafactor are unrelated." + "reason": "MLflow reporting on failed training jobs is unrelated to an Adafactor/DeepSpeed OOM." }, { "left": "issue:44734", "right": "issue:45103", "accept": false, - "reason": "Serve KV-cache indexing crash and auto_docstring annotation crash do not share the same bug." + "reason": "Different failures: serving KV-cache continuation indexing vs. auto_docstring annotation handling." }, { "left": "issue:43723", "right": "issue:44568", "accept": false, - "reason": "Tokenizer loading failure and add_special_tokens BOS/EOS regression are different issues." + "reason": "Both relate to tokenizer behavior in v5, but they are different concrete regressions affecting different APIs." }, { "left": "issue:42898", "right": "issue:45356", "accept": false, - "reason": "Both are tokenizer regressions, but they concern different models and different tokenizer logic." + "reason": "Different tokenizer regressions: clean_up_tokenization_spaces behavior vs. Kimi-K2.5 codec handling." }, { "left": "issue:44060", "right": "issue:45125", "accept": false, - "reason": "Incorrect tied-weights warning and missing _tp_plan are different Qwen3 family issues." + "reason": "Both involve Qwen3.5 MoE, but one is a tied-weights warning and the other is a missing tensor-parallel plan." }, { - "left": "issue:43874", - "right": "issue:45092", + "left": "issue:42898", + "right": "issue:44779", "accept": false, - "reason": "Missing image-patch helper and remote-code meta-init incompatibility are unrelated multimodal failures." + "reason": "Different tokenizer problems with different models and symptoms." }, { - "left": "issue:42898", - "right": "issue:44779", + "left": "issue:43874", + "right": "issue:45092", "accept": false, - "reason": "General v5 tokenization-space behavior and DeepSeek codec handling are separate tokenizer bugs." + "reason": "Different multimodal/model-init issues: missing image-patch helper vs. remote-code checkpoint/meta-init incompatibility." }, { "left": "issue:45405", "right": "issue:45712", "accept": false, - "reason": "PEFT version pinning and dummy-class cleanup are unrelated maintenance issues." + "reason": "A dependency version bump problem is unrelated to dummy classes leaking into dir(transformers)." } ] }, @@ -11219,11 +11186,11 @@ "issue:41628|issue:42757", "issue:30333|issue:34689", "issue:43723|issue:44977", - "issue:44297|issue:45310", "issue:28282|issue:30333", + "issue:44297|issue:45310", "issue:30333|issue:35141", - "issue:44263|issue:44360", "issue:44222|issue:44484", + "issue:44263|issue:44360", "issue:44483|issue:45406", "issue:44263|issue:45092", "issue:43873|issue:44945", @@ -11231,244 +11198,250 @@ "issue:44514|issue:45103", "issue:43479|issue:43644", "issue:44295|issue:44623", - "issue:45092|issue:45357", "issue:43723|issue:44871", + "issue:45092|issue:45357", "issue:44261|issue:45092", "issue:43479|issue:44913", "issue:30990|issue:38175" ], - "prepared_review_unit_hash": "2bb41230cc12819e518b1b50a2fa7ddbacdb8afc53d49fc8768589ada6543025", + "prepared_review_unit_hash": "eddab6edd75b86aab8539e17c8d9e87c54865fb8b511ddc4e6821c8ee3ee452d", "component_budget": { "node_count": 387, "item_count": 387, "soft_pair_count": 1455, - "serialized_chars": 749552, - "estimated_input_tokens": 187388, - "estimated_eval_tokens": 375032 + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 35, - "serialized_chars": 38052, - "estimated_input_tokens": 9513, - "estimated_eval_tokens": 19282 + "serialized_chars": 38054, + "estimated_input_tokens": 9514, + "estimated_eval_tokens": 19284 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "No soft edge looks like a true duplicate: the list is a mix of unrelated regressions, import errors, config issues, and save/load bugs. The closest mini-cluster is around save_pretrained/serialization, but even those are distinct failure modes.", - "confidence": 0.36, - "canonical_issue_reason": "issue:44623 is the best anchor for the only coherent subtheme here (save/load serialization regressions): it is a concrete, actionable failure with a broad enough surface to sit near other save_pretrained-related reports, while the rest of the set is too heterogeneous for a stronger canonical.", + "summary": "This cluster is a loose collection of Transformers issues around loading/saving, config serialization, processor/tokenizer behavior, and a few unrelated training/runtime bugs. I don\u2019t see any soft pair that is clearly the same underlying defect, so no duplicate merges are justified.", + "confidence": 0.27, + "canonical_issue_reason": "If a single representative is needed, issue 44297 is the most central of the serialization-related subset: it is specific, reproducible, and closest to several save/load-related complaints. Still, the cluster is too heterogeneous for a true canonical duplicate.", "canonical_pr_reason": null, - "best_issue_reason": "issue:44623 is the best overall representative for triage because it is specific, user-facing, and central to the save_pretrained cluster; however, it should not be treated as a true umbrella duplicate for the whole set.", + "best_issue_reason": "Issue 44297 is the best standalone anchor because it is concrete and sits at the intersection of the strongest soft-similarity links, but it should be treated as a thematic representative rather than a confirmed duplicate hub.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:42898", "right": "issue:44448", "accept": false, - "reason": "Both involve output/behavior changes, but one is a global v5 tokenization-space cleanup change and the other is a model-specific v4/v5 output difference." + "reason": "Both mention v5 output behavior changes, but one is tokenizer cleanup behavior and the other is Pegasus output drift; not the same code-path bug." }, { "left": "issue:43479", "right": "issue:44079", "accept": false, - "reason": "Different code paths: Phi4MultimodalConfig default initialization vs ModelOutput key assignment when a field was previously None." + "reason": "Multimodal config initialization with None is unrelated to ModelOutput key assignment semantics." }, { "left": "issue:44188", "right": "issue:44964", "accept": false, - "reason": "Unrelated bugs: torch.compile attention-kernel branching vs loading a specific multimodal model." + "reason": "Attention-kernel branching under torch.compile is unrelated to a Phi-4 multimodal loading failure." }, { "left": "issue:44295", "right": "issue:45656", "accept": false, - "reason": "Different problems: reading a registered buffer fails vs deepspeed calling optimizer.step twice." + "reason": "A position_ids buffer read error is not the same as a double optimizer-step bug under deepspeed." }, { "left": "issue:42898", "right": "issue:43295", "accept": false, - "reason": "Both are regressions, but one is tokenizer cleanup behavior and the other is a processor/tokenizer API break in v4.57.5." + "reason": "Both are regressions, but cleanup_spaces and processor.tokenizer/image handling are different failure modes and code paths." }, { "left": "issue:44933", "right": "issue:45712", "accept": false, - "reason": "Both touch imports, but one is a missing image_utils import and the other is leftover dummy classes leaking into torch-less namespace handling." + "reason": "Both touch exports/imports, but one is a missing image_utils import and the other is dummy PT stubs leaking into the public API." }, { "left": "issue:42175", "right": "issue:44933", "accept": false, - "reason": "Different subsystems: backend extras installation vs an import/export problem in image_utils." + "reason": "Backend packaging for pip extras is unrelated to an import-export issue in image_utils." }, { "left": "issue:44222", "right": "issue:44297", "accept": false, - "reason": "Both mention save_pretrained, but one is FP8/MoE model saving and the other is tokenizer metadata mismatch." + "reason": "Both involve save_pretrained, but one is FP8 MoE model saving and the other is tokenizer config metadata; different artifacts and bugs." }, { "left": "issue:43976", "right": "issue:44933", "accept": false, - "reason": "A Python version compatibility issue is not the same as a missing import from image_utils." + "reason": "A Python version compatibility issue is unrelated to a missing import from image_utils." }, { "left": "issue:44297", "right": "issue:45656", "accept": false, - "reason": "Tokenizer save metadata mismatch and duplicate optimizer stepping are unrelated failures." + "reason": "Tokenizer config mismatch on save is unrelated to an optimizer-step duplication problem." }, { "left": "issue:43479", "right": "issue:44869", "accept": false, - "reason": "Phi4MultimodalConfig defaults and Whisper timestamp decoding are separate code paths and symptoms." + "reason": "Multimodal config defaults and Whisper word-timestamp decoding are unrelated." }, { "left": "issue:44206", "right": "issue:45103", "accept": false, - "reason": "One is a feature regression in LasrFeatureExtractor, the other is an auto-docstring crash under future annotations." + "reason": "A feature regression in LasrFeatureExtractor is unrelated to an auto-docstring crash from future annotations." }, { "left": "issue:42831", "right": "issue:44623", "accept": false, - "reason": "Accuracy degradation with FineGrainedFP8 is not the same as processor.save_pretrained missing files." + "reason": "FP8 accuracy issues and processor.save_pretrained missing files are different problems." }, { "left": "issue:44222", "right": "issue:44623", "accept": false, - "reason": "Both are save_pretrained-related, but the underlying bug differs: FP8/MoE saving versus missing processor files." + "reason": "Both mention saving, but model FP8 save_pretrained and processor.save_pretrained file omission are not the same bug." }, { "left": "issue:30333", "right": "issue:36331", "accept": false, - "reason": "MLflow reporting on failed jobs and a Trainer signature change are unrelated training issues." + "reason": "MLFlow reporting on failed jobs is unrelated to a Trainer compute_loss signature regression." }, { "left": "issue:41628", "right": "issue:42757", "accept": false, - "reason": "Both are import errors, but for different symbols and modules with no shared underlying bug." + "reason": "Missing AutoImageProcessor and missing is_offline_mode are separate import breakages in different modules." }, { "left": "issue:30333", "right": "issue:34689", "accept": false, - "reason": "MLflow job-status reporting and Llama 3.2 model loading are unrelated." + "reason": "MLFlow job reporting and Llama 3.2 model-loading breakage do not share a concrete code path." }, { "left": "issue:43723", "right": "issue:44977", "accept": false, - "reason": "Tokenizer loading in v5 and flash-attention generation behavior for Qwen3.5 are different problems." + "reason": "Tokenizer loading in v5 and Qwen3.5 flash-attention generation are different subsystems." }, { - "left": "issue:44297", - "right": "issue:45310", + "left": "issue:28282", + "right": "issue:30333", "accept": false, - "reason": "Tokenizer save_pretrained metadata mismatch is not the same bug as a from_pretrained error on Qwen3.5 MoE." + "reason": "A missing PyTorch ImportError and MLFlow reporting for failed training jobs are unrelated." }, { - "left": "issue:28282", - "right": "issue:30333", + "left": "issue:44297", + "right": "issue:45310", "accept": false, - "reason": "PyTorch import availability and MLflow reporting are unrelated." + "reason": "Both are serialization/load regressions, but tokenizer_class mismatch and Qwen3.5 MoE from_pretrained failure are not the same defect." }, { "left": "issue:30333", "right": "issue:35141", "accept": false, - "reason": "Failed-job reporting in MLflow has no overlap with embeddings being reinitialized in post_init." + "reason": "Failed-job MLFlow reporting and embedding reinitialization in post_init are unrelated." }, { - "left": "issue:44263", - "right": "issue:44360", + "left": "issue:44222", + "right": "issue:44484", "accept": false, - "reason": "Same DSA indexer area, but one is about torch.split return values and the other is about a missing ReLU; not the same defect." + "reason": "FP8 model save_pretrained behavior is unrelated to the max_shard_size default question." }, { - "left": "issue:44222", - "right": "issue:44484", + "left": "issue:44263", + "right": "issue:44360", "accept": false, - "reason": "Both reference save_pretrained, but one is FP8/MoE saving and the other is the default max_shard_size behavior." + "reason": "A torch.split return-value issue in one indexer is not the same as a missing ReLU in another." }, { "left": "issue:44483", "right": "issue:45406", "accept": false, - "reason": "API request rejection in /v1/chat/completions is unrelated to Gemma4Processor missing _tokenizer." + "reason": "An API request rejection in v5.3 is unrelated to Gemma4Processor missing _tokenizer during serve." }, { "left": "issue:44263", "right": "issue:45092", "accept": false, - "reason": "A torch.split indexing issue is not the same as remote-code checkpoint incompatibility with meta init." + "reason": "A DSA indexer torch.split bug is unrelated to remote-code checkpoint incompatibility with meta init." }, { "left": "issue:43873", "right": "issue:44945", "accept": false, - "reason": "Quantization offloading and pipeline-parallel output correctness are different distributed-inference bugs." + "reason": "Quantization offloading behavior and pipeline-parallel output correctness are different runtime problems." }, { "left": "issue:44373", "right": "issue:44485", "accept": false, - "reason": "A wrong docstring and a RoPE implementation discussion are not the same code-path problem." + "reason": "A docstring error for position_ids is not the same as a RoPE implementation issue." }, { "left": "issue:44514", "right": "issue:45103", "accept": false, - "reason": "Processor batching/padding crash and future-annotations docstring crash are unrelated." + "reason": "A batched chat-template crash and an auto-docstring AttributeError are unrelated." }, { "left": "issue:43479", "right": "issue:43644", "accept": false, - "reason": "Config default initialization and non-persistent buffer corruption are different regression classes." + "reason": "Multimodal config default handling is unrelated to non-persistent buffer initialization junk." }, { "left": "issue:44295", "right": "issue:44623", "accept": false, - "reason": "Reading position_ids after buffer registration is not the same as processor.save_pretrained missing files." + "reason": "Reading a registered buffer and missing processor save files are different serialization problems." }, { - "left": "issue:45092", - "right": "issue:45357", + "left": "issue:43723", + "right": "issue:44871", "accept": false, - "reason": "Loading old remote-code checkpoints under v5 meta init is different from save_pretrained writing incorrect visual encoder keys." + "reason": "Tokenizer loading in v5 and eos_token_id configuration inconsistency are not the same issue." }, { - "left": "issue:43723", - "right": "issue:44871", + "left": "issue:45092", + "right": "issue:45357", "accept": false, - "reason": "Tokenizer loading failure and eos_token_id inconsistency are separate issues." + "reason": "Remote-code meta-init incompatibility and incorrect visual encoder keys on save_pretrained are related only loosely by model loading, not by the same bug." }, { "left": "issue:44261", "right": "issue:45092", "accept": false, - "reason": "Missing rms_norm_eps in MLA q_a_layernorm is unrelated to remote-code checkpoint/meta-init incompatibility." + "reason": "A precision error from missing rms_norm_eps is unrelated to meta-init incompatibility for old checkpoints." + }, + { + "left": "issue:43479", + "right": "issue:44913", + "accept": false, + "reason": "Both involve config persistence, but None-handling in a multimodal config and rotary_pct reset on reload are different fields and code paths." }, { "left": "issue:30990", "right": "issue:38175", "accept": false, - "reason": "Sentence Transformers getting stuck loading and unexpected zero probabilities in SigLIP2 are different model-specific bugs." + "reason": "Sentence Transformers loading stalls and SIGLIP2 zero probabilities are unrelated model/runtime problems." } ] }, @@ -11565,49 +11538,49 @@ "issue:43723|issue:44779", "issue:43723|issue:44442", "issue:43784|issue:45600", - "issue:44261|issue:44373", "issue:43122|issue:44521", "issue:42907|issue:43937", - "issue:43408|issue:43475", + "issue:44261|issue:44373", "issue:41628|issue:45278", + "issue:43408|issue:43475", "issue:44360|issue:44485", "issue:43232|issue:44743", "issue:43116|issue:45200" ], - "prepared_review_unit_hash": "359c33099548d9e59335084a2f2a46340484981a40be452da26e09445cb88ec1", + "prepared_review_unit_hash": "2bae3e6c3cce5a150a630b0ed494c7cafac813c167e8de5f00a86d26a518588c", "component_budget": { "node_count": 387, "item_count": 387, "soft_pair_count": 1455, - "serialized_chars": 749552, - "estimated_input_tokens": 187388, - "estimated_eval_tokens": 375032 + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 31, - "serialized_chars": 36503, - "estimated_input_tokens": 9126, - "estimated_eval_tokens": 18508 + "serialized_chars": 36499, + "estimated_input_tokens": 9125, + "estimated_eval_tokens": 18506 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly heterogeneous: many are unrelated import, tokenizer, generation, or model-specific regressions. I do not see any true duplicate pairs, so the best candidate is only a loose hub for the broad import-regression subset.", - "confidence": 0.44, - "canonical_issue_reason": "issue:45278 is the broadest and most central report in the set (\u201cmany import errors after update from 4.57.0 to 5.5.0\u201d), so it works best as a loose hub for the import-regression subgroup, though it is not a true duplicate of most others.", + "summary": "This cluster is mostly a set of transformer v5-era regressions and import/tokenizer issues, but the soft pairs are still too heterogeneous to treat as duplicates. I would keep one broad tokenizer-loading regression as the representative issue, but reject all soft edges.", + "confidence": 0.86, + "canonical_issue_reason": "issue:43723 is the best representative because it is a broad, concrete v5 tokenizer-loading failure centered on `AutoTokenizer.from_pretrained`, which sits near several of the other tokenizer/import regressions in this cluster.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45278 is the best overall representative because it captures the broadest symptom category in this cluster and is recent; the remaining issues are mostly narrower, model- or API-specific bugs.", + "best_issue_reason": "issue:43723 is the strongest cluster anchor: it is specific enough to be actionable, yet broad enough to cover the common tokenizer-loading regression theme better than the narrower one-off reports.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:22355", "right": "issue:28282", "accept": false, - "reason": "Different failure modes: missing transformers.onnx module vs AutoModel/PyTorch dependency import error." + "reason": "Different failure modes: missing `transformers.onnx` module vs missing PyTorch for `AutoModel`." }, { "left": "issue:42886", @@ -11619,175 +11592,175 @@ "left": "issue:41628", "right": "issue:43824", "accept": false, - "reason": "Both are import errors, but for different missing symbols/models; no clear same underlying bug." + "reason": "Both are import errors, but for different missing symbols and different model APIs." }, { "left": "issue:42831", "right": "issue:45310", "accept": false, - "reason": "Different models and symptoms: FP8 accuracy issue vs Qwen3.5 MoE from_pretrained failure." + "reason": "FineGrainedFP8 accuracy issue is unrelated to the Qwen3.5-Moe `from_pretrained` error." }, { "left": "issue:44263", "right": "issue:44373", "accept": false, - "reason": "DSA indexer split behavior vs a docstring issue; not the same bug." + "reason": "A runtime bug in `torch.split` is not the same as a wrong docstring." }, { "left": "issue:44945", "right": "issue:45753", "accept": false, - "reason": "Pipeline-parallelism wrong outputs and a Qwen3.5 image loop are distinct runtime problems." + "reason": "Pipeline parallelism output corruption and a Qwen3_5 image-specific loop are different bugs." }, { "left": "issue:42831", "right": "issue:44038", "accept": false, - "reason": "FP8 accuracy regression is not the same as Qwen3-VL-Moe compatibility/bug report." + "reason": "Different models and symptoms: FP8 accuracy vs Qwen3-VL import/runtime bug." }, { "left": "issue:42757", "right": "issue:45278", "accept": false, - "reason": "One is a specific huggingface_hub import breakage; the other is a broad umbrella report, not the same concrete bug." + "reason": "One is a specific missing import from `huggingface_hub`; the other is a broad post-upgrade import report." }, { "left": "issue:43576", "right": "issue:43784", "accept": false, - "reason": "Broken env command and sentence-transformers import NameError are unrelated." + "reason": "`transformers env` is broken, while the other is a `NameError` during sentence-transformers import." }, { "left": "issue:42886", "right": "issue:43066", "accept": false, - "reason": "Offline cache behavior and tokenizer decoder type mismatch are different tokenizer issues." + "reason": "Both involve tokenizers, but cache/offline loading is not the same as wrong decoder type in v5." }, { "left": "issue:44117", "right": "issue:44704", "accept": false, - "reason": "TOKENIZER_MAPPING_NAMES returning None and AutoProcessor kwargs forwarding are separate internals." + "reason": "`TOKENIZER_MAPPING_NAMES` returning `None` and `AutoProcessor` not forwarding kwargs are different bugs." }, { "left": "issue:43232", "right": "issue:45593", "accept": false, - "reason": "Generation kwargs update after sync_gpus and D-FINE auxiliary-loss behavior are unrelated." + "reason": "Generation kwargs handling and D-FINE auxiliary-loss behavior are unrelated." }, { "left": "issue:42371", "right": "issue:44263", "accept": false, - "reason": "TF32 configuration guidance and a torch.split bug do not share the same code path." + "reason": "TF32 API guidance and a `torch.split` indexing bug do not match." }, { "left": "issue:43756", "right": "issue:44060", "accept": false, - "reason": "Smollm3 RoPE layer count and Qwen3-Next tied-weights warning are different model implementation issues." + "reason": "Smollm3 RoPE-layer issue and Qwen3-Next tied-weights warning are different model-specific problems." }, { "left": "issue:30990", "right": "issue:33357", "accept": false, - "reason": "Loading hang in Sentence Transformers and a Mac bus error on CLIP are not the same defect." + "reason": "Sentence-transformers loading stall and a MacOS bus error with CLIP are not the same underlying bug." }, { "left": "issue:44704", "right": "issue:45290", "accept": false, - "reason": "AutoProcessor cached_file kwargs and chat-template assistant message handling are unrelated." + "reason": "`AutoProcessor.from_pretrained` kwargs handling and chat-template tool-call crashes are separate code paths." }, { "left": "issue:42907", "right": "issue:44945", "accept": false, - "reason": "Quantized model saving failure and pipeline-parallelism output corruption are different bugs." + "reason": "Quantized model saving failure and pipeline-parallelism output errors are unrelated." }, { "left": "issue:43122", "right": "issue:44514", "accept": false, - "reason": "Tokenization differences across versions and batched chat-template crash are distinct." + "reason": "Tokenizer behavior changed between versions, but the batched chat-template crash is a different failure mode." }, { "left": "issue:44295", "right": "issue:44297", "accept": false, - "reason": "Reading a registered buffer and tokenizer_config class mismatch are unrelated save/load issues." + "reason": "A buffer-read error on `position_ids` is distinct from a tokenizer class mismatch in saved config." }, { "left": "issue:43232", "right": "issue:45245", "accept": false, - "reason": "Generation bookkeeping bug and category-count runtime error are not the same code-path problem." + "reason": "Generation state update after `sync_gpus` has nothing to do with category-count limits." }, { "left": "issue:43723", "right": "issue:44779", "accept": false, - "reason": "General AutoTokenizer load failure and Deepseek tokenizer output regression are different tokenizer bugs." + "reason": "Both are v5 tokenizer regressions, but one is loading/import failure and the other is incorrect tokenization output." }, { "left": "issue:43723", "right": "issue:44442", "accept": false, - "reason": "Both mention AutoTokenizer, but one is a general v5 regression and the other a specific tokenizer-class load failure." + "reason": "`AutoTokenizer` load failure and FastSpeech2Conformer tokenizer loading failure are different registration issues." }, { "left": "issue:43784", "right": "issue:45600", "accept": false, - "reason": "Sentence-transformers import NameError and removed Sam3LiteText auto-mapping references are unrelated." - }, - { - "left": "issue:44261", - "right": "issue:44373", - "accept": false, - "reason": "Config epsilon precision issue and docstring mismatch do not indicate the same defect." + "reason": "Sentence-transformers import `NameError` and Sam3LiteText auto-mapping CI breakage are unrelated." }, { "left": "issue:43122", "right": "issue:44521", "accept": false, - "reason": "Tokenizer output changes and multimodal assistant-mask zeros are different chat/tokenization behaviors." + "reason": "Tokenization differences and zeroed assistant masks in multimodal chat templates are different bugs." }, { "left": "issue:42907", "right": "issue:43937", "accept": false, - "reason": "Quantized save failure and invalid GenerationConfig in GLM-5 are separate issues." + "reason": "Saving a dequantized model and invalid `GenerationConfig` validation are unrelated." }, { - "left": "issue:43408", - "right": "issue:43475", + "left": "issue:44261", + "right": "issue:44373", "accept": false, - "reason": "sam3_video vs sam3_tracker warning and missing fpn_position_embeddings are different SAM3 problems." + "reason": "Missing `rms_norm_eps` in MLA config and a wrong docstring for `position_ids` are not the same bug." }, { "left": "issue:41628", "right": "issue:45278", "accept": false, - "reason": "AutoImageProcessor import failure may be one instance of broad import regressions, but the reports are not the same concrete bug." + "reason": "A single missing import and a broad upgrade-wide import regression are not a duplicate pair." + }, + { + "left": "issue:43408", + "right": "issue:43475", + "accept": false, + "reason": "Same SAM 3 area, but different problems: model-type warning vs missing attribute on output object." }, { "left": "issue:44360", "right": "issue:44485", "accept": false, - "reason": "A missing ReLU in the DSA indexer and GLM-5 RoPE implementation are unrelated." + "reason": "DSA indexer activation choice and GLM-5 RoPE implementation are unrelated." }, { "left": "issue:43232", "right": "issue:44743", "accept": false, - "reason": "Generation kwargs update and Qwen3.5 recurrent-state reset are different generation bugs." + "reason": "Generation kwargs handling and recurrent-state reset in modular Qwen3.5 are different code paths." }, { "left": "issue:43116", "right": "issue:45200", "accept": false, - "reason": "Multi-label classification example output and mm_token_type_ids defaulting are unrelated application/model bugs." + "reason": "Multi-label classification example behavior and Gemma 4 token-type defaults are unrelated." } ] }, @@ -11879,14 +11852,14 @@ "issue:42757|issue:44351", "issue:43122|issue:44977", "issue:43992|issue:44704", - "issue:42907|issue:44222", "issue:43784|issue:45446", + "issue:42907|issue:44222", "issue:43937|issue:44485", "issue:42222|issue:43441", - "issue:42371|issue:43874", "issue:43976|issue:45446", - "issue:43493|issue:45753", + "issue:42371|issue:43874", "issue:44075|issue:44936", + "issue:43493|issue:45753", "issue:42831|issue:45753", "issue:43116|issue:43825", "issue:44038|issue:45478", @@ -11896,238 +11869,238 @@ "issue:45083|issue:45446", "issue:43937|issue:45563" ], - "prepared_review_unit_hash": "01f4474598c6bfb94ffe38a642689dc21c2e6b5a3fbef551f95b0d11c4ed4872", + "prepared_review_unit_hash": "b96dcd1d307b5698630cfa6f662cf9dbd5a22b3abf3a393be358a69349b2d2a5", "component_budget": { "node_count": 387, "item_count": 387, "soft_pair_count": 1455, - "serialized_chars": 749552, - "estimated_input_tokens": 187388, - "estimated_eval_tokens": 375032 + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 34, - "serialized_chars": 37411, - "estimated_input_tokens": 9353, - "estimated_eval_tokens": 18962 + "serialized_chars": 37417, + "estimated_input_tokens": 9355, + "estimated_eval_tokens": 18966 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is heterogeneous: the items span unrelated tokenizer, loading, training, saving, multimodal, and version-compatibility problems. None of the soft pairs look like the same underlying bug or change, so I reject all soft edges and do not see a meaningful canonical duplicate target.", - "confidence": 0.97, - "canonical_issue_reason": "No single canonical issue: the cluster mixes unrelated defects across many subsystems, so it does not form a duplicate set.", + "summary": "This cluster is a loose collection of unrelated Transformers issues. The soft-similarity pairs mostly share broad themes like tokenization, loading, training, or import errors, but they do not look like the same concrete underlying bug/change.", + "confidence": 0.93, + "canonical_issue_reason": "No single canonical issue fits the cluster: the items span unrelated failures across tokenization, model loading, training, multimodal processing, and API/import regressions.", "canonical_pr_reason": null, - "best_issue_reason": "No global best issue; none of the issues is representative of a coherent duplicate cluster.", + "best_issue_reason": "Issue 44779 is the clearest concrete regression-style report among the set, but only as a representative issue; it is not a true duplicate hub for the whole cluster.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44625", "right": "issue:44829", "accept": false, - "reason": "Different problems: config propagation for Qwen3.5 vs flash-attention training degeneration." + "reason": "One is a Qwen3.5 num_labels propagation bug; the other is a flash_attention_3 training degeneration issue. Different code paths and symptoms." }, { "left": "issue:29127", "right": "issue:30990", "accept": false, - "reason": "Unrelated subsystems and failures: LayoutLMv3 error messaging vs Sentence-Transformers loading hang." + "reason": "LayoutLMv3 error-message clarity and Sentence-Transformers loading hang are unrelated problems." }, { "left": "issue:43122", "right": "issue:44779", "accept": false, - "reason": "Both are tokenizer regressions, but they affect different models and likely different code paths." + "reason": "Both concern tokenizer behavior in v5, but one is a general version-to-version tokenization difference and the other is a DeepSeek-specific regression. Too broad to call the same bug." }, { "left": "issue:42907", "right": "issue:44964", "accept": false, - "reason": "Different issues: saving dequantized Ministral/Devstral weights vs loading Phi-4 multimodal." + "reason": "Saving dequantized Ministral/Devstral models is a different failure from loading Phi-4 multimodal." }, { "left": "issue:44038", "right": "issue:44297", "accept": false, - "reason": "Model loading bug vs tokenizer save metadata mismatch; not the same concrete defect." + "reason": "Qwen3-VL-Moe import/runtime issue versus tokenizer_class mismatch on save_pretrained; not the same failure." }, { "left": "issue:42907", "right": "issue:43452", "accept": false, - "reason": "Different paths: save_pretrained/dequantized model export vs AutoTokenizer/AutoModel loading gguf_file." + "reason": "Model save/dequantization failure and gguf_file load breakage are distinct loading/saving paths." }, { "left": "issue:44038", "right": "issue:44295", "accept": false, - "reason": "Different failures: Qwen3-VL-Moe model support vs position_ids buffer read error." + "reason": "Qwen3-VL-Moe bug and position_ids buffer read error are unrelated." }, { "left": "issue:39401", "right": "issue:44488", "accept": false, - "reason": "Tokenizer offset mapping bug vs model loading failure for a specific checkpoint." + "reason": "Qwen3 tokenizer offset_mapping bug and loading cjvt/sleng-bert are different tokenizer/model-loading issues." }, { "left": "issue:36296", "right": "issue:43116", "accept": false, - "reason": "Training parallelism bug vs example-script classification output issue." + "reason": "Tensor-parallel training bug and multi-label classification example failure are unrelated." }, { "left": "issue:43824", "right": "issue:44351", "accept": false, - "reason": "Both are import errors, but for different symbols with different root causes." + "reason": "Qwen2.5-VL import failure and HybridCache import failure are separate missing-symbol regressions." }, { "left": "issue:44038", "right": "issue:44222", "accept": false, - "reason": "Same broad model family, but one is a model loading/generation issue and the other is FP8 save_pretrained behavior." + "reason": "Qwen3-VL-Moe runtime issue versus FP8 save_pretrained moe issue; different subsystems." }, { "left": "issue:43066", "right": "issue:44779", "accept": false, - "reason": "Both concern tokenizer behavior, but one is decoder type handling and the other is Deepseek tokenization correctness." + "reason": "Wrong tokenizer decoder type and DeepSeek tokenizer incorrect results are both tokenizer-related, but they target different concrete defects." }, { "left": "issue:39401", "right": "issue:44521", "accept": false, - "reason": "Offset_mapping regression and assistant_masks behavior are separate tokenizer/chat-template bugs." + "reason": "Wrong offset_mapping and all-zero assistant_masks are different tokenizer/template behaviors." }, { "left": "issue:22355", "right": "issue:28218", "accept": false, - "reason": "Completely unrelated: missing transformers.onnx module vs extra space after added token." + "reason": "Missing transformers.onnx module and added-token spacing bug are unrelated." }, { "left": "issue:45216", "right": "issue:45478", "accept": false, - "reason": "Both mention Qwen3.5 and v5.4.0, but one is save_pretrained checkpoint correctness and the other is from_pretrained loading." + "reason": "Qwen3.5 save_pretrained checkpoint regression and Qwen3.5 Moe from_pretrained error are related to the same model family but not the same bug." }, { "left": "issue:42757", "right": "issue:44351", "accept": false, - "reason": "Import errors from different packages/symbols; not the same issue." + "reason": "Import errors from different missing symbols in different libraries; no shared code-path." }, { "left": "issue:43122", "right": "issue:44977", "accept": false, - "reason": "Tokenizer regression vs flash-attention generation bug; different code paths." + "reason": "General tokenizer regression in v5 versus Qwen3.5 flash-attention generation issue; different failures." }, { "left": "issue:43992", "right": "issue:44704", "accept": false, - "reason": "Different loading bugs: missing embed_tokens.weight vs AutoProcessor not forwarding kwargs." + "reason": "UMT5Encoder missing embed_tokens.weight and AutoProcessor kwargs forwarding are unrelated loading bugs." }, { - "left": "issue:42907", - "right": "issue:44222", + "left": "issue:43784", + "right": "issue:45446", "accept": false, - "reason": "Both involve model saving, but one is dequantized export failure and the other is FP8 save_pretrained for MoE." + "reason": "Sentence-Transformers import NameError and flex_attention version-check bug are unrelated." }, { - "left": "issue:43784", - "right": "issue:45446", + "left": "issue:42907", + "right": "issue:44222", "accept": false, - "reason": "Import-time NameError in sentence-transformers vs incorrect PyTorch version gating for flex_attention." + "reason": "Dequantized model save failure and FP8 moe save_pretrained bug are both saving-related, but not the same concrete problem." }, { "left": "issue:43937", "right": "issue:44485", "accept": false, - "reason": "Same model family (GLM-5) but different defects: invalid GenerationConfig vs RoPE implementation." + "reason": "GenerationConfig validation error and GLM-5 RoPE implementation discussion are different issues." }, { "left": "issue:42222", "right": "issue:43441", "accept": false, - "reason": "Unrelated model breakage: vitpose import/path issue vs Ministral FlashAttention failure." - }, - { - "left": "issue:42371", - "right": "issue:43874", - "accept": false, - "reason": "TF32 API guidance issue vs missing image-patch helper in GLM46V image processor." + "reason": "VitPose import breakage and Ministral-3 FlashAttention failure are unrelated model-specific regressions." }, { "left": "issue:43976", "right": "issue:45446", "accept": false, - "reason": "Different compatibility checks: Python version support vs PyTorch version check." + "reason": "Python version compatibility report and flex_attention import/version check bug do not match." }, { - "left": "issue:43493", - "right": "issue:45753", + "left": "issue:42371", + "right": "issue:43874", "accept": false, - "reason": "Different model families and symptoms: SigLIP2 implementation discrepancy vs Qwen3_5 infinite loop on one image." + "reason": "TF32 API guidance and missing get_number_of_image_patches are unrelated." }, { "left": "issue:44075", "right": "issue:44936", "accept": false, - "reason": "SGD argument handling vs evaluate-after-train failure are separate training issues." + "reason": "Optimizer SGD args ignored and trainer.evaluate() failing after train() are distinct training issues." + }, + { + "left": "issue:43493", + "right": "issue:45753", + "accept": false, + "reason": "SigLIP2 HF vs JAX discrepancy and Qwen3_5 infinite loop on a specific image are different model problems." }, { "left": "issue:42831", "right": "issue:45753", "accept": false, - "reason": "FineGrainedFP8 accuracy issue is unrelated to Qwen3_5 image-triggered infinite loop." + "reason": "FineGrainedFP8 accuracy issue and Qwen3_5 image-specific infinite loop are unrelated." }, { "left": "issue:43116", "right": "issue:43825", "accept": false, - "reason": "Example script output bug vs pipeline error-message wording; not the same defect." + "reason": "Example script multi-label output bug and pipeline translation-task error message bug are different example/API issues." }, { "left": "issue:44038", "right": "issue:45478", "accept": false, - "reason": "Both mention Qwen3.5/MoE-ish regressions, but one is Qwen3-VL-Moe in 5.0 and the other is a 5.4.0 from_pretrained error." + "reason": "Both involve Qwen3.5/Qwen3.5-Moe, but one is a load/runtime bug and the other is a from_pretrained error; not clearly the same defect." }, { "left": "issue:44222", "right": "issue:45478", "accept": false, - "reason": "Same MoE theme, but one is FP8 save_pretrained and the other is model loading; too different to merge." + "reason": "FP8 save_pretrained moe and Qwen3.5 Moe from_pretrained error are separate failure modes." }, { "left": "issue:43116", "right": "issue:43493", "accept": false, - "reason": "Classification example bug and SigLIP2 implementation discrepancy are unrelated." + "reason": "Multi-label classification example failure and SigLIP2 implementation discrepancy are unrelated." }, { "left": "issue:43064", "right": "issue:43673", "accept": false, - "reason": "Training-state bug under FSDP2/PEFT vs GenerationMixin cache missing during chunked prefill." + "reason": "Wrong optimizer states/params under FSDP2+PEFT and missing GenerationMixin cache in chunked_prefill are different training vs generation bugs." }, { "left": "issue:45083", "right": "issue:45446", "accept": false, - "reason": "Feature-extract helper behavior in qwen3_omni_moe vs flex_attention version-check bug." + "reason": "Helper length computation in qwen3_omni_moe and flex_attention import/version check are unrelated." }, { "left": "issue:43937", "right": "issue:45563", "accept": false, - "reason": "GLM-5 GenerationConfig validity vs stale warning for num_return_sequences are unrelated." + "reason": "GenerationConfig invalid and stale warning for num_return_sequences are different generation issues with different symptoms." } ] }, @@ -12204,8 +12177,8 @@ "issue:38617|issue:45278", "issue:33453|issue:39401", "issue:39401|issue:44779", - "issue:22355|issue:29127", "issue:44117|issue:45588", + "issue:22355|issue:29127", "issue:43827|issue:45446", "issue:43381|issue:44945", "issue:43937|issue:44977", @@ -12216,8 +12189,8 @@ "issue:42673|issue:45072", "issue:38617|issue:41628", "issue:44829|issue:45216", - "issue:44291|issue:45071", "issue:44410|issue:45310", + "issue:44291|issue:45071", "issue:43408|issue:43742", "issue:42222|issue:45542", "issue:44484|issue:45310", @@ -12230,14 +12203,14 @@ "issue:43066|issue:43122", "issue:38617|issue:44220" ], - "prepared_review_unit_hash": "dd008e89a4ec6cb660a05506002a66337b259ae8cab9755d94244a1d7955e5c4", + "prepared_review_unit_hash": "da2aab9401052791b8656bcadb5ced002e18ffc1ff965ef17eb2c7ab02c3e982", "component_budget": { "node_count": 387, "item_count": 387, "soft_pair_count": 1455, - "serialized_chars": 749552, - "estimated_input_tokens": 187388, - "estimated_eval_tokens": 375032 + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 }, "budget": { "node_count": 47, @@ -12252,186 +12225,186 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These are mostly distinct, model-specific or subsystem-specific bugs; none of the soft pairs look like the same underlying issue. The closest-looking ones are broad version-regression/import-error reports, but still not tight enough to merge.", - "confidence": 0.93, - "canonical_issue_reason": "Issue 45278 is the broadest umbrella report (many import errors after upgrading), so it is the most plausible representative issue; the other items are narrower, unrelated failures.", + "summary": "The set is heterogeneous: it contains import regressions, tokenizer bugs, model-loading failures, dtype/backend issues, and docs/problems in unrelated subsystems. None of the soft pairs look like the same underlying bug, so I reject all merges.", + "confidence": 0.96, + "canonical_issue_reason": "No single canonical issue fits the cluster; the items span unrelated bug classes and should not be collapsed into one duplicate bucket.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 45278 is the best global representative because it covers a broad regression class, unlike the others which are more specific to one model, endpoint, or code path.", + "best_issue_reason": "If forced to pick an umbrella, issue 45278 is the broadest import-error regression after upgrade, but it is still too general to serve as a safe duplicate canonical.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:42175", "right": "issue:44938", "accept": false, - "reason": "Different failures: missing TensorFlow in a torch-only install vs Python 3.14 load failure. No shared concrete bug." + "reason": "Both are import-related, but one is a missing TensorFlow backend in a torch install and the other is a Python 3.14 load failure." }, { "left": "issue:38617", "right": "issue:45278", "accept": false, - "reason": "Both are import-related, but one is a missing symbol from configuration_utils and the other is a broad multi-import regression after upgrade. Too broad to treat as one bug." + "reason": "Specific missing-symbol import vs a broad report of many import errors after upgrade; not the same concrete bug." }, { "left": "issue:33453", "right": "issue:39401", "accept": false, - "reason": "Both involve tokenization, but one is a tokenizer-loading regression and the other is wrong offset_mapping for Qwen3. Different code paths." + "reason": "Tokenizer loading regression and Qwen3 offset-mapping bug are different failure modes." }, { "left": "issue:39401", "right": "issue:44779", "accept": false, - "reason": "Both are tokenizer output issues, but they affect different models and symptoms. Not the same underlying defect." + "reason": "Both involve tokenization, but they target different models and symptoms, so this is too broad to merge." }, { - "left": "issue:22355", - "right": "issue:29127", + "left": "issue:44117", + "right": "issue:45588", "accept": false, - "reason": "Unrelated problems: missing transformers.onnx module vs LayoutLMv3 error messaging about boxes." + "reason": "TOKENIZER_MAPPING_NAMES returning None and a flash-attention crash on s_aux=None are unrelated code paths." }, { - "left": "issue:44117", - "right": "issue:45588", + "left": "issue:22355", + "right": "issue:29127", "accept": false, - "reason": "Tokenizer mapping fallback bug vs flash-attention crash on s_aux=None. Different components and failure modes." + "reason": "ONNX import failure and a LayoutLMv3 error-message issue are unrelated." }, { "left": "issue:43827", "right": "issue:45446", "accept": false, - "reason": "Docs mentioning pipeline removals vs a flex_attention import/version check bug. No shared code-path bug." + "reason": "Docs still referencing removed pipeline APIs is unrelated to the flex_attention import version check bug." }, { "left": "issue:43381", "right": "issue:44945", "accept": false, - "reason": "Gradient checkpointing in eval mode is unrelated to incorrect output under pipeline parallelism." + "reason": "Gradient-checkpointing in eval mode and incorrect output under pipeline parallelism are different problems." }, { "left": "issue:43937", "right": "issue:44977", "accept": false, - "reason": "Different generation failures for different models; one is invalid GenerationConfig, the other is flash-attention behavior." + "reason": "Both are Qwen3.5-related, but invalid GenerationConfig and flash-attention generation failure are distinct bugs." }, { "left": "issue:45405", "right": "issue:45446", "accept": false, - "reason": "Version-pin packaging issue vs an incorrect PyTorch version guard. Not the same bug." + "reason": "A PEFT version pin issue and a PyTorch version check bug are separate dependency problems." }, { "left": "issue:43450", "right": "issue:43986", "accept": false, - "reason": "Batched video shape bug vs missing torchvision crash when loading through AutoProcessor. Related area, but not the same defect." + "reason": "Batched video-shape handling and missing torchvision during AutoProcessor loading are different failures." }, { "left": "issue:39401", "right": "issue:43421", "accept": false, - "reason": "Both are tokenizer-related, but offset_mapping errors and runtime post-processor updates are different issues." + "reason": "Offset-mapping regression and runtime special-token post-processor updates are not the same issue." }, { "left": "issue:44117", "right": "issue:45081", "accept": false, - "reason": "Tokenizer mapping None-handling vs Mistral regex patch crash. Different affected paths." + "reason": "A tokenizer mapping None bug and a Mistral regex patch crash are unrelated internals." }, { "left": "issue:42673", "right": "issue:45072", "accept": false, - "reason": "VRAM leak in multi-threaded Qwen3ForCausalLM use vs dtype mismatches in bfloat16 inference. Unrelated." + "reason": "VRAM leak in Qwen3ForCausalLM and dtype mismatches in inference are different bugs." }, { "left": "issue:38617", "right": "issue:41628", "accept": false, - "reason": "Both are import errors, but they concern different symbols and modules. No evidence of one shared fix." + "reason": "Both are import errors, but they involve different missing symbols and different root causes." }, { "left": "issue:44829", "right": "issue:45216", "accept": false, - "reason": "Different training/save issues: flash_attention_3 degenerate training vs incorrect checkpoint saving for Qwen3.5." + "reason": "Flash-attention training degeneration and broken save_pretrained checkpoints are not the same concrete code path." }, { - "left": "issue:44291", - "right": "issue:45071", + "left": "issue:44410", + "right": "issue:45310", "accept": false, - "reason": "Both are version-regression breakages, but one is init_empty_weights argument handling and the other is PretrainedConfig type checking. Too distinct." + "reason": "Both concern Qwen3.5/MoE loading, but one is missing layer projections and the other is a from_pretrained failure." }, { - "left": "issue:44410", - "right": "issue:45310", + "left": "issue:44291", + "right": "issue:45071", "accept": false, - "reason": "Both touch Qwen3.5/Qwen3next loading, but one is missing projections and the other is a from_pretrained error. Not the same concrete bug." + "reason": "init_empty_weights TypeError and PretrainedConfig type-checking breakage are distinct compatibility issues." }, { "left": "issue:43408", "right": "issue:43742", "accept": false, - "reason": "Model-type warning for sam3 vs a KeyError loading MobileLLM. No shared failure mechanism." + "reason": "A model-type warning for sam3_* and a key error loading MobileLLM are unrelated model-loading issues." }, { "left": "issue:42222", "right": "issue:45542", "accept": false, - "reason": "VitPose model breakage vs TensorFlow backend detection issue. Different subsystems." + "reason": "Vitpose breakage and tensorboard-without-TensorFlow backend errors are separate subsystems." }, { "left": "issue:44484", "right": "issue:45310", "accept": false, - "reason": "save_pretrained shard-size question vs Qwen3.5 MoE from_pretrained failure. Unrelated." + "reason": "Chat/completions request rejection and Qwen3.5 MoE from_pretrained loading error are different failures." }, { "left": "issue:42175", "right": "issue:42222", "accept": false, - "reason": "TensorFlow install/backend issue vs VitPose model breakage. Different bugs." + "reason": "TensorFlow backend installation and vitpose model breakage do not share a concrete bug." }, { "left": "issue:43937", "right": "issue:44496", "accept": false, - "reason": "Invalid GenerationConfig vs unrecognized model/config.json. Different root causes." + "reason": "Invalid GenerationConfig for GLM-5 and unrecognized model/config loading for Olmo are different model-specific errors." }, { "left": "issue:44483", "right": "issue:45464", "accept": false, - "reason": "Both mention chat/completions, but the reported failures are different enough that they do not look like one concrete bug." + "reason": "Chat/completions API rejection and Qwen3.5 streaming inference failure are not the same underlying issue." }, { "left": "issue:44075", "right": "issue:45656", "accept": false, - "reason": "SGD args ignored vs optimizer.step called twice under deepspeed. Different optimizer problems." + "reason": "SGD args being ignored and deepspeed calling optimizer.step() twice are different optimizer bugs." }, { "left": "issue:44964", "right": "issue:45278", "accept": false, - "reason": "Model loading failure for Phi-4 multimodal vs broad import errors after upgrade. Not the same issue." + "reason": "A Phi-4 multimodal load failure and a broad upgrade-related import-error report are not the same concrete bug." }, { "left": "issue:43012", "right": "issue:43577", "accept": false, - "reason": "General bfloat16 warning on compile vs Blip2 dtype fields staying float32. Same theme, but distinct defects." + "reason": "A PyTorch warning during bf16 compile and Blip2 dtype staying float32 are related only loosely by dtype, not by bug." }, { "left": "issue:43066", "right": "issue:43122", "accept": false, - "reason": "Wrong decoder type in v5 vs changed tokenization output across versions. Similar area, but not the same bug." + "reason": "Both are tokenizer regressions, but one is decoder-type handling in v5 and the other is a version-to-version tokenization change." }, { "left": "issue:38617", "right": "issue:44220", "accept": false, - "reason": "ImportError for a missing symbol vs audio feature-extraction failure. No shared code-path." + "reason": "Missing configuration import and fbank feature extraction are unrelated failures." } ] }, @@ -12510,8 +12483,8 @@ "issue:43874|issue:44261", "issue:44484|issue:45478", "issue:44410|issue:45478", - "issue:44263|issue:44485", "issue:42222|issue:45459", + "issue:44263|issue:44485", "issue:43441|issue:44829", "issue:43012|issue:43408", "issue:41628|issue:45071", @@ -12522,8 +12495,8 @@ "issue:44117|issue:44987", "issue:43097|issue:43452", "issue:43986|issue:44484", - "issue:43931|issue:44038", "issue:43335|issue:45071", + "issue:43931|issue:44038", "issue:43976|issue:44945", "issue:43408|issue:44368", "issue:43874|issue:44263", @@ -12532,267 +12505,267 @@ "issue:43064|issue:44315", "issue:44464|issue:45071", "issue:43097|issue:45588", - "issue:44410|issue:45375", "issue:44062|issue:44117", + "issue:44410|issue:45375", "issue:38617|issue:44351", - "issue:44704|issue:45103", "issue:42673|issue:43577", + "issue:44704|issue:45103", "issue:44038|issue:44410", "issue:44261|issue:44485", "issue:44964|issue:45341", "issue:43986|issue:44117", "issue:43931|issue:44410" ], - "prepared_review_unit_hash": "3e5800e33797df7d931f426eb2b23d35e5c3582270a5a07d6066aedf70c96770", + "prepared_review_unit_hash": "0ee2530573861c0d660ef65316ef239599cfd141ff25d110e2a70e6419e8d31a", "component_budget": { "node_count": 387, "item_count": 387, "soft_pair_count": 1455, - "serialized_chars": 749552, - "estimated_input_tokens": 187388, - "estimated_eval_tokens": 375032 + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 37, - "serialized_chars": 38811, - "estimated_input_tokens": 9703, - "estimated_eval_tokens": 19662 + "serialized_chars": 38815, + "estimated_input_tokens": 9704, + "estimated_eval_tokens": 19664 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster looks heterogeneous: the issues span unrelated import errors, model-specific load failures, training/runtime regressions, and config/attention bugs. I do not see any soft pair that clearly describes the same underlying defect or change.", - "confidence": 0.96, - "canonical_issue_reason": "No true canonical duplicate stands out; if one issue must anchor the cluster, issue:43097 is the broadest cross-cutting regression and has the strongest evidence of wider impact/inbound references.", + "summary": "This cluster is highly heterogeneous: most pairs only share broad Transformers-v5/API similarity, not the same concrete bug or change. No soft edge looks safe to merge as a duplicate.", + "confidence": 0.93, + "canonical_issue_reason": "No single canonical issue cleanly subsumes this set; the items span unrelated import errors, model-specific loading bugs, config regressions, and training/runtime issues.", "canonical_pr_reason": null, - "best_issue_reason": "Issue:43097 is the best single representative only by breadth and reference count, not because it matches the others closely.", + "best_issue_reason": "issue:43097 is the best umbrella representative because it is a widely referenced version-wide breaking change with broader downstream impact than the others, though it still does not cover the whole cluster.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:42371", "right": "issue:44261", "accept": false, - "reason": "TF32 API settings warning vs MLA layernorm epsilon precision bug; different subsystems and failure modes." + "reason": "Different bugs: TF32 API guidance vs a FineGrainedFP8/precision regression." }, { "left": "issue:43097", "right": "issue:44493", "accept": false, - "reason": "Embedding tying deprecation/removal vs unexpected position-id key warnings; unrelated regressions." + "reason": "Different regressions: removed config flag vs unexpected position-id key warnings." }, { "left": "issue:43874", "right": "issue:44261", "accept": false, - "reason": "Missing image-patch method in GLM46V vs rms_norm_eps precision issue in MLA; not the same bug." + "reason": "Different code paths: missing image-patch helper vs RMS eps precision issue." }, { "left": "issue:44484", "right": "issue:45478", "accept": false, - "reason": "save_pretrained shard-size question vs Qwen3.5-Moe loading error; unrelated topics." + "reason": "Unrelated: shard-size default question vs Qwen3.5 MoE loading error." }, { "left": "issue:44410", "right": "issue:45478", "accept": false, - "reason": "Missing qkv/gate projections in qwen3next vs Qwen3.5-Moe from_pretrained failure; different model families." + "reason": "Same broad model family, but different defects and fixes." }, { - "left": "issue:44263", - "right": "issue:44485", + "left": "issue:42222", + "right": "issue:45459", "accept": false, - "reason": "torch.split return-value issue vs GLM-5 RoPE implementation; different code paths." + "reason": "VitPose breakage vs tokenizer protobuf error; no shared failure mode." }, { - "left": "issue:42222", - "right": "issue:45459", + "left": "issue:44263", + "right": "issue:44485", "accept": false, - "reason": "vitpose breakage vs tokenizer error masking when protobuf is absent; no shared defect." + "reason": "Different GLM issues: torch.split handling vs RoPE implementation." }, { "left": "issue:43441", "right": "issue:44829", "accept": false, - "reason": "Both mention flash attention, but one is Ministral-3 inference failure and the other is degenerate training with FA3; not mergeable as one bug." + "reason": "Both flash-attention related, but one is inference breakage and the other is training degeneration." }, { "left": "issue:43012", "right": "issue:43408", "accept": false, - "reason": "bfloat16 precision warning vs model-type mismatch warning for sam3_tracker; unrelated." + "reason": "Float16/bfloat16 warning vs SAM3 model-type warning; unrelated." }, { "left": "issue:41628", "right": "issue:45071", "accept": false, - "reason": "Missing AutoImageProcessor import vs PretrainedConfig type-checking regression; different APIs." + "reason": "Import error for AutoImageProcessor vs PretrainedConfig type-checking regression." }, { "left": "issue:45071", "right": "issue:45542", "accept": false, - "reason": "Type checking breakage in PretrainedConfig vs TensorFlow backend detection error; unrelated." + "reason": "Config type checking vs TensorFlow backend detection; different code paths." }, { "left": "issue:42673", "right": "issue:44977", "accept": false, - "reason": "VRAM leak in multithreaded dataloaders vs Qwen3.5 flash-attention generation issue; different symptoms and code paths." + "reason": "VRAM leak in dataloader threads vs flash-attention misgeneration." }, { "left": "issue:42371", "right": "issue:42831", "accept": false, - "reason": "TF32 behavior settings warning vs FineGrainedFP8 accuracy issue; only loosely related by precision." + "reason": "TF32 settings warning and FineGrainedFP8 accuracy are unrelated." }, { "left": "issue:42673", "right": "issue:43906", "accept": false, - "reason": "Qwen3ForCausalLM VRAM leak vs isolated reproduction of another issue; no evidence they share root cause." + "reason": "Different reports: a Qwen VRAM leak vs an isolated reproduction of another issue." }, { "left": "issue:44117", "right": "issue:44987", "accept": false, - "reason": "Tokenizer mapping None handling vs model loading failure for physical-intelligence/fast; different lookup layers." + "reason": "Tokenizer mapping None handling vs a specific model loading failure." }, { "left": "issue:43097", "right": "issue:43452", "accept": false, - "reason": "Embedding tying removal regression vs gguf_file loading breakage; unrelated model-loading bugs." + "reason": "Config deprecation vs gguf_file loading breakage; not the same bug." }, { "left": "issue:43986", "right": "issue:44484", "accept": false, - "reason": "AutoProcessor crash without torchvision vs max_shard_size question; no overlap in underlying defect." + "reason": "AutoProcessor/torchvision crash vs save_pretrained shard-size behavior." }, { - "left": "issue:43931", - "right": "issue:44038", + "left": "issue:43335", + "right": "issue:45071", "accept": false, - "reason": "Qwen3-VL-30B weight-shape mismatch vs Qwen3-VL-Moe regression; both are load errors but not the same concrete bug." + "reason": "SwitchTransformers sparse-layer creation bug vs config type-checking regression." }, { - "left": "issue:43335", - "right": "issue:45071", + "left": "issue:43931", + "right": "issue:44038", "accept": false, - "reason": "SwitchTransformers sparse-layer config bug vs PretrainedConfig type checking regression; unrelated." + "reason": "Different Qwen-VL failures: weight-shape mismatch vs Qwen3-VL-Moe incompatibility." }, { "left": "issue:43976", "right": "issue:44945", "accept": false, - "reason": "Python compatibility issue for Transformers 5.1.0 vs wrong output under pipeline parallelism; different problem classes." + "reason": "Python version compatibility report vs pipeline-parallelism output bug." }, { "left": "issue:43408", "right": "issue:44368", "accept": false, - "reason": "sam3_video/sam3_tracker model-type warning vs tie_word_embeddings warning in Qwen3.5; unrelated warnings." + "reason": "Model type mismatch warning vs tie_word_embeddings warning; separate issues." }, { "left": "issue:43874", "right": "issue:44263", "accept": false, - "reason": "Missing GLM46V image-patch helper vs torch.split return-values issue; no shared root cause." + "reason": "Different GLM failures: missing helper method vs torch.split semantics." }, { "left": "issue:44829", "right": "issue:45245", "accept": false, - "reason": "FlashAttention-3 training collapse vs category-count limit runtime error; unrelated." + "reason": "FlashAttention-3 training regression vs a category-count runtime limit." }, { "left": "issue:38617", "right": "issue:42757", "accept": false, - "reason": "Different import errors from different symbols/packages: layer_type_validation vs is_offline_mode." + "reason": "Both are import errors, but they are different missing symbols with different causes." }, { "left": "issue:43064", "right": "issue:44315", "accept": false, - "reason": "FSDP2/PEFT optimizer-state mismatch vs Liger Kernel not applied with model_init; different training bugs." + "reason": "Distributed optimizer-state bug vs Liger Kernel not applied; unrelated." }, { "left": "issue:44464", "right": "issue:45071", "accept": false, - "reason": "Chunked generation inconsistency under compiled forward vs PretrainedConfig type-checking regression; unrelated." + "reason": "Chunked generation inconsistency under compiled forward vs config type-checking regression." }, { "left": "issue:43097", "right": "issue:45588", "accept": false, - "reason": "Embedding tying removal regression vs flash_attention.py s_aux=None crash; different code paths." + "reason": "Config removal vs flash_attention s_aux crash; different code paths." }, { - "left": "issue:44410", - "right": "issue:45375", + "left": "issue:44062", + "right": "issue:44117", "accept": false, - "reason": "Missing qwen3next projections vs missing deepstack_visual_indexes field; different config/model issues." + "reason": "AddedToken construction bug vs tokenizer mapping None handling." }, { - "left": "issue:44062", - "right": "issue:44117", + "left": "issue:44410", + "right": "issue:45375", "accept": false, - "reason": "AddedToken special-arg duplication vs tokenizer mapping returning None; unrelated tokenizer bugs." + "reason": "Qwen3Next missing projections vs a vision-config field being dropped; similar family, different bug." }, { "left": "issue:38617", "right": "issue:44351", "accept": false, - "reason": "Different missing imports in different modules; no shared underlying defect." + "reason": "Separate import errors for different missing transformers symbols." }, { - "left": "issue:44704", - "right": "issue:45103", + "left": "issue:42673", + "right": "issue:43577", "accept": false, - "reason": "AutoProcessor kwargs propagation bug vs auto_docstring annotation crash; unrelated utility paths." + "reason": "VRAM leak vs BLIP2 dtype propagation bug." }, { - "left": "issue:42673", - "right": "issue:43577", + "left": "issue:44704", + "right": "issue:45103", "accept": false, - "reason": "VRAM leak in Qwen3ForCausalLM vs Blip2 dtype staying float32; different symptoms and fixes." + "reason": "Kwargs forwarding bug vs auto-docstring crash from future annotations." }, { "left": "issue:44038", "right": "issue:44410", "accept": false, - "reason": "Both involve Qwen3-family models, but one is a Qwen3-VL-Moe load bug and the other is qwen3next layer construction; not the same bug." + "reason": "Different Qwen failures: MoE loading issue vs missing attention projections." }, { "left": "issue:44261", "right": "issue:44485", "accept": false, - "reason": "MLA q_a_layernorm epsilon precision issue vs GLM-5 RoPE implementation; different components." + "reason": "MLA q_a_layernorm precision issue vs GLM-5 RoPE implementation; unrelated." }, { "left": "issue:44964", "right": "issue:45341", "accept": false, - "reason": "Phi-4 multimodal loading regression vs a testing_utils bug; unrelated." + "reason": "Model loading failure vs a testing_utils bug." }, { "left": "issue:43986", "right": "issue:44117", "accept": false, - "reason": "Video-model loading without torchvision vs tokenizer mapping None; different layers of the stack." + "reason": "AutoProcessor crash without torchvision vs tokenizer mapping None handling." }, { "left": "issue:43931", "right": "issue:44410", "accept": false, - "reason": "Weight-shape mismatch for Qwen3-VL-30B vs missing projections in qwen3next; unrelated model defects." + "reason": "Qwen3-VL weight mismatch vs Qwen3Next missing projections; distinct bugs." } ] }, @@ -12869,8 +12842,8 @@ "issue:43502|issue:43992", "issue:43097|issue:43873", "issue:43066|issue:43906", - "issue:43986|issue:45372", "issue:43986|issue:45290", + "issue:43986|issue:45372", "issue:45341|issue:45561", "issue:43986|issue:45341", "issue:43097|issue:43901", @@ -12879,8 +12852,8 @@ "issue:43992|issue:44484", "issue:45341|issue:45588", "issue:42886|issue:43381", - "issue:43012|issue:43550", "issue:43208|issue:43550", + "issue:43012|issue:43550", "issue:43012|issue:44368", "issue:42907|issue:45753", "issue:42907|issue:44655", @@ -12890,8 +12863,8 @@ "issue:42757|issue:43824", "issue:43012|issue:44683", "issue:44483|issue:45290", - "issue:43618|issue:43761", "issue:44038|issue:44230", + "issue:43618|issue:43761", "issue:43064|issue:44380", "issue:38617|issue:43502", "issue:44483|issue:45381", @@ -12903,20 +12876,20 @@ "issue:43352|issue:43937", "issue:43066|issue:45381" ], - "prepared_review_unit_hash": "c644993518360014f6d9732694ef6e791be4af95261c3dcc1bd5049522a54d30", + "prepared_review_unit_hash": "4ee419f0bc2a8a144b8d2c841f72f633908cb67c63025d7ef42f3a97a03b2f5c", "component_budget": { "node_count": 387, "item_count": 387, "soft_pair_count": 1455, - "serialized_chars": 749552, - "estimated_input_tokens": 187388, - "estimated_eval_tokens": 375032 + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 36, - "serialized_chars": 38200, + "serialized_chars": 38197, "estimated_input_tokens": 9550, "estimated_eval_tokens": 19356 }, @@ -12925,244 +12898,235 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This set is mostly a loose similarity cluster of Transformers v5 regressions and model-specific bugs, not a single duplicate group. I found one plausible duplicate pair in CLIP output handling; the rest look like distinct issues despite overlapping keywords (loading, saving, v5, testing utils, etc.).", - "confidence": 0.77, - "canonical_issue_reason": "issue:43761 is the clearest, most specific anchor for the CLIP v5 regression: it names the exact forward behavior (`hidden_states=None` despite `output_hidden_states=True`) and is more actionable than the sibling symptom report.", + "summary": "No soft pair looks like the same underlying bug/change. The items mostly split into unrelated v5 regressions, model-specific crashes, saving/loading bugs, and test-utils issues. The most representative issue for the cluster is the broad v5 API regression around removed behavior.", + "confidence": 0.55, + "canonical_issue_reason": "issue:43097 is the broadest and most representative regression here: it\u2019s explicitly tied to v5.0.0, describes a removed public API without guidance, and has inbound references, making it a better cluster anchor than the model-specific crashes.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43761 is the strongest issue to keep as the representative bug report because it is concrete, reproducible, and appears to capture the shared CLIP output-population regression most directly.", + "best_issue_reason": "issue:43097 is the best single issue to surface because it captures a user-facing v5 breaking change with broader impact than the narrower model or pipeline-specific reports.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43502", "right": "issue:43992", "accept": false, - "reason": "Both are load/cache-related, but one is an offline/local_files_only networking bug and the other is a missing weight during UMT5 loading; different code paths." + "reason": "Local-files-only network leakage and missing embed_tokens.weight are different loading bugs." }, { "left": "issue:43097", "right": "issue:43873", "accept": false, - "reason": "One is an API removal regression, the other is a quantization/offloading problem; same release era, not the same bug." + "reason": "Removed v5 API behavior vs quantized offloading failure; different code paths and symptoms." }, { "left": "issue:43066", "right": "issue:43906", "accept": false, - "reason": "The titles point to different tokenizer/import problems; there is no evidence they share the same concrete failure." + "reason": "Wrong tokenizer decoder type is unrelated to the unspecified isolated reproduction issue." }, { "left": "issue:43986", - "right": "issue:45372", + "right": "issue:45290", "accept": false, - "reason": "Both involve processor/model loading, but they fail for different missing dependencies and different models; not mergeable as one fix." + "reason": "AutoProcessor dependency crash vs chat-template tool-call crash are distinct failures." }, { "left": "issue:43986", - "right": "issue:45290", + "right": "issue:45372", "accept": false, - "reason": "Video-model loading without torchvision and chat-template crashes with tool calls are unrelated bugs." + "reason": "Missing torchvision crash is unrelated to mistral_common import failure in Gemma 4 loading." }, { "left": "issue:45341", "right": "issue:45561", "accept": false, - "reason": "Both mention testing_utils, but one is a vague small bug and the other is a specific xdist race on captured_info.txt; not clearly the same underlying issue." + "reason": "Generic testing_utils bug vs xdist race on captured_info.txt are not the same defect." }, { "left": "issue:43986", "right": "issue:45341", "accept": false, - "reason": "Different subsystems entirely: video processor loading vs testing utils." + "reason": "Processor loading crash and a testing_utils bug do not share the same underlying issue." }, { "left": "issue:43097", "right": "issue:43901", "accept": false, - "reason": "One concerns removed model config API, the other is stale pipeline docs; overlap is only around v5 surface area." + "reason": "Breaking API removal vs stale docs for return_all_scores are different problems." }, { "left": "issue:43825", "right": "issue:44655", "accept": false, - "reason": "Both are pipeline-related, but one is an error-message regression and the other is save_pretrained support for Pipeline objects; distinct fixes." + "reason": "Pipeline task messaging issue and save_pretrained serialization failure are unrelated." }, { "left": "issue:42617", "right": "issue:45341", "accept": false, - "reason": "3d_parallel runtime failure and testing_utils bug are unrelated." + "reason": "3d_parallel.py execution failure is unrelated to a testing-utils report." }, { "left": "issue:43992", "right": "issue:44484", "accept": false, - "reason": "Model weight-loading and chat-completions API behavior are different layers and different bugs." + "reason": "Missing checkpoint key vs max_shard_size default question are different save/load concerns." }, { "left": "issue:45341", "right": "issue:45588", "accept": false, - "reason": "Testing utils race and flash-attention sinkless-model attribute error are unrelated." + "reason": "Testing utils bug vs flash_attention AttributeError on sink-less models are unrelated." }, { "left": "issue:42886", "right": "issue:43381", "accept": false, - "reason": "Offline tokenizer cache loading and gradient checkpointing in eval mode are separate behaviors." + "reason": "Offline tokenizer cache loading and gradient-checkpointing eval-mode behavior are different code paths." }, { - "left": "issue:43012", + "left": "issue:43208", "right": "issue:43550", "accept": false, - "reason": "One is a PyTorch warning about bfloat16 compilation, the other is a compile failure for a specific model/backend combination; not the same bug." + "reason": "xLSTM training bugs are unrelated to Bamba torch.compile/SDPA failure." }, { - "left": "issue:43208", + "left": "issue:43012", "right": "issue:43550", "accept": false, - "reason": "xLSTM training blockers and Bamba SDPA compile failure are unrelated model-specific problems." + "reason": "bf16 warning during compile is not the same as a compile-time SDPA crash." }, { "left": "issue:43012", "right": "issue:44368", "accept": false, - "reason": "Both are warnings, but about different settings and different code paths." + "reason": "PyTorch precision warning and tie_word_embeddings warning are different warnings with different causes." }, { "left": "issue:42907", "right": "issue:45753", "accept": false, - "reason": "Saving dequantized Ministrals/Devstrals and Qwen3_5 infinite-loop on a specific image are unrelated." + "reason": "Quantized model save failure is unrelated to a Qwen3_5 image-generation infinite loop." }, { "left": "issue:42907", "right": "issue:44655", "accept": false, - "reason": "Both mention saving, but one is dequantized model saving and the other is Pipeline.save_pretrained; different objects and failures." + "reason": "Saving dequantized models and saving Pipeline objects are different serialization problems." }, { "left": "issue:43531", "right": "issue:45341", "accept": false, - "reason": "Qwen3-MoE sliding_window bug and testing_utils bug are unrelated." + "reason": "Qwen3-MoE sliding_window issue is unrelated to testing_utils." }, { "left": "issue:43208", "right": "issue:43526", "accept": false, - "reason": "xLSTM training bugs and BeitImageProcessorFast reduce_labels behavior are different components and symptoms." + "reason": "xLSTM training bugs do not match BeitImageProcessorFast reduce_labels behavior." }, { "left": "issue:44483", "right": "issue:44514", "accept": false, - "reason": "Chat-completions API rejection and Qwen2.5-VL batched chat-template crash are different request paths." + "reason": "v1 chat/completions rejection and batched Qwen2.5-VL chat_template crash are different failures." }, { "left": "issue:42757", "right": "issue:43824", "accept": false, - "reason": "Both are import errors, but for different missing symbols from different packages; no shared fix is evident." + "reason": "Import errors for different symbols/modules are not duplicates." }, { "left": "issue:43012", "right": "issue:44683", "accept": false, - "reason": "A warning about compiling bfloat16 is not the same as flex_attention outright failing on newer torch versions." + "reason": "bfloat16 warning and compiled flex_attention failure are distinct issues." }, { "left": "issue:44483", "right": "issue:45290", "accept": false, - "reason": "Chat completions endpoint behavior and assistant-message tool-call template handling are unrelated." - }, - { - "left": "issue:43618", - "right": "issue:43761", - "accept": true, - "reason": "Both are CLIP v5 regressions in output handling/forward results: one says attentions are no longer assigned, the other says hidden_states are missing despite being requested. This looks like one shared CLIP output-population bug." + "reason": "Chat/completions endpoint rejection is unrelated to assistant tool-call template crashes." }, { "left": "issue:44038", "right": "issue:44230", "accept": false, - "reason": "Same model family name, but one is a load regression in v5 and the other is an fp8 support request; not the same concrete bug." + "reason": "Qwen3-VL-Moe regression and Qwen3.5 fp8 inference support are different requests." + }, + { + "left": "issue:43618", + "right": "issue:43761", + "accept": false, + "reason": "Missing attentions assignment and missing hidden_states are separate CLIP regressions." }, { "left": "issue:43064", "right": "issue:44380", "accept": false, - "reason": "FSDP2/PEFT optimizer-state corruption and GPT2 attention scaling with SDPA/FlashAttention are unrelated." + "reason": "FSDP2 optimizer-state corruption and GPT2 attention scaling under SDPA are unrelated." }, { "left": "issue:38617", "right": "issue:43502", "accept": false, - "reason": "Importing a missing configuration_utils symbol and local_files_only making API requests are different issues." + "reason": "ImportError in configuration_utils and network requests despite local_files_only are different bugs." }, { "left": "issue:44483", "right": "issue:45381", "accept": false, - "reason": "OpenAI-style chat endpoint rejection and Qwen2.5-VL video position-id bug are unrelated." + "reason": "Chat/completions API rejection and Qwen2.5-VL video position_ids are unrelated." }, { "left": "issue:43299", "right": "issue:43867", "accept": false, - "reason": "Both are load/save regressions, but one is a Qwen3VL-MoE load breakage and the other is a state_dict ordering issue; not enough to call them duplicates." + "reason": "Qwen3VL loading breakage and sorted-state_dict load failure are not the same defect." }, { "left": "issue:43901", "right": "issue:44655", "accept": false, - "reason": "Docs drift for return_all_scores and Pipeline.save_pretrained failures are different categories of bugs." + "reason": "Documentation mismatch and Pipeline save_pretrained failure are unrelated." }, { "left": "issue:36683", "right": "issue:43475", "accept": false, - "reason": "Both are attribute errors on model outputs/configs, but they affect different models and missing fields; no shared code path is evident." + "reason": "Missing vocab_size on Gemma3Config and missing fpn_position_embeddings on SAM3 output are different attribute errors." }, { "left": "issue:43867", "right": "issue:45216", "accept": false, - "reason": "Both involve checkpoint/loading semantics, but one is about sorted state_dict loading and the other about incorrect saved checkpoints; too different to merge." + "reason": "State_dict sorting load error and incorrect Qwen3.5 checkpoint saving are distinct serialization issues." }, { "left": "issue:42994", "right": "issue:44655", "accept": false, - "reason": "Quantized model saving and Pipeline.save_pretrained are different save targets and likely different fixes." + "reason": "Quantized model saving failure is not the same as saving Pipeline objects." }, { "left": "issue:43352", "right": "issue:43937", "accept": false, - "reason": "FlashAttention support for Nemotron and invalid GenerationConfig for GLM-5 are unrelated." - }, - { - "left": "issue:44038", - "right": "issue:44230", - "accept": false, - "reason": "Reiterated: one is a load bug, the other is an fp8 support request; not the same underlying change." + "reason": "FlashAttention support error and invalid GenerationConfig are unrelated." }, { "left": "issue:43066", "right": "issue:45381", "accept": false, - "reason": "Tokenizer decoder-type regression and Qwen2.5-VL video position-id bug are unrelated." + "reason": "Tokenizer decoder type regression and video position_ids bug are different problems." } ] }, - "evaluator_result": { - "accept": true, - "feedback": "The summary is grounded in the packet and stays conservative. The accepted CLIP pair is plausibly the same output-population regression, while the rejected pairs are mostly supported as distinct by their titles and described failure modes. No clear overstatement beyond the evidence provided." - }, - "evaluator_used": true, + "evaluator_result": null, + "evaluator_used": false, "retried": false, - "accepted_nontrivial_soft_edge": true, + "accepted_nontrivial_soft_edge": false, "error_kind": null, "error_message": null }, @@ -13230,8 +13194,8 @@ ], "soft_pairs": [ "issue:36010|issue:42757", - "issue:44263|issue:45440", "issue:44220|issue:44655", + "issue:44263|issue:45440", "issue:43756|issue:44485", "issue:36683|issue:43525", "issue:43618|issue:45216", @@ -13265,251 +13229,254 @@ "issue:43867|issue:44484", "issue:43618|issue:44857" ], - "prepared_review_unit_hash": "6f6dcb41d6d1753e8221e43a0c7550747a846e5b005c72972430697864b1386f", + "prepared_review_unit_hash": "2b5fac5231b0b5bb7f9306640c1ab5f5f79e964e0d3e40aa312eb234c17e15f4", "component_budget": { "node_count": 387, "item_count": 387, "soft_pair_count": 1455, - "serialized_chars": 749552, - "estimated_input_tokens": 187388, - "estimated_eval_tokens": 375032 + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 35, - "serialized_chars": 37786, - "estimated_input_tokens": 9447, - "estimated_eval_tokens": 19150 + "serialized_chars": 37791, + "estimated_input_tokens": 9448, + "estimated_eval_tokens": 19152 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is very heterogeneous: most items are unrelated bug reports across loading, config, attention, TP, and test infrastructure. I do not see any soft pair that is clearly the same underlying bug/change, so I would not merge any of them as duplicates.", - "confidence": 0.86, - "canonical_issue_reason": "If a single anchor is required, issue #36683 is the best representative because it is the most active/open issue and matches the nearby family of model-config attribute/loading regressions; however, the overall set is not a true duplicate cluster.", + "summary": "This cluster is mostly heterogeneous model-specific bug reports and regressions. The only strong duplicate/soft-match is the import-time slowdown pair (`import transformers takes long sometimes` vs. the `import *` performance regression/fix); the rest are different code paths, models, or symptoms and should stay separate.", + "confidence": 0.89, + "canonical_issue_reason": "`issue:44246` is the broadest, most general issue in the set and best represents the import-performance theme; `issue:41669` reads as a specific cause/fix for that same slowdown.", "canonical_pr_reason": null, - "best_issue_reason": "Issue #36683 is the strongest global issue candidate by discussion volume and because several nearby items are also model-loading/config-attribute problems, making it the closest reusable triage anchor.", + "best_issue_reason": "`issue:44246` is the best representative issue because it describes the user-facing symptom without narrowing to one implementation detail or model.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:36010", "right": "issue:42757", "accept": false, - "reason": "Different import failures in different modules/dependencies; same symptom, not the same bug." + "reason": "Both are import errors, but they involve different missing symbols in different packages and different fixes." }, { - "left": "issue:44263", - "right": "issue:45440", + "left": "issue:44220", + "right": "issue:44655", "accept": false, - "reason": "Unrelated code paths: GlmMoeDsaIndexer torch.split behavior vs DeepSeekV3 implementation divergence." + "reason": "Audio feature extraction bug vs. pipeline save_pretrained problem; unrelated code paths." }, { - "left": "issue:44220", - "right": "issue:44655", + "left": "issue:44263", + "right": "issue:45440", "accept": false, - "reason": "Different subsystems: fbank feature extraction vs pipeline save_pretrained." + "reason": "Different model internals and failures; no shared underlying bug." }, { "left": "issue:43756", "right": "issue:44485", "accept": false, - "reason": "Both mention RoPE, but they are different model-specific implementation issues." + "reason": "Both mention RoPE, but they target different models and different implementation issues." }, { "left": "issue:36683", "right": "issue:43525", "accept": false, - "reason": "Both are missing config attributes, but for different models and different fields; too broad to treat as one bug." + "reason": "Both are missing config attributes, but for different models and different fields." }, { "left": "issue:43618", "right": "issue:45216", "accept": false, - "reason": "Different regressions: CLIPOutput attentions assignment vs Qwen3.5 save_pretrained checkpoint keys." + "reason": "CLIP attentions assignment bug vs. Qwen save_pretrained regression; not the same issue." }, { "left": "issue:43906", "right": "issue:45753", "accept": false, - "reason": "No clear shared code path; one is a reproduction of an older issue, the other is an infinite-loop report on a specific image." + "reason": "One is a reproduction of another issue; the other is an infinite-loop report on a specific image." }, { "left": "issue:36683", "right": "issue:44242", "accept": false, - "reason": "Different model families and failure modes; config attribute error vs load-balancing loss logic." + "reason": "Config attribute bug vs. MoE load-balancing-loss logic bug; different failures." }, { "left": "issue:36683", "right": "issue:43881", "accept": false, - "reason": "Both are loading failures, but the specific missing attribute/model context differs." + "reason": "Different model-loading failure and different missing field; not a duplicate." }, { "left": "issue:43299", "right": "issue:45753", "accept": false, - "reason": "Different Qwen model variants and different failure modes: model loading vs generation loop." + "reason": "Both are Qwen3.5-related, but one is MoE loading and the other is an image-specific generation loop." }, { "left": "issue:43352", "right": "issue:43976", "accept": false, - "reason": "Unrelated: FlashAttention support for a model vs Python-version compatibility." + "reason": "FlashAttention model-support issue vs. Python-version compatibility issue." }, { "left": "issue:43441", "right": "issue:44380", "accept": false, - "reason": "Different attention bugs in different models; not the same concrete fix." + "reason": "Both touch attention backends, but they concern different models and different bugs." }, { "left": "issue:43352", "right": "issue:44380", "accept": false, - "reason": "Same broad attention area, but different model/backend behaviors and no shared code-path evidence." + "reason": "Generic FlashAttention backend behavior vs. model-specific unsupported-feature error." }, { "left": "issue:43992", "right": "issue:45310", "accept": false, - "reason": "Different models and different missing-key/loading failures." + "reason": "Different Qwen/UMT5 loading problems with no clear shared code-path." }, { "left": "issue:44380", "right": "issue:45479", "accept": false, - "reason": "Attention backend scaling bug vs classification-loss degeneration; unrelated." + "reason": "Attention scaling bug vs. classification loss degeneration; unrelated." }, { "left": "issue:41669", "right": "issue:44246", - "accept": false, - "reason": "Possible performance overlap, but one is a concrete import-star cleanup and the other is a vague slow-import report; not enough to merge." + "accept": true, + "reason": "These describe the same import-time slowdown: one identifies the `import *` cause, the other reports the user-visible slow import symptom." }, { "left": "issue:43867", "right": "issue:45375", "accept": false, - "reason": "Different loading/config issues; state_dict ordering vs a missing vision config field." + "reason": "Both are load/config issues, but one is about state_dict ordering and the other is a missing vision config field." }, { "left": "issue:43064", "right": "issue:44492", "accept": false, - "reason": "Completely unrelated: distributed optimizer-state bug vs typo in cache-strategy docs/code." + "reason": "Distributed training optimizer-state bug vs. a cache-strategy typo; unrelated." }, { "left": "issue:43618", "right": "issue:44380", "accept": false, - "reason": "Different bugs in different paths; CLIP output field assignment vs GPT2 backend scaling." + "reason": "Different attention/output bugs in different model families." }, { "left": "issue:43352", "right": "issue:45478", "accept": false, - "reason": "Same general FlashAttention theme, but different models and reports." + "reason": "Same general FlashAttention theme, but different models and distinct breakage." }, { "left": "issue:44355", "right": "issue:44855", "accept": false, - "reason": "Both involve import/runtime errors, but one is about compiled Python files and the other is a Python 3.13 JIT parsing issue." + "reason": "Compiled-Python import error vs. TorchScript parsing error; different root causes." }, { "left": "issue:36683", "right": "issue:44117", "accept": false, - "reason": "Both are loading-related, but one is a missing config attribute and the other is tokenizer mapping returning None." + "reason": "Missing config attribute vs. tokenizer mapping returning None; separate failures." }, { "left": "issue:42890", "right": "issue:45561", "accept": false, - "reason": "Test flakiness in different places: missing set_seed vs xdist file-write race." + "reason": "Both are test-related, but one is missing set_seed and the other is an xdist file-write race." }, { "left": "issue:43097", "right": "issue:43352", "accept": false, - "reason": "Unrelated: tie_embeddings API removal vs FlashAttention support." + "reason": "Tie-embeddings API removal vs. FlashAttention support error; not the same bug." }, { "left": "issue:43867", "right": "issue:45357", "accept": false, - "reason": "Different Qwen3.5 regressions: generic load failure vs incorrect visual-encoder keys on save." + "reason": "Different Qwen3.5 save/load regressions with no shared underlying code-path." }, { "left": "issue:43976", "right": "issue:44938", "accept": false, - "reason": "These are both compatibility complaints, but one is Python 3.9/3.10 support and the other is Python 3.14 load failure." + "reason": "Python version compatibility issue vs. Python 3.14 import failure; related only loosely." }, { "left": "issue:43992", "right": "issue:45478", "accept": false, - "reason": "Different models and failure modes; not the same from_pretrained bug." + "reason": "Different Qwen3.5 loading failures; one is UMT5 encoder weights, the other is MoE from_pretrained." }, { "left": "issue:44360", "right": "issue:45440", "accept": false, - "reason": "Different DSA/indexing logic vs DeepSeek model divergence." + "reason": "DSA indexer math bug vs. DeepseekV3MoE implementation divergence; not the same issue." }, { "left": "issue:43064", "right": "issue:44655", "accept": false, - "reason": "Distributed training optimizer-state bug vs pipeline save_pretrained; unrelated." + "reason": "Optimizer-state corruption in training vs. pipeline save_pretrained failure; unrelated." }, { "left": "issue:36296", "right": "issue:43575", "accept": false, - "reason": "Both mention tensor parallelism, but one is a training bug and the other is a load-time OOM; not the same concrete issue." + "reason": "Tensor-parallel training bug vs. load-time OOM for a specific model." }, { "left": "issue:44077", "right": "issue:44380", "accept": false, - "reason": "Different subsystems and fixes; post_init validation vs attention scaling." + "reason": "Optional post_init restriction vs. attention scaling ignored; different functionality." }, { "left": "issue:43618", "right": "issue:45588", "accept": false, - "reason": "Different bugs in unrelated code paths; output field assignment vs flash_attention s_aux handling." + "reason": "CLIP attentions assignment bug vs. flash_attention crash on sink-less models; different models and code paths." }, { "left": "issue:28282", "right": "issue:33453", "accept": false, - "reason": "Different regressions: missing PyTorch import vs tokenizer loading behavior." + "reason": "PyTorch-missing ImportError vs. tokenizer loading regression; unrelated root causes." }, { "left": "issue:43867", "right": "issue:44484", "accept": false, - "reason": "Different topics: load failure with sorted state_dict vs save_pretrained shard-size default." + "reason": "State-dict loading issue vs. max_shard_size default question; not duplicates." }, { "left": "issue:43618", "right": "issue:44857", "accept": false, - "reason": "Unrelated issues: CLIP output metadata vs AMP/CUDA crash in LwDetrImageLoss." + "reason": "Different model-specific runtime errors with no shared underlying bug." } ] }, - "evaluator_result": null, - "evaluator_used": false, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet, and the only accepted soft edge (issue:41669 \u2194 issue:44246) is a conservative duplicate match based on the shared import-time slowdown theme. The remaining pairwise rejections are appropriately cautious and do not overclaim duplicates or shared root causes." + }, + "evaluator_used": true, "retried": false, - "accepted_nontrivial_soft_edge": false, + "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null }, @@ -13578,8 +13545,8 @@ "soft_pairs": [ "issue:43618|issue:44977", "issue:43901|issue:44938", - "issue:42890|issue:44964", "issue:44655|issue:44938", + "issue:42890|issue:44964", "issue:42890|issue:44112", "issue:42890|issue:45600", "issue:43502|issue:44062", @@ -13591,8 +13558,8 @@ "issue:44355|issue:44749", "issue:30990|issue:31795", "issue:45125|issue:45478", - "issue:44387|issue:44556", "issue:45125|issue:45375", + "issue:44387|issue:44556", "issue:43010|issue:45278", "issue:30333|issue:33453", "issue:43723|issue:44556", @@ -13606,8 +13573,8 @@ "issue:44355|issue:45071", "issue:44222|issue:44230", "issue:43827|issue:44961", - "issue:33453|issue:34689", "issue:43010|issue:45656", + "issue:33453|issue:34689", "issue:43299|issue:44556", "issue:42757|issue:43576", "issue:33453|issue:36296", @@ -13615,262 +13582,220 @@ "issue:43576|issue:43824", "issue:44961|issue:45468" ], - "prepared_review_unit_hash": "300f2f3a9d637e22b527b85b843073a146a8af89e03d840c7c32e8617d7fd8de", + "prepared_review_unit_hash": "c347ea7360e09f82d5c604cc106352e32a55a887a9531993145c9d6527cc7544", "component_budget": { "node_count": 387, "item_count": 387, "soft_pair_count": 1455, - "serialized_chars": 749552, - "estimated_input_tokens": 187388, - "estimated_eval_tokens": 375032 + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 38, - "serialized_chars": 38498, - "estimated_input_tokens": 9625, - "estimated_eval_tokens": 19506 + "serialized_chars": 38503, + "estimated_input_tokens": 9626, + "estimated_eval_tokens": 19508 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a broad set of unrelated Transformers v5 regressions and compatibility reports; the soft-similarity pairs are mostly false positives. I would treat issue:45278 as the umbrella canonical issue because it best captures the broad \u201cmany import errors after upgrading to v5.5.0\u201d theme.", - "confidence": 0.86, - "canonical_issue_reason": "issue:45278 is the broadest and most representative report in the cluster, summarizing the dominant theme of widespread import/load breakage after upgrading to Transformers 5.5.0.", + "summary": "This cluster is mostly a grab bag of unrelated Transformers v5 regressions, import/load failures, and a few docs/tests issues. The broadest representative is the open umbrella report about many import errors after upgrading to 5.5.0; the rest are not close enough to treat as duplicates.", + "confidence": 0.78, + "canonical_issue_reason": "Issue 45278 is the best umbrella for this cluster because it broadly captures the recurring theme of import/load regressions after upgrading to v5, while the other issues are mostly narrower variants or unrelated docs/test failures.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45278 best serves as the cluster anchor because it describes the general regression class that many of the other import/load complaints resemble.", + "best_issue_reason": "Issue 45278 is the most useful global representative: it is broad, current, and aligns with the dominant pattern in the cluster (v5 upgrade regressions causing import/load breakage).", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43618", "right": "issue:44977", "accept": false, - "reason": "Different failures: missing CLIPOutput attentions vs Qwen3.5 flash-attention generation behavior." + "reason": "Different model families and symptoms: CLIP attentions assignment vs Qwen3.5 flash-attention generation." }, { "left": "issue:43901", "right": "issue:44938", "accept": false, - "reason": "Documentation mismatch for text-classification output vs Python 3.14 import/load failure; unrelated symptoms." + "reason": "Docs behavior mismatch vs Python 3.14 load failure; unrelated problems." }, { - "left": "issue:42890", - "right": "issue:44964", + "left": "issue:44655", + "right": "issue:44938", "accept": false, - "reason": "Test seed flakiness in sam_hq vs model loading failure for Phi-4 multimodal; not the same bug." + "reason": "Pipeline save_pretrained issue is not the same as interpreter compatibility/loading failure." }, { - "left": "issue:44655", - "right": "issue:44938", + "left": "issue:42890", + "right": "issue:44964", "accept": false, - "reason": "Pipeline save_pretrained issue vs Python 3.14 load failure; distinct code paths." + "reason": "Different model/test failure and a separate latest-transformers loading regression." }, { "left": "issue:42890", "right": "issue:44112", "accept": false, - "reason": "Both are test-related, but they concern different tests and failure causes." + "reason": "Both are CI/test failures, but on different models with different causes." }, { "left": "issue:42890", "right": "issue:45600", "accept": false, - "reason": "Sam HQ test seeding vs auto_mappings stale config references; unrelated." + "reason": "Different CI regressions: missing seeds in tests vs removed config references in auto_mappings." }, { "left": "issue:43502", "right": "issue:44062", "accept": false, - "reason": "Local-files-only network leak vs AddedToken constructor error; different problems." + "reason": "local_files_only network leakage and AddedToken kwarg duplication are different code paths." }, { "left": "issue:43010", "right": "issue:44075", "accept": false, - "reason": "Cache update no_grad discussion vs SGD optimizer args ignored; unrelated." + "reason": "Cache update no_grad behavior is unrelated to SGD optimizer argument handling." }, { "left": "issue:42890", "right": "issue:45341", "accept": false, - "reason": "Different test stability issues in different files/components." + "reason": "Test seeding flake vs a separate testing_utils bug; not the same issue." }, { "left": "issue:42548", "right": "issue:44351", "accept": false, - "reason": "Both are import errors, but for different symbols and likely different export regressions." + "reason": "Both are import errors, but for different symbols and likely different causes." }, { "left": "issue:44222", "right": "issue:45125", "accept": false, - "reason": "FP8 save_pretrained moE bug vs missing tensor-parallel plan; different concrete code paths." + "reason": "FP8 save_pretrained bug and missing tensor-parallel plan are different failures." }, { "left": "issue:43502", "right": "issue:44485", "accept": false, - "reason": "Local_files_only network behavior vs GLM-5 RoPE implementation; unrelated." + "reason": "Tokenizer/network isolation bug vs GLM RoPE implementation discussion." }, { "left": "issue:44355", "right": "issue:44749", "accept": false, - "reason": "Compiled Python file loading errors vs performance regression after upgrade; not the same bug." + "reason": "Compiled Python file errors are unrelated to the reported slowdown when filtering data." }, { "left": "issue:30990", "right": "issue:31795", "accept": false, - "reason": "Loading hang for Sentence Transformers vs documentation confusion in model.forward; unrelated." + "reason": "Loading hang and forward-argument documentation confusion are not duplicates." }, { "left": "issue:45125", "right": "issue:45478", "accept": false, - "reason": "Same model family, but one is missing _tp_plan and the other is a generic from_pretrained failure; too little evidence of one concrete duplicate bug." + "reason": "Same model family, but one is tensor-parallel metadata and the other is a from_pretrained regression; not one concrete bug." }, { - "left": "issue:44387", - "right": "issue:44556", + "left": "issue:45125", + "right": "issue:45375", "accept": false, - "reason": "Int4 quantization OOM vs checkpoint reload incompatibility after upgrade; different failure modes." + "reason": "Different missing config fields and different failure modes." }, { - "left": "issue:45125", - "right": "issue:45375", + "left": "issue:44387", + "right": "issue:44556", "accept": false, - "reason": "Tensor-parallel metadata issue vs missing deepstack_visual_indexes field; unrelated model config bugs." + "reason": "Memory/OOM regression under int4 quantization is distinct from checkpoint reload compatibility after upgrading." }, { "left": "issue:43010", "right": "issue:45278", "accept": false, - "reason": "No-grad cache update discussion vs broad import errors after upgrade; unrelated." + "reason": "Cache update semantics do not match the broad import-error regression report." }, { "left": "issue:30333", "right": "issue:33453", "accept": false, - "reason": "MLFlow reporting on failed training jobs vs tokenizer loading regression; unrelated." + "reason": "MLflow job status reporting and tokenizer loading regression are unrelated." }, { "left": "issue:43723", "right": "issue:44556", "accept": false, - "reason": "Tokenizer loading issue vs reloading old checkpoints after upgrade; related in theme but not the same concrete bug." - }, - { - "left": "issue:43010", - "right": "issue:44368", - "accept": false, - "reason": "Cache/layer update decoration vs tie_word_embeddings warning during fine-tuning; unrelated." - }, - { - "left": "issue:43502", - "right": "issue:44222", - "accept": false, - "reason": "Network access with local_files_only vs FP8 save_pretrained moe; different code paths." - }, - { - "left": "issue:43502", - "right": "issue:44961", - "accept": false, - "reason": "Local file download behavior vs a standalone issue titled 'racoon'; no meaningful relation." - }, - { - "left": "issue:44315", - "right": "issue:44360", - "accept": false, - "reason": "Liger kernel application bug vs DSA indexer ReLU discussion; unrelated." - }, - { - "left": "issue:42831", - "right": "issue:43874", - "accept": false, - "reason": "FP8 accuracy issue vs missing image patch-count API in GLM46V; distinct bugs." - }, - { - "left": "issue:30333", - "right": "issue:36296", - "accept": false, - "reason": "MLFlow reporting on failed jobs vs tensor parallel training bug; different subsystems." + "reason": "Tokenizer AutoTokenizer loading issue vs generic checkpoint reload breakage; different loaders." }, { "left": "issue:44355", "right": "issue:44655", "accept": false, - "reason": "Compiled file loading errors vs pipeline save_pretrained failure; unrelated." + "reason": "Compiled-file execution errors are unrelated to saving Pipeline objects." }, { "left": "issue:44355", "right": "issue:45071", "accept": false, - "reason": "Compiled file loading errors vs PretrainedConfig type checking regression; different issues." + "reason": "Runtime compiled-file issue and PretrainedConfig type-checking regression are different bugs." }, { "left": "issue:44222", "right": "issue:44230", "accept": false, - "reason": "Both mention FP8/Qwen, but one is save_pretrained moE and the other is inference support; not the same concrete bug." + "reason": "FP8 save_pretrained and FP8 inference support are different change requests/issues." }, { "left": "issue:43827", "right": "issue:44961", "accept": false, - "reason": "Docs still referencing pipeline() vs unrelated 'racoon' issue." + "reason": "Documentation cleanup after pipeline removals is unrelated to the placeholder issue." }, { - "left": "issue:33453", - "right": "issue:34689", + "left": "issue:43010", + "right": "issue:45656", "accept": false, - "reason": "Tokenizer loading regression vs model loading breakage for Llama 3.2 multimodal; different model-specific issues." + "reason": "Cache update no_grad concern is not the same as optimizer stepping twice under deepspeed." }, { - "left": "issue:43010", - "right": "issue:45656", + "left": "issue:33453", + "right": "issue:34689", "accept": false, - "reason": "Cache/layer no-grad topic vs optimizer step being called twice with deepspeed; unrelated." + "reason": "Tokenizer loading regression and model loading regression are different concrete failures." }, { "left": "issue:43299", "right": "issue:44556", "accept": false, - "reason": "Qwen3VL MoE model loading breakage in v5 dev vs checkpoint reload incompatibility after upgrade; similar theme but different concrete failures." + "reason": "Specific Qwen3VL MoE load breakage is not the same as generic checkpoint reload incompatibility." }, { "left": "issue:42757", "right": "issue:43576", "accept": false, - "reason": "Import error for is_offline_mode vs transformers env command broken; different symptoms and code paths." - }, - { - "left": "issue:33453", - "right": "issue:36296", - "accept": false, - "reason": "Tokenizer loading regression vs tensor parallel training bug; unrelated." + "reason": "Dependency import error and the v5 env command breakage do not show the same code-path problem." }, { "left": "issue:43576", "right": "issue:45071", "accept": false, - "reason": "Broken CLI command vs PretrainedConfig type checking regression; not the same bug." + "reason": "Env command breakage and config type-checking are unrelated." }, { "left": "issue:43576", "right": "issue:43824", "accept": false, - "reason": "CLI env command issue vs missing Qwen2.5-VL export; unrelated." + "reason": "Env command bug and Qwen2.5-VL import error are different failures." }, { "left": "issue:44961", "right": "issue:45468", "accept": false, - "reason": "Placeholder 'racoon' issue vs Gemma-4 audio positional encoding bug; no relation." + "reason": "No meaningful relation between the two issues." } ] }, @@ -13895,7 +13820,6 @@ "cluster_id": "cluster-15354-387", "nodes": [ "issue:31795", - "issue:33453", "issue:36296", "issue:38617", "issue:41628", @@ -13932,6 +13856,7 @@ "issue:44829", "issue:44861", "issue:44908", + "issue:44938", "issue:44961", "issue:45003", "issue:45071", @@ -13944,12 +13869,12 @@ "issue:45753" ], "soft_pairs": [ - "issue:43010|issue:43976", "issue:31795|issue:38617", + "issue:43010|issue:43976", "issue:41628|issue:42548", "issue:38617|issue:43576", "issue:44908|issue:44961", - "issue:43576|issue:44351", + "issue:43576|issue:44938", "issue:43010|issue:43232", "issue:36296|issue:43901", "issue:41628|issue:44355", @@ -13961,8 +13886,8 @@ "issue:44393|issue:45753", "issue:43576|issue:44556", "issue:44393|issue:45071", - "issue:41669|issue:43673", "issue:44038|issue:45125", + "issue:41669|issue:43673", "issue:44556|issue:45092", "issue:44373|issue:44961", "issue:44704|issue:44908", @@ -13982,98 +13907,96 @@ "issue:44393|issue:45478", "issue:44315|issue:44829", "issue:44077|issue:44861", - "issue:43827|issue:44908", - "issue:31795|issue:33453", - "issue:41669|issue:44908" + "issue:43827|issue:44908" ], - "prepared_review_unit_hash": "a8175a4d333f06e873478f26fb7c530d41426721820a44a661813735d0aa546e", + "prepared_review_unit_hash": "fbab0301503245c833b30e8d235e2d597de5a6ad5a62b741b93e3a7ff3073e85", "component_budget": { "node_count": 387, "item_count": 387, "soft_pair_count": 1455, - "serialized_chars": 749552, - "estimated_input_tokens": 187388, - "estimated_eval_tokens": 375032 + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 41, - "serialized_chars": 39349, - "estimated_input_tokens": 9838, - "estimated_eval_tokens": 19932 + "soft_pair_count": 39, + "serialized_chars": 38756, + "estimated_input_tokens": 9689, + "estimated_eval_tokens": 19634 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a noisy mix of unrelated docs, model-specific bugs, and import/regression reports. The only soft duplicate I\u2019d accept is the lazy-loading/import-path pair (44273/45003); everything else is too broad or clearly different.", - "confidence": 0.69, - "canonical_issue_reason": "issue:41669 is the broadest concrete import-system report in the set and best captures the cluster\u2019s shared theme of top-level import/lazy-loading regressions.", + "summary": "The set is mostly a grab bag of unrelated bug reports; the soft-similarity pairs are not strong enough to treat as duplicates. I rejected all soft edges and picked the most central import/performance report as the best issue representative.", + "confidence": 0.91, + "canonical_issue_reason": "Issue 44246 is the clearest broad import-performance report in this set and best represents the import/loading theme shared by several nearby issues.", "canonical_pr_reason": null, - "best_issue_reason": "issue:41669 is the strongest representative because it is specific, system-wide, and plausibly connected to several of the import-related reports here.", + "best_issue_reason": "Issue 44246 has the strongest signal and widest relevance among the issues here; it is the most suitable overall issue representative.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:43010", - "right": "issue:43976", + "left": "issue:31795", + "right": "issue:38617", "accept": false, - "reason": "Cache/no_grad bug vs Python-version support issue; unrelated." + "reason": "Documentation confusion vs an ImportError for a different symbol; unrelated bugs." }, { - "left": "issue:31795", - "right": "issue:38617", + "left": "issue:43010", + "right": "issue:43976", "accept": false, - "reason": "Doc wording confusion vs import error for a different symbol; not the same bug." + "reason": "Cache/no_grad behavior is unrelated to Python version compatibility." }, { "left": "issue:41628", "right": "issue:42548", "accept": false, - "reason": "Both are import errors, but for different top-level symbols; too broad to merge." + "reason": "Both are import errors, but for different classes/symbols and likely different causes." }, { "left": "issue:38617", "right": "issue:43576", "accept": false, - "reason": "Different failures: missing import vs broken env command." + "reason": "ImportError for layer_type_validation is unrelated to the v5 env command breakage." }, { "left": "issue:44908", "right": "issue:44961", "accept": false, - "reason": "Scheduler kwargs bug vs unrelated placeholder issue." + "reason": "Scheduler kwargs bug has nothing to do with the unrelated 'racoon' issue." }, { "left": "issue:43576", - "right": "issue:44351", + "right": "issue:44938", "accept": false, - "reason": "v5 env command failure vs missing HybridCache export; different concrete problems." + "reason": "Broken env command and Python 3.14 load failure are different failure modes." }, { "left": "issue:43010", "right": "issue:43232", "accept": false, - "reason": "Both generation-related, but different code paths and symptoms." + "reason": "Both touch generation internals, but one is no_grad decoration and the other is sync_gpus state update; not the same bug." }, { "left": "issue:36296", "right": "issue:43901", "accept": false, - "reason": "Tensor parallel training bug vs documentation mismatch." + "reason": "Tensor parallel training bug and docs mentioning return_all_scores are unrelated." }, { "left": "issue:41628", "right": "issue:44355", "accept": false, - "reason": "Import symbol failure vs compiled-file runtime error; unrelated." + "reason": "Importing AutoImageProcessor and running compiled Python files are different issues." }, { "left": "issue:36296", "right": "issue:43526", "accept": false, - "reason": "Tensor-parallel bug vs BeitImageProcessorFast label handling; different area." + "reason": "Tensor parallel training and BeitImageProcessorFast label reduction are unrelated." }, { "left": "issue:44261", @@ -14084,180 +14007,177 @@ { "left": "issue:44273", "right": "issue:45003", - "accept": true, - "reason": "Both point to lazy-loading/import machinery being broken; 45003 gives a likely root cause for 44273." + "accept": false, + "reason": "Lazy loading is broad; sys.modules access is a separate implementation bug." }, { "left": "issue:44623", "right": "issue:45230", "accept": false, - "reason": "Processor save-pretrained bug vs generic bug report." + "reason": "Processor save_pretrained file output bug and an unspecified bug report are not the same issue." }, { "left": "issue:44273", "right": "issue:45542", "accept": false, - "reason": "Lazy loading vs TF backend detection; different failures." + "reason": "Lazy loading problems are unrelated to the TensorFlow backend error." }, { "left": "issue:44393", "right": "issue:45753", "accept": false, - "reason": "Different Qwen3-VL/Qwen3.5 image bugs with unrelated symptoms." + "reason": "Two different Qwen3 image-model symptoms; not enough evidence of the same root cause." }, { "left": "issue:43576", "right": "issue:44556", "accept": false, - "reason": "v5 env command regression vs checkpoint reload incompatibility; not the same change." + "reason": "Different v5 regressions: env command vs checkpoint reload compatibility." }, { "left": "issue:44393", "right": "issue:45071", "accept": false, - "reason": "Model output bug vs PretrainedConfig type-checking regression." + "reason": "Qwen3-VL output issue and PretrainedConfig type checking are unrelated." }, { - "left": "issue:41669", - "right": "issue:43673", + "left": "issue:44038", + "right": "issue:45125", "accept": false, - "reason": "Import-performance/import-star issue vs generation cache missing; too different." + "reason": "Qwen3-VL bug and missing tensor-parallel plan for Qwen3_5Moe are different problems." }, { - "left": "issue:44038", - "right": "issue:45125", + "left": "issue:41669", + "right": "issue:43673", "accept": false, - "reason": "Different Qwen3-MoE/tensor-parallel issues; not enough evidence of same defect." + "reason": "Import star performance regression and missing GenerationMixin cache are different code paths." }, { "left": "issue:44556", "right": "issue:45092", "accept": false, - "reason": "Both involve v5 incompatibility, but one is general checkpoint reload and the other is a specific remote-code/meta-init case." + "reason": "Both concern v5 checkpoint loading, but for different models and distinct failure causes." }, { "left": "issue:44373", "right": "issue:44961", "accept": false, - "reason": "Docstring issue vs unrelated placeholder issue." + "reason": "Docstring typo and unrelated placeholder issue." }, { "left": "issue:44704", "right": "issue:44908", "accept": false, - "reason": "AutoProcessor kwarg forwarding vs scheduler kwargs; unrelated." + "reason": "Passing kwargs to cached_file and scheduler kwargs are unrelated bugs." }, { "left": "issue:42890", "right": "issue:44908", "accept": false, - "reason": "Test flakiness from missing seeds vs scheduler kwargs bug." + "reason": "Missing set_seed in tests is unrelated to scheduler behavior." }, { "left": "issue:44246", "right": "issue:44273", "accept": false, - "reason": "Slow import report vs lazy-loading malfunction; related area but not the same bug." + "reason": "Import slowness and lazy loading malfunction are related at a high level, but not the same concrete bug." }, { "left": "issue:43408", "right": "issue:44315", "accept": false, - "reason": "Model-type warning vs Liger kernel application bug; different code paths." + "reason": "Model-type warning and Liger Kernel model_init behavior are unrelated." }, { "left": "issue:43976", "right": "issue:44393", "accept": false, - "reason": "Python compatibility issue vs image output bug." + "reason": "Python support issue and Qwen3-VL visual output bug are unrelated." }, { "left": "issue:43352", "right": "issue:44393", "accept": false, - "reason": "Flash Attention support error vs Qwen3-VL output bug." + "reason": "Flash Attention support warning is unrelated to Qwen3-VL hallucination/output errors." }, { "left": "issue:44297", "right": "issue:45230", "accept": false, - "reason": "Tokenizer save metadata bug vs generic bug report." + "reason": "Tokenizer class mismatch on save is not the same as a generic bug report." + }, + { + "left": "issue:43576", + "right": "issue:44908", + "accept": false, + "reason": "CLI env breakage and scheduler kwargs handling are unrelated." + }, + { + "left": "issue:44908", + "right": "issue:45071", + "accept": false, + "reason": "Scheduler kwargs bug and PretrainedConfig type-checking regression are unrelated." }, { "left": "issue:44295", "right": "issue:45230", "accept": false, - "reason": "Position_ids buffer read error vs generic bug report." + "reason": "position_ids buffer access error does not match an unspecified bug report." }, { "left": "issue:43673", "right": "issue:44492", "accept": false, - "reason": "Generation cache regression vs typo in cache-strategy docs/code." + "reason": "Missing cache during generation and a typo in cache strategies are different issues." }, { "left": "issue:41669", "right": "issue:44351", "accept": false, - "reason": "Import-star/lazy-load cleanup vs missing HybridCache export; too indirect." + "reason": "Import-star performance problem and missing HybridCache import are different symptoms with different fixes." }, { "left": "issue:41669", "right": "issue:45341", "accept": false, - "reason": "Import-system issue vs testing_utils bug; unrelated." + "reason": "Import performance regression and a testing_utils bug are unrelated." }, { "left": "issue:41669", "right": "issue:44492", "accept": false, - "reason": "Import-system issue vs cache-strategy typo; unrelated." + "reason": "Import-star slowdown and a cache-strategy typo are unrelated." }, { "left": "issue:44393", "right": "issue:45478", "accept": false, - "reason": "Different Qwen3-related bugs with no shared concrete fix." + "reason": "Different Qwen3.5/Qwen3-VL issues with no shared concrete failure path." }, { "left": "issue:44315", "right": "issue:44829", "accept": false, - "reason": "Liger-kernel application bug vs flash_attention_3 training degradation; different issues." + "reason": "Liger Kernel application in model_init and flash_attention_3 training degeneration are unrelated." }, { "left": "issue:44077", "right": "issue:44861", "accept": false, - "reason": "Different model utility bugs; no shared code-path evidence." + "reason": "Optional post_init policy and tied-weight-key AttributeError are different model-internals bugs." }, { "left": "issue:43827", "right": "issue:44908", "accept": false, - "reason": "Docs still mention pipeline() vs scheduler kwargs bug." - }, - { - "left": "issue:31795", - "right": "issue:33453", - "accept": false, - "reason": "Doc confusion vs tokenizer loading regression; unrelated." - }, - { - "left": "issue:41669", - "right": "issue:44908", - "accept": false, - "reason": "Import/lazy-loading performance issue vs scheduler kwargs bug; unrelated." + "reason": "Docs still mentioning pipeline() and scheduler kwargs handling are unrelated." } ] }, - "evaluator_result": { - "accept": true, - "feedback": "The summary is grounded in the packet and stays mostly conservative. The lone accepted soft duplicate (44273/45003) is plausible from the issue titles, and the other rejections do not overclaim. The representative/canonical choice of 41669 is reasonable as a broad import-system report, even if the cluster overall is noisy." - }, - "evaluator_used": true, + "evaluator_result": null, + "evaluator_used": false, "retried": false, - "accepted_nontrivial_soft_edge": true, + "accepted_nontrivial_soft_edge": false, "error_kind": null, "error_message": null }, @@ -14316,6 +14236,8 @@ "issue:45468" ], "soft_pairs": [ + "issue:31795|issue:33453", + "issue:41669|issue:44908", "issue:42371|issue:44492", "issue:44655|issue:44861", "issue:43576|issue:44273", @@ -14357,274 +14279,286 @@ "issue:31515|issue:33453", "issue:43298|issue:44230" ], - "prepared_review_unit_hash": "69cea6aa46574de52c1f5e53ae96815b3105f4f24a9d72b2123173aedf465a8d", + "prepared_review_unit_hash": "568487fc00857f754cbd0b5a58bc54212cfdd0bd2cd4a7021b536d64b79ece3a", "component_budget": { "node_count": 387, "item_count": 387, "soft_pair_count": 1455, - "serialized_chars": 749552, - "estimated_input_tokens": 187388, - "estimated_eval_tokens": 375032 + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 }, "budget": { "node_count": 40, "item_count": 40, - "soft_pair_count": 40, - "serialized_chars": 34373, - "estimated_input_tokens": 8594, - "estimated_eval_tokens": 17444 + "soft_pair_count": 42, + "serialized_chars": 34970, + "estimated_input_tokens": 8743, + "estimated_eval_tokens": 17742 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This set is mostly a grab-bag of unrelated issue reports across loading, tokenizer, training, docs, and model-specific bugs. I did not find any soft edge that confidently represents the same underlying defect, so no duplicates should be merged here.", - "confidence": 0.89, - "canonical_issue_reason": "No single canonical issue stands out: the items span many unrelated bugs, and the few tokenizer/v5-loading reports are similar only at a high level, not clearly the same concrete defect.", + "summary": "No soft-edge pair looks like the same underlying bug/change; they mostly share only broad subsystem words. No PRs were provided, so issue triage is the only meaningful choice here.", + "confidence": 0.91, + "canonical_issue_reason": "issue:43502 is the clearest, most concrete user-facing bug: `local_files_only=True` still triggering API requests. It is narrowly scoped, reproducible, and broadly useful as the representative issue.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43066 is the most concrete and actionable issue in the loosely related tokenizer/v5 subset, so it is the best anchor if one must be chosen.", + "best_issue_reason": "issue:43502 is the best overall candidate because it describes a specific, high-signal loading/networking regression with an unambiguous expected behavior.", "best_pr_reason": null, "soft_edge_verdicts": [ + { + "left": "issue:31795", + "right": "issue:33453", + "accept": false, + "reason": "Different problems: documentation confusion for forward inputs vs a tokenizer loading regression." + }, + { + "left": "issue:41669", + "right": "issue:44908", + "accept": false, + "reason": "Unrelated topics: import-star performance overhead vs scheduler kwargs being ignored." + }, { "left": "issue:42371", "right": "issue:44492", "accept": false, - "reason": "Different topics: TF32 configuration warning vs a cache-strategy typo." + "reason": "Different bugs: TF32 API guidance vs a typo in cache strategies." }, { "left": "issue:44655", "right": "issue:44861", "accept": false, - "reason": "Both touch save/pretrained paths, but one is pipeline serialization and the other is tied-weights key handling; not the same bug." + "reason": "Both involve saving/model internals, but one is pipeline save_pretrained failure and the other is a tied-weights helper AttributeError." }, { "left": "issue:43576", "right": "issue:44273", "accept": false, - "reason": "Unrelated: v5 env command breakage vs lazy loading failure." + "reason": "Different features: v5 `transformers env` command breakage vs lazy loading malfunction." }, { "left": "issue:44360", "right": "issue:45468", "accept": false, - "reason": "Different model/code-path problems with no clear overlap." + "reason": "Distinct code paths and model components; not the same bug." }, { "left": "issue:45230", "right": "issue:45310", "accept": false, - "reason": "One is a vague bug report; the other is a specific Qwen3.5 MoE from_pretrained regression." + "reason": "Generic bug report vs a specific Qwen3.5 MoE from_pretrained failure; too vague to be the same issue." }, { "left": "issue:43827", "right": "issue:44492", "accept": false, - "reason": "Documentation cleanup vs a cache-strategy typo; unrelated." + "reason": "Docs still mentioning removed pipeline APIs is unrelated to a cache strategy typo." }, { "left": "issue:44273", "right": "issue:44908", "accept": false, - "reason": "Lazy loading and inverse_sqrt scheduler kwargs are unrelated subsystems." + "reason": "Lazy loading behavior and inverse_sqrt scheduler kwargs are unrelated." }, { "left": "issue:43519", "right": "issue:44485", "accept": false, - "reason": "Different model-specific bugs: timestamp calculation vs RoPE implementation." + "reason": "Different model-specific bugs: Qwen3VL timestamp math vs GLM-5 RoPE implementation." }, { "left": "issue:41669", "right": "issue:43704", "accept": false, - "reason": "Import-star performance cleanup vs VRAM leak in a model; not the same defect." + "reason": "Import-star performance issue is unrelated to VRAM leakage in multi-threaded dataloading." }, { "left": "issue:43519", "right": "issue:44908", "accept": false, - "reason": "Timestamp calculation bug and scheduler kwargs bug are unrelated." + "reason": "Different subsystems: processor timestamp calculation vs learning-rate scheduler config handling." }, { "left": "issue:31795", "right": "issue:34689", "accept": false, - "reason": "Documentation confusion vs model loading regression; different scope." + "reason": "Documentation wording issue vs model-loading regression for Llama 3.2 Vision." }, { "left": "issue:31795", "right": "issue:37428", "accept": false, - "reason": "Docs issue vs import error in flash attention utils; unrelated." + "reason": "Forward-argument docs confusion is unrelated to a flash-attention import error." }, { "left": "issue:44315", "right": "issue:45092", "accept": false, - "reason": "Both involve model initialization, but one is Liger kernel application and the other is meta-init compatibility for remote-code checkpoints; not the same bug." + "reason": "Liger Kernel application in model_init is not the same as remote-code checkpoint incompatibility with meta init." }, { "left": "issue:43704", "right": "issue:44485", "accept": false, - "reason": "VRAM leak in Qwen3ForCausalLM vs GLM-5 RoPE implementation issue; unrelated." + "reason": "Different model/runtime bugs: VRAM leak vs RoPE implementation." }, { "left": "issue:44485", "right": "issue:45468", "accept": false, - "reason": "Two different model components and failure modes." + "reason": "Separate model components and failures; not a single concrete bug." }, { "left": "issue:42371", "right": "issue:43704", "accept": false, - "reason": "TF32 behavior guidance vs a VRAM leak report; unrelated." + "reason": "TF32 configuration guidance is unrelated to the dataloader-thread VRAM leak." }, { "left": "issue:43502", "right": "issue:43519", "accept": false, - "reason": "Local-files-only network requests vs timestamp computation in a processor; different bugs." + "reason": "Network access during loading vs Qwen3VL timestamp calculation are different bugs." }, { "left": "issue:43519", "right": "issue:44623", "accept": false, - "reason": "Processor timestamp math vs processor.save_pretrained missing files; not the same issue." + "reason": "Processor timestamp math and missing files in processor.save_pretrained are distinct issues." }, { "left": "issue:43519", "right": "issue:43704", "accept": false, - "reason": "Different model/processor bugs with no shared code-path evidence." + "reason": "Different areas: processor timestamp computation vs VRAM leakage in threaded loading." }, { "left": "issue:44263", "right": "issue:45468", "accept": false, - "reason": "DSA indexer return-value issue vs Gemma4 audio positional encoding issue; unrelated." + "reason": "torch.split return handling in a DSA indexer is unrelated to Gemma4 audio positional encoding." }, { "left": "issue:43606", "right": "issue:45468", "accept": false, - "reason": "CPU offload device mismatch vs Gemma4 audio positional encoding; unrelated." + "reason": "CPU offload device mismatch has no clear relation to Gemma4 audio positional encoding." }, { "left": "issue:44261", "right": "issue:45468", "accept": false, - "reason": "Precision/config discussion for MLA q_a_layernorm vs Gemma4 audio encoding; unrelated." + "reason": "Missing rms_norm_eps in MLA q_a_layernorm is a different numerical bug than Gemma4 audio positional encoding." }, { "left": "issue:43704", "right": "issue:45341", "accept": false, - "reason": "VRAM leak in a model vs a testing_utils bug; not the same defect." + "reason": "VRAM leak in multi-threaded dataloading is unrelated to a testing_utils bug." }, { "left": "issue:42913", "right": "issue:43066", "accept": false, - "reason": "Same broad tokenizer-v5 theme, but one is a general behavior change report and the other is a specific decoder-type bug; not enough to call them the same underlying issue." + "reason": "Both are tokenizer/v5 related, but one is a broad behavior difference and the other is a specific wrong decoder type bug." }, { "left": "issue:44908", "right": "issue:44998", "accept": false, - "reason": "Scheduler kwargs bug vs an unrelated placeholder issue title." + "reason": "Scheduler kwargs bug is unrelated to the unrelated 'Unemployment' issue." }, { "left": "issue:44861", "right": "issue:45290", "accept": false, - "reason": "Tied-weights crash in saving logic vs chat-template crash on tool-call messages; different code paths." + "reason": "Tied-weights helper crash is unrelated to chat template tool-call handling." }, { "left": "issue:43525", "right": "issue:44861", "accept": false, - "reason": "Config attribute missing vs tied-weights key handling; unrelated." + "reason": "Missing pad_token_id in Llama4Config is a different config bug than tied-weights handling." }, { "left": "issue:43704", "right": "issue:44998", "accept": false, - "reason": "VRAM leak vs placeholder issue; unrelated." + "reason": "VRAM leak bug and the unrelated issue titled 'Unemployment' are clearly not the same." }, { "left": "issue:41669", "right": "issue:44998", "accept": false, - "reason": "Import-star cleanup vs placeholder issue; unrelated." + "reason": "Import-star performance issue is unrelated to the unrelated issue titled 'Unemployment'." }, { "left": "issue:43572", "right": "issue:44861", "accept": false, - "reason": "Missing pad_token_idx in a config vs tied-weights key crash; unrelated." + "reason": "StableLmConfig pad_token_idx removal issue is unrelated to tied-weights AttributeError." }, { "left": "issue:43066", "right": "issue:44998", "accept": false, - "reason": "Tokenizer v5 decoder-type bug vs placeholder issue; unrelated." + "reason": "Wrong tokenizer decoder type is unrelated to the unrelated issue titled 'Unemployment'." }, { "left": "issue:31515", "right": "issue:38617", "accept": false, - "reason": "Slow checkpoint loading vs an import error in configuration_utils; different defects." + "reason": "Slow checkpoint loading and an ImportError from configuration_utils are different failure modes." }, { "left": "issue:44360", "right": "issue:44998", "accept": false, - "reason": "DSA indexer bug vs placeholder issue; unrelated." + "reason": "DSA indexer functionality issue is unrelated to the unrelated issue titled 'Unemployment'." }, { "left": "issue:44230", "right": "issue:44623", "accept": false, - "reason": "FP8 model support request vs processor.save_pretrained file omission; unrelated." + "reason": "FP8 inference support for Qwen models is unrelated to processor.save_pretrained missing files." }, { "left": "issue:30333", "right": "issue:31515", "accept": false, - "reason": "MLflow reporting on failed training jobs vs slow from_pretrained checkpoint loading; unrelated." + "reason": "MLflow reporting on failed training jobs is unrelated to slow from_pretrained checkpoint loading." }, { "left": "issue:31515", "right": "issue:36296", "accept": false, - "reason": "Checkpoint loading performance vs tensor-parallel training bug; unrelated." + "reason": "Checkpoint loading performance is unrelated to a tensor parallel training bug." }, { "left": "issue:31515", "right": "issue:31795", "accept": false, - "reason": "Slow loading vs documentation confusion; different kinds of issues." + "reason": "Slow checkpoint loading and forward-argument documentation confusion are unrelated." }, { "left": "issue:44230", "right": "issue:45310", "accept": false, - "reason": "FP8 support request vs a Qwen3.5 MoE loading regression; not the same bug." + "reason": "FP8 support request and a Qwen3.5 MoE from_pretrained error are different issues." }, { "left": "issue:31515", "right": "issue:33453", "accept": false, - "reason": "Slow checkpoint loading vs tokenizer loading regression; different symptoms and likely different fixes." + "reason": "Checkpoint loading slowness is not the same as the tokenizer loading regression." }, { "left": "issue:43298", "right": "issue:44230", "accept": false, - "reason": "Continuous batching audio-model support vs FP8 inference support; unrelated." + "reason": "Continuous batching audio-model support and FP8 inference support are different feature gaps, not the same bug." } ] }, @@ -14654,7 +14588,7 @@ "soft_pairs": [ "issue:26350|issue:33260" ], - "prepared_review_unit_hash": "65fba15f8445349534702d92ac64990a4c1dcee2aa8b0752196446d58fe5d48b", + "prepared_review_unit_hash": "010abed66cc8c1d5a28f2876a9b7a7d8dc8445c25b6616c5d45cf08051f29b41", "component_budget": { "node_count": 2, "item_count": 2, @@ -14676,18 +14610,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are both community-contribution feature requests in the same general area, but they target different capabilities: Flash Attention 2 support vs GGUF support. They do not look like the same underlying issue.", - "confidence": 0.96, - "canonical_issue_reason": "Issue 26350 is the better canonical issue only if the cluster is meant to represent Flash Attention 2 architecture support requests, because it is the more specific and earlier request. But it is not a canonical representative for the GGUF request.", + "summary": "These are two unrelated community contribution issues: one requests Flash Attention 2 support for more architectures, the other requests GGUF support for more architectures. They do not describe the same underlying bug or change.", + "confidence": 0.97, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "Neither issue is a global best representative of the other; they request different features. If forced to choose a single issue for the cluster, 26350 is slightly more specific and established, but the two should not be merged as duplicates.", + "best_issue_reason": "No canonical issue exists here because the requests target different features; if one had to be picked as a representative issue, #26350 is the older and more discussed feature request.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:26350", "right": "issue:33260", "accept": false, - "reason": "Different feature requests: Flash Attention 2 support for more architectures versus GGUF support for more architectures. Same broad subsystem, but not the same underlying bug or change." + "reason": "Different feature requests with no shared code path or explicit linkage: Flash Attention 2 support vs GGUF support." } ] }, @@ -14724,12 +14658,12 @@ "issue:43834|issue:44365", "issue:44365|issue:44370" ], - "prepared_review_unit_hash": "89b811526feaa1e170b4ad18cf1b70b9a0c833b8c97eed4ee16b27e565b0275d", + "prepared_review_unit_hash": "2f2d8e8a2ec10bccf707ea15462303fc48908d959146f7e0b89e4ca9dd1da267", "component_budget": { "node_count": 4, "item_count": 4, "soft_pair_count": 6, - "serialized_chars": 4453, + "serialized_chars": 4454, "estimated_input_tokens": 1114, "estimated_eval_tokens": 2484 }, @@ -14737,7 +14671,7 @@ "node_count": 4, "item_count": 4, "soft_pair_count": 6, - "serialized_chars": 4453, + "serialized_chars": 4454, "estimated_input_tokens": 1114, "estimated_eval_tokens": 2484 }, @@ -14746,48 +14680,48 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These look like separate i18n translation-request issues rather than duplicates; only #27088 is concrete about Turkish, while the others are generic template instances.", - "confidence": 0.89, - "canonical_issue_reason": "#27088 is the most specific and actionable issue because it names a concrete target language (Turkish) instead of a placeholder template.", + "summary": "All four items are translation-tracking issues for docs, but they are not the same underlying request: one is Turkish-specific and the others are generic placeholder issues for different languages. No PRs are present.", + "confidence": 0.92, + "canonical_issue_reason": "Issue 27088 is the only concrete, language-specific report ([i18n-TR] Turkish); the others are template-style placeholders and not a better canonical anchor.", "canonical_pr_reason": null, - "best_issue_reason": "#27088 is the best representative issue in this cluster due to its specificity and clearer scope.", + "best_issue_reason": "27088 is the most specific and actionable issue, while the others are generic duplicates/templates rather than a concrete unique request.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:27088", "right": "issue:44370", "accept": false, - "reason": "Both are about docs translation, but one is concrete Turkish and the other is a generic placeholder; they are not clearly the same underlying request." + "reason": "Same broad translation-tracking template, but not the same language/request; different underlying issue." }, { "left": "issue:27088", "right": "issue:44365", "accept": false, - "reason": "The shared template suggests similar category, but the language target is not the same concrete issue, so this is not a duplicate." + "reason": "These are different translation targets, so they are not duplicates of the same bug/request." }, { "left": "issue:27088", "right": "issue:43834", "accept": false, - "reason": "Generic translation template versus a specific Turkish request; no evidence they describe the same change." + "reason": "Turkish-specific issue vs a generic language placeholder; related topic but not the same underlying issue." }, { "left": "issue:43834", "right": "issue:44370", "accept": false, - "reason": "These are template-like translation issues with no concrete language match shown, so they should not be merged as duplicates." + "reason": "Both are generic translation placeholders, but they do not identify the same concrete language or request." }, { "left": "issue:43834", "right": "issue:44365", "accept": false, - "reason": "Same template text alone is insufficient; they could refer to different languages and therefore different requests." + "reason": "Template similarity only; the issues are for different language translations, not one shared request." }, { "left": "issue:44365", "right": "issue:44370", "accept": false, - "reason": "Both are placeholder translation requests, but there is no concrete evidence they refer to the same language or same underlying task." + "reason": "They share the same template wording, but no evidence they refer to the same translation target or change." } ] }, @@ -14856,12 +14790,12 @@ "issue:43837|issue:45027", "issue:43837|issue:45177" ], - "prepared_review_unit_hash": "8435e75656feb7c954bc1806a8a411222bc132c7fa0e47c0dd948b4e145f7e07", + "prepared_review_unit_hash": "7ab19ac4bf0378c06c5dae49fbf63d38d86575ae7e6d809da951d59571ca8159", "component_budget": { "node_count": 14, "item_count": 14, "soft_pair_count": 28, - "serialized_chars": 17652, + "serialized_chars": 17650, "estimated_input_tokens": 4413, "estimated_eval_tokens": 9082 }, @@ -14869,7 +14803,7 @@ "node_count": 14, "item_count": 14, "soft_pair_count": 28, - "serialized_chars": 17652, + "serialized_chars": 17650, "estimated_input_tokens": 4413, "estimated_eval_tokens": 9082 }, @@ -14878,24 +14812,24 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are mostly separate model-request issues across different architectures and modalities. They share a generic \"add support\" theme, but not the same underlying change, so no soft edges should be merged.", - "confidence": 0.96, - "canonical_issue_reason": "Issue 37015 is the best representative of the cluster: it is a typical, active model-add request and the set has no clear true duplicate to collapse into a single canonical bug.", - "canonical_pr_reason": null, - "best_issue_reason": "37015 is the strongest overall representative by discussion activity and because it matches the common pattern in this cluster: a concrete request to add a model to the library.", - "best_pr_reason": null, + "summary": "This cluster is a set of separate model-addition requests, not duplicate reports of the same bug/change. The soft edges are mostly name-similarity across different model families or modalities (text, audio, vision), so none should be merged.", + "confidence": 0.97, + "canonical_issue_reason": "No single canonical issue fits the cluster: these are distinct model requests for different architectures and modalities, not one underlying defect or feature.", + "canonical_pr_reason": "No PRs are present in this cluster.", + "best_issue_reason": "No single issue is a good global representative; any choice would omit unrelated requests in other model families/modalities.", + "best_pr_reason": "No PRs are present in this cluster.", "soft_edge_verdicts": [ { "left": "issue:36817", "right": "issue:41211", "accept": false, - "reason": "Different requested models (EuroBert vs DEIMv2); same broad \"add model\" theme only." + "reason": "EuroBert config support and DEIMv2 are unrelated model additions." }, { "left": "issue:36817", "right": "issue:43671", "accept": false, - "reason": "EuroBert config work and Qwen3-TTS support are unrelated model additions." + "reason": "EuroBert is a text-model config request; Qwen3-TTS is an audio generation request." }, { "left": "issue:37015", @@ -14907,61 +14841,61 @@ "left": "issue:37015", "right": "issue:44121", "accept": false, - "reason": "NeoBERT and the weight-sparse transformer request are distinct architectures and code paths." + "reason": "NeoBERT and circuit-sparsity/OpenAI weight-sparse transformer are unrelated model requests." }, { "left": "issue:37015", "right": "issue:43909", "accept": false, - "reason": "NeoBERT and LFM2.5 Audio target different model types and implementations." + "reason": "NeoBERT is a text model; LFM2.5 Audio is an audio model." }, { "left": "issue:37015", "right": "issue:42503", "accept": false, - "reason": "NeoBERT and ModernVBERT are separate model requests, not the same bug/change." + "reason": "NeoBERT and ModernVBERT are different model architectures, not the same change." }, { "left": "issue:37015", "right": "issue:45027", "accept": false, - "reason": "NeoBERT and Voxtral TTS are different models with different support requirements." + "reason": "NeoBERT and Voxtral TTS support are unrelated model additions." }, { "left": "issue:41115", "right": "issue:43837", "accept": false, - "reason": "MiniCPM3 and Qwen3-ASR are unrelated model additions." + "reason": "MiniCPM3 architecture support and Qwen3-ASR support are different concrete changes." }, { "left": "issue:41115", "right": "issue:44121", "accept": false, - "reason": "MiniCPM3 support and the weight-sparse transformer request do not describe the same concrete change." + "reason": "MiniCPM3 and the weight-sparse transformer request are not the same model or code path." }, { "left": "issue:41115", "right": "issue:42503", "accept": false, - "reason": "MiniCPM3 and ModernVBERT are different model architectures." + "reason": "MiniCPM3 and ModernVBERT are separate model additions." }, { "left": "issue:41115", "right": "issue:45027", "accept": false, - "reason": "MiniCPM3 and Voxtral-4B-TTS-2603 are unrelated requests." + "reason": "MiniCPM3 support and Voxtral TTS support are distinct requests." }, { "left": "issue:41211", "right": "issue:43671", "accept": false, - "reason": "DEIMv2 and Qwen3-TTS are different model families and tasks." + "reason": "DEIMv2 and Qwen3-TTS are unrelated model families with different implementations." }, { "left": "issue:42503", "right": "issue:43837", "accept": false, - "reason": "ModernVBERT models and Qwen3-ASR are not the same underlying support request." + "reason": "ModernVBERT and Qwen3-ASR are different models and different tasks." }, { "left": "issue:42503", @@ -14973,85 +14907,79 @@ "left": "issue:42503", "right": "issue:42738", "accept": false, - "reason": "ModernVBERT and the generic BERT-like/RoPE request overlap loosely in domain, but they are not the same concrete change." + "reason": "ModernVBERT and BERT-like RoPE support are not the same concrete model change." }, { "left": "issue:42738", "right": "issue:45027", "accept": false, - "reason": "A generic BERT-like RoPE request and Voxtral TTS support are not the same model or implementation path." + "reason": "BERT-like RoPE support and Voxtral TTS support address different model classes." }, { "left": "issue:43671", "right": "issue:43837", "accept": false, - "reason": "Qwen3-TTS and Qwen3-ASR share a family name but are different modality-specific features." + "reason": "TTS and ASR are related audio domains but separate model requests and code paths." }, { "left": "issue:43671", "right": "issue:45027", "accept": false, - "reason": "Different TTS models; no evidence they are the same support change." + "reason": "Both are TTS-related, but they request different models and likely different integrations." }, { "left": "issue:43671", "right": "issue:45177", "accept": false, - "reason": "Qwen3-TTS and DeepSeek-OCR2 are unrelated model requests." - }, - { - "left": "issue:43671", - "right": "issue:44121", - "accept": false, - "reason": "A TTS support request and a sparse transformer architecture request are not duplicates." + "reason": "Qwen3-TTS and DeepSeek-OCR2 are different modalities and unrelated changes." }, { "left": "issue:43837", "right": "issue:45322", "accept": false, - "reason": "Qwen3-ASR and EUPE are unrelated additions." + "reason": "ASR support and EUPE (a perception encoder) are not the same model addition." }, { "left": "issue:43837", "right": "issue:43909", "accept": false, - "reason": "Qwen3-ASR and LFM2.5 Audio are different audio-model requests, not one concrete fix." + "reason": "Qwen3-ASR and LFM2.5 Audio are different audio models with different support needs." }, { "left": "issue:43909", "right": "issue:44205", "accept": false, - "reason": "Audio model support and SAM3-LiteText are different model types with different code paths." + "reason": "LFM2.5 Audio and SAM3-LiteText are different model requests with no shared underlying bug." }, { "left": "issue:43671", "right": "issue:43909", "accept": false, - "reason": "Qwen3-TTS and LFM2.5 Audio are separate audio-model requests." + "reason": "Qwen3-TTS and LFM2.5 Audio are separate audio-model integrations." }, { "left": "issue:44121", "right": "issue:45322", "accept": false, - "reason": "Weight-sparse transformer support and EUPE are different model additions." + "reason": "Weight-sparse transformer support and EUPE are unrelated model requests." }, { "left": "issue:43837", "right": "issue:44121", "accept": false, - "reason": "Qwen3-ASR and the weight-sparse transformer request are unrelated." + "reason": "Qwen3-ASR and the weight-sparse transformer request are different concrete code changes." }, { "left": "issue:43837", "right": "issue:45027", "accept": false, - "reason": "Qwen3-ASR and Voxtral TTS are different task/model combinations." + "reason": "Qwen3-ASR and Voxtral TTS are different audio tasks and integrations." }, { "left": "issue:43837", "right": "issue:45177", "accept": false, - "reason": "Qwen3-ASR and DeepSeek-OCR2 are unrelated model requests." + "reason": "ASR support and DeepSeek-OCR2 are different modalities and model families." } ] }, @@ -15092,12 +15020,12 @@ "issue:43668|issue:44031", "issue:43668|issue:43813" ], - "prepared_review_unit_hash": "83ce738025620504797af04b706c29b3de06d1c1ea0542e268a6ef2d4291fbc9", + "prepared_review_unit_hash": "3a21da2f46fbef8e76b79c47ec7b9c8b2646d1585ebe7dc1bf347aeffa8541a8", "component_budget": { "node_count": 6, "item_count": 6, "soft_pair_count": 8, - "serialized_chars": 6292, + "serialized_chars": 6291, "estimated_input_tokens": 1573, "estimated_eval_tokens": 3402 }, @@ -15105,7 +15033,7 @@ "node_count": 6, "item_count": 6, "soft_pair_count": 8, - "serialized_chars": 6292, + "serialized_chars": 6291, "estimated_input_tokens": 1573, "estimated_eval_tokens": 3402 }, @@ -15114,60 +15042,60 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The cluster is a loose collection of unrelated closed issues across models, configs, loss plumbing, and tokenizer warnings. None of the soft pairs look like the same underlying bug or change, so no duplicate merges are justified.", - "confidence": 0.92, - "canonical_issue_reason": "No single issue is a clear canonical duplicate target because the items describe different bugs in different components rather than one shared defect.", + "summary": "These issues are not duplicates; they span unrelated bugs across different models, config typing, tokenizer warnings, and a typo. All soft edges should be rejected.", + "confidence": 0.97, + "canonical_issue_reason": "No clear canonical issue: the cluster is heterogeneous, with unrelated topics such as normalization semantics, loss-plumbing, config type hints, a typo, and tokenizer warnings.", "canonical_pr_reason": null, - "best_issue_reason": "issue:44031 is the strongest standalone issue in the set: it has the most discussion activity and describes a concrete, user-visible regression with a specific version trigger.", + "best_issue_reason": "No issue is a strong cluster representative; 44031 is the broadest title, but it is still unrelated to the other reports.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:37934", "right": "issue:43668", "accept": false, - "reason": "One is about Llama4TextL2Norm possibly implementing RMSNorm; the other is a ModernBERTConfig type-hint problem. Different models and different failure modes." + "reason": "Both mention normalization-related terms, but one is about Llama4TextL2Norm semantics while the other is a ModernBERTConfig type hint; different bug classes." }, { "left": "issue:42936", "right": "issue:43240", "accept": false, - "reason": "Mask2Former ignore_value being unused is a model-loss configuration bug, while kwargs not being passed affects loss-function plumbing in general. Not the same concrete bug." + "reason": "Mask2former ignore_value handling and loss kwargs plumbing are different code paths and different failure modes." }, { "left": "issue:42936", "right": "issue:44031", "accept": false, - "reason": "Mask2Former loss ignore_value and tokenizer regex warning regression are unrelated subsystems and symptoms." + "reason": "A model loss-argument bug and a tokenizer regex warning after 4.57.3 are unrelated." }, { "left": "issue:42936", "right": "issue:43813", "accept": false, - "reason": "A missing ignore_value usage in Mask2Former is unrelated to a typo in a quantization operation name." + "reason": "Ignore_value usage in Mask2former is not the same underlying issue as a quantization-attribute typo." }, { "left": "issue:42936", "right": "issue:43668", "accept": false, - "reason": "These concern different models and different kinds of issues: ignored loss parameter vs incorrect config type annotation." + "reason": "One is a functional bug in Mask2former loss handling; the other is an incorrect type hint in a config class." }, { "left": "issue:43240", "right": "issue:43668", "accept": false, - "reason": "Loss kwargs forwarding and ModernBERTConfig type hints do not describe the same bug or fix path." + "reason": "Passing kwargs into a loss function and fixing a config type hint are unrelated changes." }, { "left": "issue:43668", "right": "issue:44031", "accept": false, - "reason": "Incorrect norm_eps type hint in ModernBERTConfig is unrelated to tokenizer regex warning behavior." + "reason": "A ModernBERTConfig annotation error and a tokenizer regex warning are different subsystems and different fixes." }, { "left": "issue:43668", "right": "issue:43813", "accept": false, - "reason": "A config type-hint issue and a typo in quantization_operation are different code paths and fixes." + "reason": "Incorrect type hints and a typo in orig_conversion.quantization_operation are distinct issues." } ] }, @@ -15197,7 +15125,7 @@ "soft_pairs": [ "issue:41084|issue:43037" ], - "prepared_review_unit_hash": "36f56e65eef3ba129411d2ddad4b7a8cb299bf350c1aaa08ce1bd755a2e88d06", + "prepared_review_unit_hash": "4be7d4c65e5d2f7b80f1a57a96bd17fa29b8d56328e91c23d5ffb197dbee0c76", "component_budget": { "node_count": 2, "item_count": 2, @@ -15219,18 +15147,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The two issues are not duplicates: one is about block decoding, the other is about DeepSeek v3.2 model support. They only share weak textual similarity.", - "confidence": 0.96, - "canonical_issue_reason": null, + "summary": "These two issues are not duplicates: one is a low-level block decoding request, while the other is a model-specific DeepSeek v3.2 support request.", + "confidence": 0.97, + "canonical_issue_reason": "Neither issue is a clear canonical duplicate target for the other; they describe different problem types and different expected changes.", "canonical_pr_reason": null, - "best_issue_reason": "43037 is the more specific and actionable issue title; 41084 is broader and appears unrelated to it.", + "best_issue_reason": "Issue 43037 is the better standalone tracking issue because it is more specific and actionable than the vague 'Set Block Decoding' request.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:41084", "right": "issue:43037", "accept": false, - "reason": "Different underlying topics and code paths: block decoding vs model support. The similarity is only superficial." + "reason": "The titles and likely scopes differ substantially: one concerns block decoding mechanics, the other concerns adding support for a specific model version. They do not appear to describe the same underlying bug or change." } ] }, @@ -15260,7 +15188,7 @@ "soft_pairs": [ "issue:43598|issue:43600" ], - "prepared_review_unit_hash": "a5dd25a823e112c738c606ddb1193508254c89eb821b2d014ed2496a667b9718", + "prepared_review_unit_hash": "df977e7412363678a5c4d3a2f4b760f1e1c75ee16ed995b979db0031df6fc483", "component_budget": { "node_count": 2, "item_count": 2, @@ -15282,18 +15210,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The two issues are both about Trainer customization, but they target different behaviors: one is about `remove_unused_column`, the other about FLOPS estimation for custom trainers. They do not look like the same underlying bug or change.", - "confidence": 0.94, - "canonical_issue_reason": "No clear canonical issue: the topics are related but distinct, with different user-facing behavior and different likely code paths.", + "summary": "These are related to Trainer customization, but they describe different problems: one is about `remove_unused_column` being too rigid, the other is about FLOPS estimation for custom trainers. They are not duplicates.", + "confidence": 0.91, + "canonical_issue_reason": "Issue 43598 is the broader, more central Trainer customizability request; 43600 is a separate FLOPS-estimation edge case.", "canonical_pr_reason": null, - "best_issue_reason": "Neither issue is a strong global representative of the other since they address separate Trainer concerns rather than the same defect.", + "best_issue_reason": "43598 is the better representative issue for this cluster because it captures the broader customization concern and is less narrowly scoped than 43600.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43598", "right": "issue:43600", "accept": false, - "reason": "Reject: both mention custom trainers, but one concerns unused-column removal and the other concerns FLOPS estimation; these are separate problems and would not be merged as one fix." + "reason": "Different underlying bugs: one changes how Trainer drops unused columns, the other concerns FLOPS accounting for custom trainers. Shared theme only, not the same fix." } ] }, @@ -15323,7 +15251,7 @@ "soft_pairs": [ "issue:43684|issue:44405" ], - "prepared_review_unit_hash": "02fbdd5e5cba129baa67486036c14d08e05084365a522d7633107b5bff76228c", + "prepared_review_unit_hash": "45a4006c06bab89496ef3cc59191cc57f9b29428dee00755a44cf3ea26366afa", "component_budget": { "node_count": 2, "item_count": 2, @@ -15345,18 +15273,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These issues are related only at a very high level: both ask for Qwen model support, but they target different model families and different auto classes/tasks, so they are not duplicates.", - "confidence": 0.96, - "canonical_issue_reason": "No single canonical issue: #43684 is about registering Qwen3-Omni for AutoModel/AutoModelForConditionalGeneration, while #44405 is about AutoModelForSequenceClassification support for Qwen3.5 (Qwen3_5Config).", + "summary": "The two issues are about different Qwen variants and different AutoModel heads, so they are not duplicates and should not be merged.", + "confidence": 0.97, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "#43684 is broader and more representative of a registration gap, but it should not subsume #44405 because the requested task and config are different.", + "best_issue_reason": "Issue 43684 is the broader registration request, but it does not represent the same bug/feature as 44405; there is no true canonical issue here.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43684", "right": "issue:44405", "accept": false, - "reason": "Different underlying changes: one is model registration for Qwen3-Omni generative auto classes, the other is sequence-classification support for Qwen3.5/Qwen3_5Config. Same broad theme, but not the same bug or fix." + "reason": "Different model families and different heads: Qwen3-Omni registration for AutoModel/AutoModelForConditionalGeneration vs Qwen3.5 sequence-classification support. Not the same underlying change." } ] }, @@ -15393,12 +15321,12 @@ "issue:43845|issue:43859", "issue:43846|issue:43859" ], - "prepared_review_unit_hash": "17a8216cf0d0005bcbed37b49031f0f9dcfa1f5d42ad052c0592dfc5e8a169f5", + "prepared_review_unit_hash": "8e38dbeff14c4affdcb488e1c9d23498e08297ab0fd021c4e3b8c3c5041653ea", "component_budget": { "node_count": 4, "item_count": 4, "soft_pair_count": 6, - "serialized_chars": 4267, + "serialized_chars": 4265, "estimated_input_tokens": 1067, "estimated_eval_tokens": 2390 }, @@ -15406,7 +15334,7 @@ "node_count": 4, "item_count": 4, "soft_pair_count": 6, - "serialized_chars": 4267, + "serialized_chars": 4265, "estimated_input_tokens": 1067, "estimated_eval_tokens": 2390 }, @@ -15415,54 +15343,54 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The three issues titled \"huggingface \" look like duplicates of the same report; the \"Fett\" issue appears unrelated. I would cluster the huggingface issues together and keep issue 43845 as the representative.", - "confidence": 0.93, - "canonical_issue_reason": "Issue 43845 is the earliest of the near-identical \"huggingface \" reports and is a good representative for that duplicate set.", + "summary": "Three closed issues share the exact title \"huggingface \" and look like duplicates; the \"Fett\" issue does not look related from the available metadata.", + "confidence": 0.77, + "canonical_issue_reason": "Issue 43845 is the earliest of the identical-title cluster \"huggingface \" and is a clean representative for the duplicate set.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 43845 is the best global issue candidate because it is one of the duplicate huggingface reports and is the earliest among them; the other issue title \"Fett\" is unrelated.", + "best_issue_reason": "Issue 43845 is the best overall issue candidate because it is part of the clear duplicate cluster and was created first; 43859 has slightly more discussion but no stronger signal of being distinct.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43835", "right": "issue:43845", "accept": false, - "reason": "Different titles and no evidence they describe the same bug or request; \"Fett\" does not match the huggingface reports." + "reason": "Different title and no shared targets/files or body detail to indicate the same underlying issue." }, { "left": "issue:43835", "right": "issue:43846", "accept": false, - "reason": "Different titles and no shared concrete problem description; this looks unrelated to the huggingface issues." + "reason": "Different title and no concrete evidence that this is the same bug report." }, { "left": "issue:43835", "right": "issue:43859", "accept": false, - "reason": "Different titles and no indication of the same underlying issue; reject as unrelated." + "reason": "Different title with no supporting metadata tying it to the huggingface duplicates." }, { "left": "issue:43845", "right": "issue:43846", "accept": true, - "reason": "Same title \"huggingface \" and highly similar low-information reports; they plausibly refer to the same underlying issue." + "reason": "Exact same title, very close creation times, and no evidence of distinct scope." }, { "left": "issue:43845", "right": "issue:43859", "accept": true, - "reason": "Same title \"huggingface \" and similar sparse reports; this is likely the same duplicate issue." + "reason": "Exact same title and strong duplicate signal from the repeated report pattern." }, { "left": "issue:43846", "right": "issue:43859", "accept": true, - "reason": "Same title and no distinguishing details; these look like duplicates of the same huggingface report." + "reason": "Exact same title and same apparent issue report as the other huggingface entries." } ] }, "evaluator_result": { "accept": true, - "feedback": "The clustering is grounded in the packet: issue 43835 has a different title (\u201cFett\u201d) and no concrete linkage to the three \u201chuggingface \u201d issues, while 43845/43846/43859 share the same title and are reasonably treated as a duplicate cluster. The choice of 43845 as representative is also conservative and supported by the earliest created_at among that group." + "feedback": "The summary is grounded in the packet: 43845/43846/43859 all share the exact title \"huggingface \", while 43835 is clearly different. The issue-order and canonical choice are also supported by the timestamps. The soft-edge accepts are based on limited metadata, but they do not overstate beyond title/timing similarity." }, "evaluator_used": true, "retried": false, @@ -15500,88 +15428,88 @@ "issue:44572|issue:44937", "issue:44573|issue:44937" ], - "prepared_review_unit_hash": "ac6d4023436f2e7b4c8a6d790faa113b8fa49bfa9c0e0aaa836759f3729b7690", + "prepared_review_unit_hash": "fa113db7967fa440bbc1575ba422726461f06ac3a2c10b53d727cf97d195ba35", "component_budget": { "node_count": 5, "item_count": 5, "soft_pair_count": 9, - "serialized_chars": 5788, - "estimated_input_tokens": 1447, - "estimated_eval_tokens": 3150 + "serialized_chars": 5794, + "estimated_input_tokens": 1449, + "estimated_eval_tokens": 3154 }, "budget": { "node_count": 5, "item_count": 5, "soft_pair_count": 9, - "serialized_chars": 5788, - "estimated_input_tokens": 1447, - "estimated_eval_tokens": 3150 + "serialized_chars": 5794, + "estimated_input_tokens": 1449, + "estimated_eval_tokens": 3154 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These look like unrelated low-signal/spam issues rather than duplicates. One item is a generic documentation request, while the others are spammy or promotional titles with no concrete shared bug/change.", - "confidence": 0.96, - "canonical_issue_reason": "issue:44418 is the only non-spam-like, plausibly legitimate request in the set, so it is the best available anchor even though the cluster overall is weak.", + "summary": "The cluster appears to be a set of unrelated low-quality/spam issues with superficial title similarity. No pair has enough substantive content to justify duplicate merging as the same underlying bug/change.", + "confidence": 0.97, + "canonical_issue_reason": "Issue 44371 is the earliest of the clearly spam/placeholder issues and is the best available anchor for this cluster, though it is not a meaningful product bug report.", "canonical_pr_reason": null, - "best_issue_reason": "issue:44418 is the most suitable representative because it is the only item with a conventional issue title; the others are clearly spam/promotional or empty.", + "best_issue_reason": "44371 is the most suitable issue anchor because it is an early exact-match placeholder spam item; the rest are similarly non-substantive and do not provide a better canonical signal.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44371", "right": "issue:44572", "accept": false, - "reason": "Both are titled \"\" and there is no evidence they describe the same underlying problem beyond being spam." + "reason": "Both are just '' placeholders; there is no concrete bug or change description to establish the same underlying issue." }, { "left": "issue:44371", "right": "issue:44573", "accept": false, - "reason": "Both are titled \"\" with no body detail; this is not enough to treat them as the same issue." + "reason": "Both are just '' placeholders; identical titles alone are not enough to confirm a real duplicate bug/change." }, { "left": "issue:44371", "right": "issue:44418", "accept": false, - "reason": "Spam title versus a generic documentation suggestion; no shared concrete bug, feature, or target." + "reason": "One is spam and the other is a generic documentation suggestion; the overlap is only superficial wording, not the same underlying issue." }, { "left": "issue:44371", "right": "issue:44937", "accept": false, - "reason": "Spam title versus promotional \"Check out 'Google LLC'\"; clearly unrelated." + "reason": "The titles are unrelated spam-like phrases and do not indicate the same concrete bug or request." }, { "left": "issue:44418", "right": "issue:44572", "accept": false, - "reason": "Generic documentation enhancement versus spam; no same underlying change or bug." + "reason": "Generic docs-suggestion wording does not match a placeholder spam report well enough to call them the same issue." }, { "left": "issue:44418", "right": "issue:44573", "accept": false, - "reason": "Generic documentation enhancement versus spam; no evidence of duplication." + "reason": "Generic docs-suggestion wording does not match a placeholder spam report well enough to call them the same issue." }, { "left": "issue:44572", "right": "issue:44573", "accept": false, - "reason": "Both are spam titles, but there is no concrete issue text showing they are the same underlying report." + "reason": "Although both are '', there is no substantive report content to confirm they describe the same underlying bug/change." }, { "left": "issue:44572", "right": "issue:44937", "accept": false, - "reason": "Spam versus promotional title; unrelated." + "reason": "A placeholder spam title and a specific promotional title are not evidence of the same underlying issue." }, { "left": "issue:44573", "right": "issue:44937", "accept": false, - "reason": "Spam versus promotional title; unrelated." + "reason": "A placeholder spam title and a specific promotional title are not evidence of the same underlying issue." } ] }, @@ -15611,7 +15539,7 @@ "soft_pairs": [ "issue:45145|issue:45151" ], - "prepared_review_unit_hash": "6ebd917169c87d3de0cebc41a1b7f28bc53b05b3fd31a76ee12626be5edb920a", + "prepared_review_unit_hash": "c87c66c34e215718d3d6c1e4eacb20c2cfc1241fdbc8fdd809ce4a9ff21272bf", "component_budget": { "node_count": 2, "item_count": 2, @@ -15633,28 +15561,28 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "#45145 and #45151 look highly similar and may be the same report, but the packet only shows title/body-length similarity plus a soft-similarity signal. There isn\u2019t enough direct evidence here to confidently mark them as duplicates, so this should be treated as a likely related / possible duplicate pair rather than a confirmed one. #45145 is still the better canonical candidate because it has an inbound reference.", - "confidence": 0.69, - "canonical_issue_reason": "#45145 is the better canonical candidate if these turn out to be duplicates because it already has an inbound reference, but the evidence available here is only weak-to-moderate similarity rather than a confirmed duplicate link.", + "summary": "These two closed issues are strong near-duplicates: they share the same title, have substantial soft similarity, and #45145 is earlier with an inbound reference, making it the best canonical representative.", + "confidence": 0.92, + "canonical_issue_reason": "#45145 is the better canonical issue because it was created first and already has an inbound reference, while #45151 appears to be a later duplicate with no stronger distinguishing signal.", "canonical_pr_reason": null, - "best_issue_reason": "#45145 is the best issue to keep as the canonical thread for now because it has slightly more visibility via an inbound reference, though the duplication claim is only provisional.", + "best_issue_reason": "#45145 is the strongest issue for the cluster: it is the earliest instance, has the same title as #45151, and has the only inbound reference activity.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45145", "right": "issue:45151", - "accept": false, - "reason": "The titles are identical and the body lengths are close, but the packet does not provide the actual bodies or any explicit linkage, so this is not enough to confidently confirm the same underlying issue." + "accept": true, + "reason": "Accept as a near-duplicate: the titles match exactly and the soft-similarity signal is strong, with #45145 also predating #45151 and carrying the only inbound reference. We do not have body text here, so the judgment is based on title and available similarity signals rather than body-level evidence." } ] }, "evaluator_result": { "accept": false, - "feedback": "The summary overstates the evidence. The packet only shows strong title similarity, similar body length, and a soft-similarity signal; it does not provide the actual bodies or any explicit linkage confirming duplication. Since the pair\u2019s deterministic verdict is false, marking the issue pair as accepted/duplicate is too strong. The canonical choice of #45145 is plausible because it has an inbound reference, but the confidence and duplicate claim should be toned down to something more conservative like \u2018likely related/possible duplicate.\u2019" + "feedback": "The overall duplicate/near-duplicate conclusion is plausible, but the soft-edge rationale overstates what is in the packet: we do not have the issue bodies, so \"highly similar bodies\" is not directly grounded, and \"same closed issue type\" is not a meaningful distinction here. The verdict should be phrased more conservatively based on same title plus soft similarity evidence and earlier timestamp/inbound reference, without claiming stronger body-level evidence than provided." }, "evaluator_used": true, "retried": true, - "accepted_nontrivial_soft_edge": false, + "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null }, @@ -15731,7 +15659,7 @@ "soft_pairs": [ "issue:45672|issue:45674" ], - "prepared_review_unit_hash": "a336c419d30b9bd852e12e00cf1b7bab6e3b7d03203821ef6ed200de7aa47f78", + "prepared_review_unit_hash": "19b5752beca17e4e02fb32bbef0c5c56e57bd37c447436c2a8168588232d9592", "component_budget": { "node_count": 2, "item_count": 2, @@ -15753,18 +15681,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are two different 4-bit inference bugs: one is Gemma4 audio-module dtype handling with uint8 weights and `torch.finfo()`, the other is `BitsAndBytesConfig.llm_int8_skip_modules` unintentionally removing the default `lm_head` exclusion and triggering an `AssertionError`.", - "confidence": 0.97, - "canonical_issue_reason": null, + "summary": "These are two different 4-bit inference bugs: one in Gemma4/audio-module handling with uint8 weights and torch.finfo(), the other in BitsAndBytesConfig skip-module handling clearing the default lm_head exclusion. They are related by quantized inference but not the same underlying defect.", + "confidence": 0.96, + "canonical_issue_reason": "No single canonical issue fits both reports; they describe separate failure modes. If forced to pick the clearer representative of its own bug, issue #45674 is the more general BitsAndBytesConfig regression report.", "canonical_pr_reason": null, - "best_issue_reason": null, + "best_issue_reason": "Issue #45674 is the stronger standalone issue: it names the exact configuration regression, states the observed AssertionError, and has a clearer path to reproduction/fix than the Gemma4-specific torch.finfo() error.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45672", "right": "issue:45674", "accept": false, - "reason": "Different underlying code paths and failure modes: dtype/math handling in Gemma4 audio modules versus configuration logic for module skipping in BitsAndBytes. They are not the same bug and should not be merged." + "reason": "Different bugs and code paths: Gemma4/audio uint8 torch.finfo() TypeError vs. BitsAndBytesConfig llm_int8_skip_modules clearing lm_head exclusion and triggering an AssertionError. They are not mergeable as one fix." } ] }, @@ -15794,7 +15722,7 @@ "soft_pairs": [ "issue:45762|issue:45763" ], - "prepared_review_unit_hash": "b060f37cf83d61d8baf7a394652e5b28bced900fb14661d662e31793ffd05a28", + "prepared_review_unit_hash": "25fba1d63b8fab9cd7d84219496f61f159fc149e6e4640904ae244d4c18fbba1", "component_budget": { "node_count": 2, "item_count": 2, @@ -15816,18 +15744,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The two issues are title-only entries about different named individuals with different dates, and there is no substantive evidence they describe the same underlying problem.", - "confidence": 0.96, + "summary": "The two issues look like unrelated name-and-date entries with no shared description, references, or code context. They should not be clustered as duplicates.", + "confidence": 0.98, "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "Neither issue is a good canonical representative: both are closed, bodyless, and appear to be unrelated personal-name entries rather than duplicate bug reports.", + "best_issue_reason": null, "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45762", "right": "issue:45763", "accept": false, - "reason": "Different names and dates with no body, links, or shared technical context; they do not look like the same underlying issue." + "reason": "The titles only share a surname and date-like text; there is no evidence of the same underlying bug, request, or actionable issue." } ] }, @@ -15857,7 +15785,7 @@ "soft_pairs": [ "issue:45764|issue:45765" ], - "prepared_review_unit_hash": "d358283d1bfa5c844cbdf51a537ccb8b1a086761fe166e5dbc4dd530c3abe8e9", + "prepared_review_unit_hash": "fe409b76adca8244431eda270b257edc5801e5988bc6364f3a25952ac74ae71c", "component_budget": { "node_count": 2, "item_count": 2, @@ -15879,18 +15807,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These two closed issues do not appear to describe the same bug or change; they look like unrelated name/date entries with no technical content to deduplicate.", - "confidence": 0.96, - "canonical_issue_reason": "issue:45764 is only marginally preferable as the anchor because it was created slightly earlier, but neither issue is a credible canonical since both titles are just standalone name/date strings with no shared problem description.", + "summary": "The two closed issues appear to be unrelated name/date entries rather than the same bug or request, so this is not a duplicate cluster.", + "confidence": 0.98, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "Neither issue is a good representative for the cluster; there is no body, discussion, or technical context suggesting a real report to canonicalize.", + "best_issue_reason": null, "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45764", "right": "issue:45765", "accept": false, - "reason": "Reject: the titles are unrelated personal-name/date strings and there is no evidence they describe the same underlying issue." + "reason": "Titles only contain different personal names and dates with no shared problem statement, code path, or explicit linkage; they do not look like duplicates." } ] }, @@ -15949,154 +15877,154 @@ "pull_request:43823|pull_request:45795", "pull_request:43251|pull_request:43254" ], - "prepared_review_unit_hash": "36f6934f01baf6d8bde55f1db2a2cfa90ede3f4d88fe74340fec552a8f8ffbab", + "prepared_review_unit_hash": "335bfacae392884c7a669f7690c251829a1be6d44468f0d7f741390a8612b531", "component_budget": { "node_count": 13, "item_count": 13, "soft_pair_count": 19, - "serialized_chars": 35203, - "estimated_input_tokens": 8801, - "estimated_eval_tokens": 17858 + "serialized_chars": 35199, + "estimated_input_tokens": 8800, + "estimated_eval_tokens": 17856 }, "budget": { "node_count": 13, "item_count": 13, "soft_pair_count": 19, - "serialized_chars": 35203, - "estimated_input_tokens": 8801, - "estimated_eval_tokens": 17858 + "serialized_chars": 35199, + "estimated_input_tokens": 8800, + "estimated_eval_tokens": 17856 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "This cluster is mostly unrelated PRs that only share broad repository similarity. The only true duplicate-looking pair is the two `fixed_cross_entropy` patches for issue 43240; the rest are distinct model additions, docs, or refactors.", - "confidence": 0.9, + "summary": "The cluster is mostly heterogeneous and the soft similarities are superficial. The only clear duplicate-like pair is the two fixed_cross_entropy PRs, which address the same loss-wrapper bug and could plausibly be merged into one change.", + "confidence": 0.97, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 43251 is the clearest representative of the only duplicated change in the cluster: it directly fixes issue 43240 in `fixed_cross_entropy`, and its implementation already covers the kwargs supported by `nn.functional.cross_entropy`.", + "canonical_pr_reason": "PR 43251 is the better canonical representative: it directly fixes the reported issue, has the explicit issue link, and implements the kwargs forwarding in the loss wrapper in a complete way.", "best_issue_reason": null, - "best_pr_reason": "PR 43251 is the best canonical PR because it is issue-linked, narrowly scoped, and most directly expresses the underlying fix; PR 43254 is effectively the same change with a slightly different presentation.", + "best_pr_reason": "PR 43251 is the best global representative because it is the clearest standalone bug fix in the cluster and the only one with an explicit issue target; 43254 is a narrower variant of the same change.", "soft_edge_verdicts": [ { "left": "pull_request:39895", "right": "pull_request:43424", "accept": false, - "reason": "Videoprism model addition vs executorch dynamic-shape test; unrelated changes." + "reason": "Unrelated model-addition vs export test changes; same broad ecosystem only, not the same bug/change." }, { "left": "pull_request:41855", "right": "pull_request:43424", "accept": false, - "reason": "Tokenizer method additions vs executorch export test; no shared bug or code path." + "reason": "Tokenizer API additions and export-shape testing are unrelated changes." }, { "left": "pull_request:41855", "right": "pull_request:43823", "accept": false, - "reason": "Mistral tokenizer helpers vs adding MobileLLM; different models and purposes." + "reason": "Different models and different workstreams; no shared concrete code-path problem." }, { "left": "pull_request:42668", "right": "pull_request:43823", "accept": false, - "reason": "Processor robustness refactor vs adding MobileLLM; not the same underlying change." + "reason": "Processor robustness fixes for existing models versus adding a new model; not the same underlying change." }, { "left": "pull_request:43251", "right": "pull_request:43823", "accept": false, - "reason": "Cross-entropy kwargs fix vs new model implementation; unrelated." + "reason": "Loss-function kwargs fix is unrelated to adding MobileLLM." }, { "left": "pull_request:43251", "right": "pull_request:44827", "accept": false, - "reason": "Loss utility fix vs Mistral4 test/model changes; different code paths." + "reason": "Loss wrapper kwargs handling is unrelated to Mistral4 test/model changes." }, { "left": "pull_request:43424", "right": "pull_request:44827", "accept": false, - "reason": "Executorch test coverage vs Mistral4 test/model fixes; only generic testing similarity." + "reason": "Executorch dynamic-shape test coverage is unrelated to Mistral4 fixes." }, { "left": "pull_request:43424", "right": "pull_request:43823", "accept": false, - "reason": "Executorch export tests vs MobileLLM model addition; unrelated." + "reason": "Different functionality and code paths; only both are framework PRs." }, { "left": "pull_request:43823", "right": "pull_request:45655", "accept": false, - "reason": "New model addition vs config resolution ordering fix; no shared concrete bug." + "reason": "New model addition and class config resolution order fix do not share the same underlying bug/change." }, { "left": "pull_request:43823", "right": "pull_request:43995", "accept": false, - "reason": "MobileLLM addition vs Falcon output-interface refactor; different subsystems." - }, - { - "left": "pull_request:43823", - "right": "pull_request:45213", - "accept": false, - "reason": "New model addition vs model-creation skill/docs; documentation/tooling, not same change." - }, - { - "left": "pull_request:43823", - "right": "pull_request:45795", - "accept": false, - "reason": "MobileLLM model addition vs docs for audio/video processors; only broad docs similarity." + "reason": "New model addition versus Falcon refactoring; different targets and problems." }, { "left": "pull_request:43995", "right": "pull_request:45213", "accept": false, - "reason": "Falcon refactor vs model-creation skill/docs; unrelated." + "reason": "Falcon output-interface refactor and a model-creation skill are not the same change." }, { "left": "pull_request:43995", "right": "pull_request:44827", "accept": false, - "reason": "Falcon output-collection refactor vs Mistral4 test/model fixes; different code paths." + "reason": "Different models and different issues; no mergeable shared fix." }, { "left": "pull_request:44827", "right": "pull_request:45170", "accept": false, - "reason": "Mistral4 tests/model changes vs CLIP-like layernorm rename cleanup; unrelated." + "reason": "Mistral4 test fixes and CLIP-style layernorm renaming are unrelated." }, { "left": "pull_request:44827", "right": "pull_request:45795", "accept": false, - "reason": "Mistral4 fixes vs docs for audio/video processors; no shared underlying bug." + "reason": "Mistral4 test fixes and docs for audio/video processors are unrelated." }, { "left": "pull_request:44827", "right": "pull_request:45655", "accept": false, - "reason": "Mistral4 fixes vs `cls.config` resolution order fix; unrelated implementation areas." + "reason": "Different code paths: model tests versus config-resolution behavior." }, { "left": "pull_request:44827", "right": "pull_request:45213", "accept": false, - "reason": "Mistral4 fixes vs model-creation skill/docs; different scope." + "reason": "Only a shared auto/modeling file appears in a broad cluster sense; the actual changes are unrelated." + }, + { + "left": "pull_request:43823", + "right": "pull_request:45213", + "accept": false, + "reason": "Adding a model and adding an internal creation skill are not the same underlying change." + }, + { + "left": "pull_request:43823", + "right": "pull_request:45795", + "accept": false, + "reason": "Model addition versus docs-only processor guide; not the same change." }, { "left": "pull_request:43251", "right": "pull_request:43254", "accept": true, - "reason": "Both patch the same `fixed_cross_entropy` code path for issue 43240 and would plausibly be merged into one PR." + "reason": "Both PRs patch the same fixed_cross_entropy wrapper to support passing through cross-entropy kwargs, so they address the same concrete bug in the same code path and could be merged into one PR." } ] }, "evaluator_result": { "accept": true, - "feedback": "The analysis is conservative and grounded in the packet. The only accepted pair, 43251/43254, is clearly the same `fixed_cross_entropy` bugfix in the same file for the same issue target and could plausibly be merged into one PR. The remaining soft pairs are appropriately rejected as broad similarity rather than shared bugfix or mergeable duplicate changes." + "feedback": "The summary is grounded in the packet: PR 43251 and PR 43254 share the same file and explicit issue target, and the change in both is the same fixed_cross_entropy kwargs support. The other pair verdicts stay conservative by rejecting broad ecosystem similarity without claiming a shared bugfix. " }, "evaluator_used": true, "retried": false, @@ -16123,7 +16051,7 @@ "soft_pairs": [ "pull_request:40546|pull_request:42781" ], - "prepared_review_unit_hash": "496d198fd8e55089431df5366c84948ad129888071b2fb776086ac134feb7386", + "prepared_review_unit_hash": "48e29c877547d507c890160961185154419cd433d2329b245c04ffb9b850f5bb", "component_budget": { "node_count": 2, "item_count": 2, @@ -16145,18 +16073,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are related VibeVoice PRs, but they are not the same change: one is the base VibeVoice implementation, the other adds a separate realtime variant with additional tokenizer/feature/generation work. They should not be merged as duplicates.", - "confidence": 0.92, + "summary": "These two pull requests are related but not duplicates: PR 40546 adds the base VibeVoice model integration, while PR 42781 extends the family with VibeVoice Realtime and an acoustic tokenizer. They share several VibeVoice files, but the change goals and code paths are not the same.", + "confidence": 0.94, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #40546 is the better canonical PR because it is the original, broader VibeVoice implementation and appears more mature (non-draft, much higher review activity).", + "canonical_pr_reason": "PR 40546 is the better canonical representative because it introduces the core VibeVoice implementation and broad Transformers integration; PR 42781 is a later, narrower realtime extension with additional tokenizer and generation changes.", "best_issue_reason": null, - "best_pr_reason": "PR #40546 is the best overall PR to represent the cluster since it establishes the core model integration; #42781 is a follow-on realtime extension rather than the same concrete change.", + "best_pr_reason": "PR 40546 is the best single representative of the cluster because it is the original base VibeVoice PR and covers the main model import/integration work rather than the later realtime variant.", "soft_edge_verdicts": [ { "left": "pull_request:40546", "right": "pull_request:42781", "accept": false, - "reason": "Do not merge: both are in the same VibeVoice area, but #40546 implements the base model while #42781 is a draft for VibeVoice Realtime with extra acoustic tokenizer and generation changes. They touch overlapping files, but they are not the same underlying bug or feature and would not plausibly be one PR." + "reason": "Reject: they overlap on shared VibeVoice scaffolding, but 42781 is a distinct realtime/tokenizer extension rather than the same underlying bug or one mergeable change set." } ] }, @@ -16186,12 +16114,12 @@ "soft_pairs": [ "pull_request:40755|pull_request:41224" ], - "prepared_review_unit_hash": "a53f7de13e647e0aac5e405bd150be0a917d8410c457133530b34b29040e90b7", + "prepared_review_unit_hash": "a4f0dd266deaf10ea8b584c35b2e4d4bdc9f4b1b34c671751053549431a5fa2d", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4676, + "serialized_chars": 4675, "estimated_input_tokens": 1169, "estimated_eval_tokens": 2594 }, @@ -16199,7 +16127,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4676, + "serialized_chars": 4675, "estimated_input_tokens": 1169, "estimated_eval_tokens": 2594 }, @@ -16208,7 +16136,7 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The two pull requests are unrelated feature additions in different model families: TimesFM covariate forecasting vs. DINOv3 image classification support. They do not describe the same underlying change and should not be clustered as duplicates.", + "summary": "The two pull requests are unrelated feature additions in different model families: TimesFM covariate forecasting vs DINOv3 image classification support. They do not describe the same bug or change.", "confidence": 0.99, "canonical_issue_reason": null, "canonical_pr_reason": null, @@ -16219,7 +16147,7 @@ "left": "pull_request:40755", "right": "pull_request:41224", "accept": false, - "reason": "Different models, different code paths, and different user-facing features; they are not the same bug or change and could not be merged into one PR." + "reason": "Different code paths and domains: one adds TimesFM covariate forecasting utilities/docs/tests, the other adds DINOv3 image classification support and related auto/model changes. They are not the same underlying change." } ] }, @@ -16314,12 +16242,12 @@ "pull_request:41797|pull_request:45186", "pull_request:45186|pull_request:45630" ], - "prepared_review_unit_hash": "9cbc7197164be5125ef4c5d5474a90d9b8a13aa11f6c7ec7d9a0f87366aa9f78", + "prepared_review_unit_hash": "565e46cba748ec0ccc128fc81d8213c1930f227e3cfdd2973297dc84f2da1457", "component_budget": { "node_count": 3, "item_count": 3, "soft_pair_count": 2, - "serialized_chars": 10427, + "serialized_chars": 10428, "estimated_input_tokens": 2607, "estimated_eval_tokens": 5470 }, @@ -16327,7 +16255,7 @@ "node_count": 3, "item_count": 3, "soft_pair_count": 2, - "serialized_chars": 10427, + "serialized_chars": 10428, "estimated_input_tokens": 2607, "estimated_eval_tokens": 5470 }, @@ -16336,24 +16264,24 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are three separate new-model addition PRs. They share common Transformers boilerplate (docs, auto mappings, model registration), but each targets a different model family and code path, so they are not duplicates.", + "summary": "These are three distinct model-addition pull requests (DeepSeek OCR, Isaac, and Kimi2-6). They share framework plumbing and docs/auto-registration patterns, but they do not fix the same concrete code path or represent the same change.", "confidence": 0.96, "canonical_issue_reason": null, - "canonical_pr_reason": null, + "canonical_pr_reason": "No canonical duplicate PR: each changes a different model integration, with different files, configs, and model-specific implementations.", "best_issue_reason": null, - "best_pr_reason": null, + "best_pr_reason": "PR 45186 is the most representative of the shared pattern (a new model integration), but it is still not a duplicate of the others.", "soft_edge_verdicts": [ { "left": "pull_request:41797", "right": "pull_request:45186", "accept": false, - "reason": "Different model families and implementations (DeepSeek OCR vs Isaac); overlap is limited to standard model-registration/docs boilerplate." + "reason": "Different models and codepaths: DeepSeek OCR vs Isaac. Shared registration/docs scaffolding is not enough to treat them as the same change." }, { "left": "pull_request:45186", "right": "pull_request:45630", "accept": false, - "reason": "Different additions (Isaac vs Kimi2-6) with different model-specific files and tests; similarity is only in shared integration scaffolding." + "reason": "Different models and integrations: Isaac vs Kimi2-6. Similar onboarding pattern, but not the same underlying bug or feature." } ] }, @@ -16383,7 +16311,7 @@ "soft_pairs": [ "pull_request:42975|pull_request:42976" ], - "prepared_review_unit_hash": "bfe1ef52932bdfadbc58ede293c19d08192c4ea7af1681546f38ad3100c69513", + "prepared_review_unit_hash": "86b0f308da69d8582d76e0bbb3b9e0b24ca7192adfe20e25b5e6f15b35eb6f66", "component_budget": { "node_count": 2, "item_count": 2, @@ -16405,18 +16333,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are two separate maintenance PRs touching overlapping GitHub workflow files, but they upgrade different actions and address different dependency bumps, so they should not be clustered as duplicates.", - "confidence": 0.94, + "summary": "Both PRs are broad GitHub Actions dependency bumps, but they fix different things: #42975 upgrades several workflow actions for Node 24 compatibility, while #42976 mainly bumps docker/build-push-action in Docker image workflows. They overlap in files, but not in underlying change or failure mode.", + "confidence": 0.95, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 42975 is the broader, more representative workflow-runtime compatibility upgrade: it updates multiple GitHub actions to Node 24-compatible versions across many workflows. PR 42976 is a narrower Docker action version bump set and does not cover the same change set.", + "canonical_pr_reason": "PR #42975 is the better canonical representative because it is the broader, explicitly stated Node 24 compatibility upgrade across many workflows.", "best_issue_reason": null, - "best_pr_reason": "PR 42975 is the better cluster representative because it is the larger, more comprehensive GitHub Actions compatibility update and better matches the overall maintenance theme of the shared workflow files.", + "best_pr_reason": "PR #42975 is the strongest overall representative of the cluster: it has the clearer problem statement and broader workflow-action modernization scope.", "soft_edge_verdicts": [ { "left": "pull_request:42975", "right": "pull_request:42976", "accept": false, - "reason": "Reject: both are workflow maintenance PRs, but 42975 upgrades checkout/cache/upload-artifact for Node 24 compatibility while 42976 bumps docker/build-push-action and a few related workflow actions. They touch some of the same files, but not the same underlying bug or change." + "reason": "They are related workflow maintenance changes, but not the same concrete bug/change. #42975 upgrades assorted GitHub Actions for Node 24 compatibility; #42976 specifically updates docker/build-push-action versions in Docker build workflows. Different actions, different code-path risk, so do not merge as duplicates." } ] }, @@ -16453,70 +16381,70 @@ "pull_request:43448|pull_request:45490", "pull_request:43451|pull_request:45490" ], - "prepared_review_unit_hash": "3c1c824e4d3481e33d4e3fd62665b51287688afd0e382d1c94e4e4dee835e4e4", + "prepared_review_unit_hash": "2770b640ce9fae6d15117d28f3fd7fae7902094fee9d82c798399bbdc54c133f", "component_budget": { "node_count": 4, "item_count": 4, "soft_pair_count": 6, - "serialized_chars": 14079, - "estimated_input_tokens": 3520, - "estimated_eval_tokens": 7296 + "serialized_chars": 14074, + "estimated_input_tokens": 3519, + "estimated_eval_tokens": 7294 }, "budget": { "node_count": 4, "item_count": 4, "soft_pair_count": 6, - "serialized_chars": 14079, - "estimated_input_tokens": 3520, - "estimated_eval_tokens": 7296 + "serialized_chars": 14074, + "estimated_input_tokens": 3519, + "estimated_eval_tokens": 7294 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "All four PRs are separate model-addition efforts. They share generic Transformers scaffolding (docs, auto mappings, tests), but each targets a different architecture and code path, so none should be deduplicated together.", - "confidence": 0.97, + "summary": "These are four separate model-addition PRs (ViT NEPA, Molmo, Molmo2, and CTSM) that only overlap at the pattern level: docs, auto-registration, and model scaffolding. They do not appear to fix the same bug or implement the same change, so no soft-duplicate edges should be accepted.", + "confidence": 0.98, "canonical_issue_reason": null, - "canonical_pr_reason": "No canonical PR: these are distinct model additions (ViT NEPA, Molmo, Molmo2, CTSM) with different implementations and filenames, not the same underlying change.", + "canonical_pr_reason": "PR 43451 (Molmo2) is the most substantial and best-documented representative of the cluster, with the broadest set of files and the highest review activity, but it is still a distinct model addition rather than a duplicate of the others.", "best_issue_reason": null, - "best_pr_reason": "No single PR is a better duplicate representative; all four are separate feature PRs rather than near-duplicates.", + "best_pr_reason": "PR 43451 is the strongest stand-alone PR in the set: it adds a full Molmo2 stack including model, processors, video processing, auto mappings, docs, and tests.", "soft_edge_verdicts": [ { "left": "pull_request:42978", "right": "pull_request:43448", "accept": false, - "reason": "Different model additions (ViT NEPA vs Molmo); shared docs/auto-registration files are generic and do not indicate the same underlying change." + "reason": "Different model families and code paths: ViT NEPA vs Molmo. Shared docs/auto files are just boilerplate for adding a new model." }, { "left": "pull_request:42978", "right": "pull_request:43451", "accept": false, - "reason": "Different model additions (ViT NEPA vs Molmo2) with different model, processor, and conversion files; not the same code-path fix." + "reason": "Different implementations for different models (ViT NEPA vs Molmo2); overlap is only in generic integration files like docs and auto registration." }, { "left": "pull_request:42978", "right": "pull_request:45490", "accept": false, - "reason": "ViT NEPA and CTSM are unrelated model integrations; overlapping docs/auto files are incidental scaffolding." + "reason": "CTSM is a separate time-series model addition, not the same bug/change as ViT NEPA." }, { "left": "pull_request:43448", "right": "pull_request:43451", "accept": false, - "reason": "Molmo and Molmo2 are separate model families with different implementation and processing stacks, so they are not duplicates." + "reason": "Molmo and Molmo2 are related by naming but are distinct model releases with different files, processors, and model code; not the same concrete change." }, { "left": "pull_request:43448", "right": "pull_request:45490", "accept": false, - "reason": "Molmo and CTSM target different models and codepaths; similarity is limited to shared registration/documentation changes." + "reason": "Molmo vs CTSM are unrelated model additions; similarity is limited to scaffolding and documentation." }, { "left": "pull_request:43451", "right": "pull_request:45490", "accept": false, - "reason": "Molmo2 and CTSM are distinct model additions with different modules and tests; not mergeable as one PR." + "reason": "Molmo2 and CTSM touch different model stacks and solve different integration tasks, so they are not duplicates." } ] }, @@ -16549,52 +16477,52 @@ "pull_request:43291|pull_request:44815", "pull_request:43665|pull_request:44815" ], - "prepared_review_unit_hash": "90ca72f5dd0a336e9de1b2ca1c6c61ce3fc8aaa1d4f3ff8fba117e8bc17b7e87", + "prepared_review_unit_hash": "92f252785ab5953aa569048892289f192753151ab51023b67ad90ee2add91997", "component_budget": { "node_count": 3, "item_count": 3, "soft_pair_count": 3, - "serialized_chars": 7091, - "estimated_input_tokens": 1773, - "estimated_eval_tokens": 3802 + "serialized_chars": 7093, + "estimated_input_tokens": 1774, + "estimated_eval_tokens": 3804 }, "budget": { "node_count": 3, "item_count": 3, "soft_pair_count": 3, - "serialized_chars": 7091, - "estimated_input_tokens": 1773, - "estimated_eval_tokens": 3802 + "serialized_chars": 7093, + "estimated_input_tokens": 1774, + "estimated_eval_tokens": 3804 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These PRs are not duplicates: one fixes Whisper tokenization behavior, one changes unrelated CLIP/ViT tests, and one fixes FP8/dequant loading logic. The only common signal is generic \"fix\" wording.", - "confidence": 0.98, + "summary": "All three PRs are unrelated: two are test-only changes for different models, and one is a substantial FP8/dequantization fix across loading/quantization code. None look like duplicates of the same underlying bug or change.", + "confidence": 0.95, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 44815 is the strongest standalone change: it contains substantive implementation work across loading and FP8 quantization code, plus a model-specific test, unlike the test-only or narrow Whisper PRs.", + "canonical_pr_reason": "pull_request:44815 is the only substantive code-fix PR; it targets FP8 dequant/padding/loading behavior across core loading and finegrained FP8 code, unlike the other two test-only PRs.", "best_issue_reason": null, - "best_pr_reason": "PR 44815 is the best representative of the cluster because it is the most complete, code-path-level fix rather than a test tweak or a narrowly scoped tokenizer change.", + "best_pr_reason": "pull_request:44815 is the best representative PR because it addresses a concrete implementation bug and has the broadest real code impact; the others are not comparable as change candidates.", "soft_edge_verdicts": [ { "left": "pull_request:43291", "right": "pull_request:43665", "accept": false, - "reason": "Different subsystems and failure modes: Whisper tokenizer decode logic versus CLIP/ViT test edits. No shared concrete bug or mergeable combined fix." + "reason": "Both are test-focused, but they touch different models (Whisper vs CLIP/ViT) and do not describe the same bug or code-path fix." }, { "left": "pull_request:43291", "right": "pull_request:44815", "accept": false, - "reason": "Whisper decoding/timestamp handling is unrelated to FP8 dequantization and model loading. These do not look like the same underlying bug." + "reason": "Whisper tokenization/test cleanup is unrelated to the FP8 dequant/loading fix in 44815; they do not share a concrete underlying issue." }, { "left": "pull_request:43665", "right": "pull_request:44815", "accept": false, - "reason": "The CLIP/ViT PR is just test assertions in model tests, while the other changes FP8 dequant/loading internals. Not the same code-path problem." + "reason": "43665 is just noisy test edits for CLIP/ViT, while 44815 is a real FP8 quantization/dequantization fix; no same-bug overlap." } ] }, @@ -16624,12 +16552,12 @@ "soft_pairs": [ "pull_request:43488|pull_request:43532" ], - "prepared_review_unit_hash": "a5df7642c7aad0719f1c3e6b458ae8d24fee99f39e45b27a1059cfa29b339c3a", + "prepared_review_unit_hash": "23b02b32c3859e2edf7cd896c98172c5aa0fbab223c1cc6d8ed403b12551559f", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4562, + "serialized_chars": 4563, "estimated_input_tokens": 1141, "estimated_eval_tokens": 2538 }, @@ -16637,7 +16565,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4562, + "serialized_chars": 4563, "estimated_input_tokens": 1141, "estimated_eval_tokens": 2538 }, @@ -16646,18 +16574,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The two pull requests are not duplicates: they only overlap on one test file, while their actual changes target different areas (one is a repo-bot/setup tweak, the other is workflow/testing/notification changes).", - "confidence": 0.91, + "summary": "The two PRs are not duplicates: #43488 is a small packaging/version-bump and test-bot check, while #43532 is a CI/workflow and testing-utils refactor. The only overlap is a shared test file, but the underlying changes are different.", + "confidence": 0.97, "canonical_issue_reason": null, - "canonical_pr_reason": "No true canonical PR here; if forced, 43532 is the more substantive and informative change set, but it is not the same fix as 43488.", + "canonical_pr_reason": "No single canonical PR: #43488 and #43532 do not address the same code-path problem, and their shared filename is incidental rather than evidence of duplication.", "best_issue_reason": null, - "best_pr_reason": "43532 is the better representative of the cluster because it has the broader, more detailed change set and clearer context; 43488 looks like a narrow bot-check/formatting PR.", + "best_pr_reason": "#43532 is the more substantial and representative PR in this cluster, but only as a cluster representative\u2014not because it duplicates #43488.", "soft_edge_verdicts": [ { "left": "pull_request:43488", "right": "pull_request:43532", "accept": false, - "reason": "Reject: shared evidence is only soft similarity and one common test file, but the diffs address different code paths and purposes. These would not plausibly be merged as one PR." + "reason": "Reject: the overlap is only a shared test file. #43488 changes setup/versioning and a dependency table; #43532 changes GitHub workflows, testing utilities, and related test behavior. They are not the same underlying bug or change." } ] }, @@ -16687,7 +16615,7 @@ "soft_pairs": [ "pull_request:43612|pull_request:43613" ], - "prepared_review_unit_hash": "318ba506395ee91d22e9d1811b6ecce4c886b92ff844ba9e7cfaa744a929e4cb", + "prepared_review_unit_hash": "44e9e754f2f01dbcc382f42f9207e83aeb8a0979a3dbf0e6e35d09c0cf0d0221", "component_budget": { "node_count": 2, "item_count": 2, @@ -16709,18 +16637,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Two closely related but distinct feature PRs: one adds a Promptable Concept Segmentation pipeline, the other adds a Promptable Visual Segmentation pipeline. They share some documentation and registration plumbing, but they target different pipelines/models and should not be merged as duplicates.", + "summary": "These are two related but distinct PRs: one adds a Promptable Concept Segmentation pipeline, the other adds a Promptable Visual Segmentation pipeline. They share documentation and pipeline plumbing, but they target different tasks and model integrations, so they are not duplicates.", "confidence": 0.96, "canonical_issue_reason": null, - "canonical_pr_reason": "No canonical PR for deduplication: these are separate feature additions with different task definitions and model support.", + "canonical_pr_reason": "No duplicate PR exists here. If a single representative is needed, PR 43613 is the broader integration, but it is still a different feature from PR 43612.", "best_issue_reason": null, - "best_pr_reason": "PR #43613 is the more comprehensive representative if one is needed, with broader model coverage and more integration/test surface, but it is not a duplicate of #43612.", + "best_pr_reason": "PR 43613 is the better standalone representative because it covers the broader visual segmentation stack and more related integrations/tests, while PR 43612 is a separate concept-segmentation feature.", "soft_edge_verdicts": [ { "left": "pull_request:43612", "right": "pull_request:43613", "accept": false, - "reason": "Different pipeline features and code paths: concept segmentation vs visual segmentation. Shared docs/registration files are not enough to treat them as the same change." + "reason": "Shared docs and pipeline scaffolding do not make these the same change. They implement different segmentation pipelines with different targets and supporting model/test files, so they should not be merged as duplicates." } ] }, @@ -16750,12 +16678,12 @@ "soft_pairs": [ "pull_request:43651|pull_request:43663" ], - "prepared_review_unit_hash": "5875e643e1d0f2220db75d6a57fa23f0798fe1c6997bb0c4131622650ae6f1f8", + "prepared_review_unit_hash": "379982c43366d9f992a3139f94113da4913e4e668bfc6c291405d06a324add00", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4292, + "serialized_chars": 4291, "estimated_input_tokens": 1073, "estimated_eval_tokens": 2402 }, @@ -16763,7 +16691,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4292, + "serialized_chars": 4291, "estimated_input_tokens": 1073, "estimated_eval_tokens": 2402 }, @@ -16772,18 +16700,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Two open PRs in trainer.py that both add customization hooks, but they address different behaviors: gradient-accumulation loss scaling vs. signature-column filtering. Not duplicates.", - "confidence": 0.97, + "summary": "Both PRs touch `src/transformers/trainer.py` and add customization hooks, but they address different trainer behaviors: gradient-accumulation loss scaling vs. signature-column filtering. They are not the same underlying change.", + "confidence": 0.95, "canonical_issue_reason": null, - "canonical_pr_reason": "No single canonical PR: 43651 changes loss scaling in training_step, while 43663 adds a separate override hook for column filtering; they solve different problems.", + "canonical_pr_reason": null, "best_issue_reason": null, - "best_pr_reason": "Neither PR is a global duplicate representative. If forced to pick a cleaner standalone change, 43651 is slightly more self-contained, but the pair should not be merged as duplicates.", + "best_pr_reason": "PR #43651 is the cleaner representative of the cluster because it makes a specific training-step behavior change with a clear code-path impact, while #43663 is a separate override hook for dataset column filtering.", "soft_edge_verdicts": [ { "left": "pull_request:43651", "right": "pull_request:43663", "accept": false, - "reason": "Different extension points and code paths in Trainer: one controls GA loss normalization, the other controls dataset column filtering/signature inspection. Same file only, not the same concrete bug/change." + "reason": "They only share the same file and general Trainer customization theme. One changes gradient accumulation loss scaling in `training_step`, the other extracts signature-column logic for feature filtering. These are different code paths and not plausibly a single merged fix/change." } ] }, @@ -16815,7 +16743,7 @@ "pull_request:43656|pull_request:43842", "pull_request:43836|pull_request:43842" ], - "prepared_review_unit_hash": "5e6961e986f2af7939a2f3d40703c0f29ed20fd147de96349f133eb79c10332f", + "prepared_review_unit_hash": "be7c77d81a19ce46f69580ae4a1a1e7b79143bfbc777eef0f74401e5f901aed7", "component_budget": { "node_count": 3, "item_count": 3, @@ -16837,33 +16765,33 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "PR #43842 is the clearest match for the cluster: it is a narrow, explicit fix for the TypeAdapter NameError in `src/transformers/cli/serve.py`. PR #43836 looks related and plausibly part of the same serve.py annotation-evaluation fix, but it is broader. PR #43656 is weaker evidence: it touches the same file, but the packet only supports it as a more general annotation/future-import cleanup tied to a different issue target, not clearly the same bugfix.", - "confidence": 0.7, + "summary": "Two PRs appear to target the same TypeAdapter import/annotation NameError in `src/transformers/cli/serve.py`; the third PR is a broader typing cleanup that is not the same concrete fix.", + "confidence": 0.73, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #43842 is the best canonical PR because it most directly and minimally fixes the concrete TypeAdapter NameError, with the clearest issue linkage and least surrounding refactor noise.", + "canonical_pr_reason": "PR #43842 is the most focused and explicit fix: it directly addresses the `TypeAdapter` NameError when `pydantic` is absent by moving the import under `TYPE_CHECKING` and quoting the annotation.", "best_issue_reason": null, - "best_pr_reason": "PR #43842 is the strongest overall representative: it is the narrowest, best-scoped fix for the specific runtime failure in `serve.py` and is easiest to use as the cluster anchor.", + "best_pr_reason": "PR #43842 is the best representative because it is narrowly scoped to the runtime NameError and is less speculative than the broader typing changes in the other PRs.", "soft_edge_verdicts": [ { "left": "pull_request:43656", "right": "pull_request:43842", "accept": false, - "reason": "Same file, but the evidence is too broad and the linked issue differs; this looks more like a related annotation cleanup than the same concrete TypeAdapter bugfix." + "reason": "Both touch `serve.py` and mention `TypeAdapter`, but #43656 looks like a broader annotation/type-hint cleanup, while #43842 specifically fixes the `TypeAdapter` import-time NameError when `pydantic` is not installed. Different concrete fixes, so not a duplicate pair." }, { "left": "pull_request:43836", "right": "pull_request:43842", "accept": true, - "reason": "Both PRs target the same `serve.py` TypeAdapter/annotation NameError path and share the same issue target, so they plausibly represent the same underlying fix despite different scope." + "reason": "Both PRs are working around the same `TypeAdapter` NameError in the CLI serve path, with overlapping changes to quoting/type-checking annotations in the same file. They look like two versions of the same underlying fix and could plausibly be merged into one PR." } ] }, "evaluator_result": { - "accept": false, - "feedback": "The summary is a bit too strong for the packet. PR #43842 is well supported as the narrow TypeAdapter fix, and PR #43836 is plausibly related because it shares the same file and issue target. But PR #43656 is not grounded as clearly: the packet only shows broader annotation/future-import cleanup and a different linked issue, so calling it the same TypeAdapter bugfix and accepting it as mergeable with #43842 overstates the evidence. A more conservative summary would distinguish #43842 as the clearest match and describe the others as related annotation-evaluation fixes rather than the same bug with confidence." + "accept": true, + "feedback": "The summary is grounded in the packet: #43842 is the most focused fix for the `TypeAdapter` NameError, #43656 is broader annotation cleanup in the same file, and #43836 shares the same issue target and file with #43842. The soft-edge verdicts are conservative overall: rejecting #43656 vs #43842 is justified by the broader/different-looking change, and accepting #43836 vs #43842 is reasonable because they appear to be two implementations of the same underlying `TypeAdapter` bugfix in `src/transformers/cli/serve.py`." }, "evaluator_used": true, - "retried": true, + "retried": false, "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null @@ -17050,148 +16978,148 @@ "pull_request:44044|pull_request:44154", "pull_request:44024|pull_request:44076" ], - "prepared_review_unit_hash": "c7d106248907fefa412ed608d45a28f50f7db65198db25558b712b8d7b5ca374", + "prepared_review_unit_hash": "f2efc4714add0223714f7bdae4365e9624d6f01c309f19753206e571ae5cc251", "component_budget": { "node_count": 28, "item_count": 28, "soft_pair_count": 77, - "serialized_chars": 84502, - "estimated_input_tokens": 21126, - "estimated_eval_tokens": 42508 + "serialized_chars": 84508, + "estimated_input_tokens": 21127, + "estimated_eval_tokens": 42510 }, "budget": { "node_count": 28, "item_count": 28, "soft_pair_count": 72, - "serialized_chars": 76935, - "estimated_input_tokens": 19234, - "estimated_eval_tokens": 38724 + "serialized_chars": 76940, + "estimated_input_tokens": 19235, + "estimated_eval_tokens": 38726 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a set of model-specific output-tracing refactors, not true duplicates across different architectures. The only clear duplicate-like overlaps are within the GPT-Neo and GPT-J/CodeGen refactor branches.", - "confidence": 0.93, + "summary": "This cluster is mostly a wave of model-specific PRs migrating output tracing to `capture_outputs`/`can_return_tuple`. Only the same-model GPT-Neo and GPT-J pairs look like true duplicates; most other pairs are parallel refactors across different architectures and should stay separate.", + "confidence": 0.82, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 44722 is the strongest representative of the overlapping GPT-J/CodeGen tracing refactors: it targets the same concrete output-capture code path as the earlier GPT-J PRs and appears to be the most complete/latest variant.", + "canonical_pr_reason": "PR 44007 is the broadest, clearest representative of the initiative: it has an explicit issue link, touches multiple related vision backbones, and shows the standardized output-tracing refactor in a concrete, reusable form.", "best_issue_reason": null, - "best_pr_reason": "PR 44722 is the best single representative for this cluster because it covers the same concrete GPT-J/CodeGen output-tracing change as the closest overlapping PRs and has the clearest scope among the accepted duplicates.", + "best_pr_reason": "PR 44007 is the best single anchor for the cluster because it is issue-linked, relatively broad, and most representative of the model-by-model output-tracing migration pattern.", "soft_edge_verdicts": [ { "left": "pull_request:43996", "right": "pull_request:44085", "accept": false, - "reason": "Different model families; only a broad standardized-output theme overlaps." + "reason": "Different model families and code paths; same umbrella refactor theme only." }, { "left": "pull_request:43996", "right": "pull_request:44044", "accept": false, - "reason": "Different model families and code paths; same refactor theme only." + "reason": "Different architectures (FNet/CVT vs DeBERTa); not the same concrete change." }, { "left": "pull_request:44066", "right": "pull_request:44085", "accept": true, - "reason": "Same GPT-J/CodeGen output-tracing refactor family with overlapping model code paths; could be merged into one PR." + "reason": "Same GPT-J output-tracing refactor in the same model file; these look like overlapping versions of the same change." }, { "left": "pull_request:44007", "right": "pull_request:44072", "accept": false, - "reason": "RegNet vs EfficientNet are separate model implementations; no shared concrete bug/change." + "reason": "Both are output-tracing migrations, but for different vision models and code paths." }, { "left": "pull_request:44072", "right": "pull_request:44722", "accept": false, - "reason": "Different models and forward paths; only a generic decorator migration overlaps." + "reason": "EfficientNet and GPT-J/CodeGen are unrelated implementations; same initiative is not enough." }, { "left": "pull_request:44066", "right": "pull_request:44072", "accept": false, - "reason": "GPT-J/CodeGen output tracing is unrelated to EfficientNet output tracing." + "reason": "GPT-J/CodeGen vs EfficientNet are separate model refactors, not one duplicate bug fix." }, { "left": "pull_request:44013", "right": "pull_request:44044", "accept": false, - "reason": "MobileNetV2 and DeBERTa touch unrelated code paths." + "reason": "MobileNetV2 and DeBERTa are different model stacks with unrelated output-tracing code." }, { "left": "pull_request:44066", "right": "pull_request:44086", "accept": false, - "reason": "Different model families; no shared concrete fix." + "reason": "GPT-J/CodeGen and MGP-STR are different code paths; same pattern only." }, { "left": "pull_request:44066", "right": "pull_request:44071", "accept": false, - "reason": "GPT-J/CodeGen refactor vs MPT refactor; same pattern, different code path." + "reason": "GPT-J/CodeGen vs MPT are distinct model implementations, so not duplicates." }, { "left": "pull_request:44018", "right": "pull_request:44068", "accept": true, - "reason": "Both are GPT-Neo output-tracing refactors on the same model/code path and overlap enough to be merged together." + "reason": "Same GPT-Neo file and same decorator-based output-tracing migration; one is effectively a rewrite of the other." }, { "left": "pull_request:44066", "right": "pull_request:44068", "accept": false, - "reason": "GPT-J/CodeGen vs GPT-Neo; different models and code paths." + "reason": "GPT-J/CodeGen vs GPT-Neo are different model families and should not be merged." }, { "left": "pull_request:44018", "right": "pull_request:44066", "accept": false, - "reason": "GPT-Neo and GPT-J/CodeGen are distinct model implementations." + "reason": "Both are output-tracing refactors, but for different models (GPT-Neo vs GPT-J/CodeGen)." }, { "left": "pull_request:44019", "right": "pull_request:44722", "accept": false, - "reason": "ResNet and GPT-J/CodeGen are unrelated model families." + "reason": "ResNet and GPT-J/CodeGen are unrelated code paths." }, { "left": "pull_request:44086", "right": "pull_request:44722", "accept": false, - "reason": "MGP-STR and GPT-J/CodeGen are different code paths." + "reason": "MGP-STR and GPT-J/CodeGen are different model implementations." }, { "left": "pull_request:44019", "right": "pull_request:44085", "accept": false, - "reason": "ResNet output tracing is unrelated to GPT-J." + "reason": "ResNet and GPT-J are different model families; not a duplicate change." }, { "left": "pull_request:44018", "right": "pull_request:44019", "accept": false, - "reason": "GPT-Neo vs ResNet; same broad refactor theme only." + "reason": "GPT-Neo and ResNet are unrelated; only the refactor style overlaps." }, { "left": "pull_request:44019", "right": "pull_request:44071", "accept": false, - "reason": "ResNet and MPT are separate models with different forward logic." + "reason": "ResNet and MPT are different code paths with no shared bug/change." }, { "left": "pull_request:44068", "right": "pull_request:44116", "accept": false, - "reason": "GPT-Neo vs Flaubert/XLM are unrelated code paths." + "reason": "GPT-Neo and Flaubert are separate model refactors, not one mergeable PR." }, { "left": "pull_request:44068", "right": "pull_request:44722", "accept": false, - "reason": "GPT-Neo and GPT-J/CodeGen are different model families." + "reason": "GPT-Neo vs GPT-J/CodeGen are different model families despite similar output-tracing mechanics." }, { "left": "pull_request:44071", @@ -17203,19 +17131,19 @@ "left": "pull_request:44018", "right": "pull_request:44116", "accept": false, - "reason": "GPT-Neo and Flaubert/XLM do not share the same concrete bug or change." + "reason": "GPT-Neo and Flaubert are different models; no same concrete code-path problem." }, { "left": "pull_request:44018", "right": "pull_request:44086", "accept": false, - "reason": "GPT-Neo and MGP-STR are unrelated model code paths." + "reason": "GPT-Neo and MGP-STR are different model refactors." }, { "left": "pull_request:44018", "right": "pull_request:44085", "accept": false, - "reason": "GPT-Neo and GPT-J are different models." + "reason": "GPT-Neo and GPT-J are distinct model families." }, { "left": "pull_request:44068", @@ -17223,173 +17151,263 @@ "accept": false, "reason": "GPT-Neo and MGP-STR are unrelated implementations." }, - { - "left": "pull_request:44071", - "right": "pull_request:44086", - "accept": false, - "reason": "MPT and MGP-STR have no shared concrete fix." - }, - { - "left": "pull_request:44059", - "right": "pull_request:44086", - "accept": false, - "reason": "GPT-2 and MGP-STR are unrelated model families." - }, - { - "left": "pull_request:44085", - "right": "pull_request:44086", - "accept": false, - "reason": "GPT-J and MGP-STR do not touch the same code path." - }, { "left": "pull_request:44098", "right": "pull_request:44101", "accept": false, - "reason": "ViLT and Flaubert/XLM are separate model stacks." + "reason": "ViLT and Flaubert are different architectures; same output-handling theme only." }, { "left": "pull_request:44059", "right": "pull_request:44722", "accept": false, - "reason": "GPT-2 and GPT-J/CodeGen are different concrete refactors." + "reason": "GPT-2 and GPT-J/CodeGen are separate refactors with different code paths." }, { "left": "pull_request:44018", "right": "pull_request:44071", "accept": false, - "reason": "GPT-Neo and MPT are different model implementations." + "reason": "GPT-Neo and MPT are unrelated model families." }, { "left": "pull_request:44017", "right": "pull_request:44722", "accept": false, - "reason": "SegFormer and GPT-J/CodeGen are unrelated code paths." + "reason": "Segformer and GPT-J/CodeGen are different models; not one duplicate change." }, { "left": "pull_request:44017", "right": "pull_request:44161", "accept": false, - "reason": "SegFormer and LongT5 do not share the same bug/change." + "reason": "Segformer and LongT5 are separate output-tracing migrations." }, { "left": "pull_request:44017", "right": "pull_request:44073", "accept": false, - "reason": "SegFormer and VisualBERT are unrelated model families." + "reason": "Segformer and VisualBert are different model code paths." }, { "left": "pull_request:44068", "right": "pull_request:44071", "accept": false, - "reason": "GPT-Neo and MPT are different models with different output logic." + "reason": "GPT-Neo and MPT are unrelated despite similar decorator usage." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44073", + "accept": false, + "reason": "RegNet and VisualBert are different model implementations." + }, + { + "left": "pull_request:44017", + "right": "pull_request:44018", + "accept": false, + "reason": "Segformer and GPT-Neo are different architectures." + }, + { + "left": "pull_request:44073", + "right": "pull_request:44101", + "accept": false, + "reason": "VisualBert and Flaubert are different model stacks." + }, + { + "left": "pull_request:44059", + "right": "pull_request:44071", + "accept": false, + "reason": "GPT-2 and MPT are unrelated code paths." + }, + { + "left": "pull_request:44059", + "right": "pull_request:44068", + "accept": false, + "reason": "GPT-2 and GPT-Neo are different model families; same refactor style is not enough." + }, + { + "left": "pull_request:44056", + "right": "pull_request:44101", + "accept": false, + "reason": "MPNet and Flaubert are different models with separate output-tracing code." + }, + { + "left": "pull_request:44073", + "right": "pull_request:44161", + "accept": false, + "reason": "VisualBert and LongT5 are unrelated model refactors." + }, + { + "left": "pull_request:44101", + "right": "pull_request:44161", + "accept": false, + "reason": "Flaubert and LongT5 are different model families." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44017", + "accept": false, + "reason": "RegNet and Segformer are different vision backbones; not the same change." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44056", + "accept": false, + "reason": "RegNet and MPNet are different model code paths." }, { "left": "pull_request:44071", "right": "pull_request:44076", "accept": false, - "reason": "MPT and ImageGPT are unrelated code paths." + "reason": "MPT and ImageGPT are separate model implementations." + }, + { + "left": "pull_request:44074", + "right": "pull_request:44101", + "accept": false, + "reason": "TextNet and Flaubert are unrelated refactors." + }, + { + "left": "pull_request:44056", + "right": "pull_request:44161", + "accept": false, + "reason": "MPNet and LongT5 are different models; same initiative only." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44018", + "accept": false, + "reason": "RegNet and GPT-Neo are unrelated code paths." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44161", + "accept": false, + "reason": "RegNet and LongT5 are different model families." + }, + { + "left": "pull_request:44074", + "right": "pull_request:44161", + "accept": false, + "reason": "TextNet and LongT5 are different model implementations." }, { "left": "pull_request:43997", + "right": "pull_request:44071", + "accept": false, + "reason": "RegNet and MPT are unrelated models." + }, + { + "left": "pull_request:44010", + "right": "pull_request:44071", + "accept": false, + "reason": "SqueezeBert and MPT are different architectures." + }, + { + "left": "pull_request:44010", "right": "pull_request:44073", "accept": false, - "reason": "RegNet and VisualBERT are separate model implementations." + "reason": "SqueezeBert and VisualBert are separate model refactors." }, { - "left": "pull_request:44017", - "right": "pull_request:44018", + "left": "pull_request:44018", + "right": "pull_request:44129", "accept": false, - "reason": "SegFormer and GPT-Neo are unrelated models." + "reason": "GPT-Neo and SpeechT5 are different code paths." }, { - "left": "pull_request:44017", - "right": "pull_request:44056", + "left": "pull_request:43997", + "right": "pull_request:44074", + "accept": false, + "reason": "RegNet and TextNet are unrelated model families." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44129", "accept": false, - "reason": "SegFormer and MPNet are different architectures." + "reason": "RegNet and SpeechT5 are different implementations." }, { "left": "pull_request:44071", "right": "pull_request:44129", "accept": false, - "reason": "MPT and SpeechT5 are unrelated model families." + "reason": "MPT and SpeechT5 are unrelated code paths." }, { "left": "pull_request:43996", "right": "pull_request:44161", "accept": false, - "reason": "CVT/FNet and LongT5 do not share the same concrete refactor." + "reason": "FNet/CVT and LongT5 are different model families." }, { "left": "pull_request:43996", "right": "pull_request:44017", "accept": false, - "reason": "CVT/FNet and SegFormer are separate model code paths." + "reason": "FNet/CVT and Segformer are separate vision model refactors." }, { "left": "pull_request:44010", "right": "pull_request:44044", "accept": false, - "reason": "SqueezeBERT and DeBERTa are unrelated implementations." + "reason": "SqueezeBert and DeBERTa are different model code paths." }, { "left": "pull_request:44044", "right": "pull_request:44074", "accept": false, - "reason": "DeBERTa and TextNet are different model families." + "reason": "DeBERTa and TextNet are unrelated refactors." }, { "left": "pull_request:44066", "right": "pull_request:44722", "accept": true, - "reason": "Same GPT-J/CodeGen output-tracing change set; strong overlap on the same concrete code path." + "reason": "Same GPT-J output-tracing refactor in the same file; 44722 looks like the broader follow-on version." }, { "left": "pull_request:44085", "right": "pull_request:44722", "accept": true, - "reason": "Both touch GPT-J output tracing on the same underlying code path and are close enough to merge together." + "reason": "Both target GPT-J output tracing in the same file, so they are likely overlapping variants of the same change." }, { "left": "pull_request:44085", "right": "pull_request:44129", "accept": false, - "reason": "GPT-J and SpeechT5 are unrelated model implementations." + "reason": "GPT-J and SpeechT5 are different model families." }, { "left": "pull_request:43996", "right": "pull_request:44007", "accept": false, - "reason": "CVT/FNet vs RegNet/ResNet/RT-DETR are unrelated model paths." + "reason": "FNet/CVT and RegNet/ResNet/RT-DETR are different model refactors." }, { "left": "pull_request:43996", "right": "pull_request:44129", "accept": false, - "reason": "CVT/FNet and SpeechT5 are different models." + "reason": "FNet/CVT and SpeechT5 are unrelated implementations." }, { "left": "pull_request:44044", "right": "pull_request:44129", "accept": false, - "reason": "DeBERTa and SpeechT5 are unrelated code paths." + "reason": "DeBERTa and SpeechT5 are different model code paths." }, { "left": "pull_request:44044", "right": "pull_request:44154", "accept": false, - "reason": "DeBERTa and VITS are different model families." + "reason": "DeBERTa and VITS are separate refactors." }, { "left": "pull_request:44024", "right": "pull_request:44076", "accept": false, - "reason": "FocalNet and ImageGPT are unrelated implementations." + "reason": "FocalNet and ImageGPT are different model families; same output-tracing style only." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is grounded in the packet: it correctly treats this as mostly model-specific output-tracing refactors and identifies the GPT-Neo and GPT-J/CodeGen branches as the only real overlap clusters. The accepted soft-edge pairs are also conservative enough and look mergeable into single bugfix PRs for the same model family/code path." + "feedback": "The summary is grounded in the packet: it correctly characterizes this as a broad output-tracing migration wave and avoids claiming cross-architecture pairs are duplicates. The two accepted soft edges are also conservative because they involve the same model/file family and look like overlapping implementations of the same refactor, while the rejected pairs are mostly different architectures or unrelated code paths." }, "evaluator_used": true, "retried": false, @@ -17424,20 +17442,20 @@ "pull_request:43997|pull_request:44010", "pull_request:44010|pull_request:44129" ], - "prepared_review_unit_hash": "34900a8774c8bba9e9a5ceff83a6371d8168fb0816eb1b8224a8cf8f13331497", + "prepared_review_unit_hash": "e4e07bf558766a13c786a35a4daed48b3e2eb9b0576517dc5f570736d88125e0", "component_budget": { "node_count": 28, "item_count": 28, "soft_pair_count": 77, - "serialized_chars": 84502, - "estimated_input_tokens": 21126, - "estimated_eval_tokens": 42508 + "serialized_chars": 84508, + "estimated_input_tokens": 21127, + "estimated_eval_tokens": 42510 }, "budget": { "node_count": 6, "item_count": 6, "soft_pair_count": 5, - "serialized_chars": 13067, + "serialized_chars": 13068, "estimated_input_tokens": 3267, "estimated_eval_tokens": 6790 }, @@ -17446,42 +17464,42 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These PRs are all model-specific refactors toward standardized output tracing/capture_outputs, but they are not duplicates of one another because each targets a different model implementation and code path.", - "confidence": 0.92, + "summary": "These PRs are all model-specific refactors toward standardized output tracing/capture, but they target different architectures and code paths. They are similar in style, not duplicates of the same fix, so the soft edges should be rejected.", + "confidence": 0.95, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 44129 is the most representative single item: it has an explicit issue target, a clear refactor title, and a focused but substantial SpeechT5 output-tracing migration.", + "canonical_pr_reason": "PR 44073 is the cleanest representative of the shared refactor pattern: it introduces the standardized tracing helpers in a single model file with a focused change set.", "best_issue_reason": null, - "best_pr_reason": "PR 44129 is the strongest standalone exemplar for the cluster because it is tightly scoped, traceable to an issue, and clearly reflects the standardized output-capture change.", + "best_pr_reason": "PR 44073 is the best cluster representative because it is self-contained, clearly about the standardized output-tracing migration, and is not entangled with unrelated model changes.", "soft_edge_verdicts": [ { "left": "pull_request:44073", "right": "pull_request:44074", "accept": false, - "reason": "Both are output-tracing refactors, but they touch different models and different code paths; this is a shared pattern, not the same underlying change." + "reason": "Both are output-tracing migrations, but for different models (VisualBert vs TextNet). They touch different concrete code paths and are not mergeable as one fix." }, { "left": "pull_request:44056", "right": "pull_request:44073", "accept": false, - "reason": "MPNet and VisualBert are separate model implementations; the PRs are analogous refactors, not a single duplicate fix." + "reason": "Same general refactor theme, but MPNet and VisualBert are separate implementations with different tracing hooks and output paths." }, { "left": "pull_request:44056", "right": "pull_request:44074", "accept": false, - "reason": "Different model files and model-specific tracing logic; no evidence they fix the same concrete bug or are mergeable as one PR." + "reason": "Both migrate output capture, but they are model-specific refactors in different files with different layer types and behaviors." }, { "left": "pull_request:43997", "right": "pull_request:44010", "accept": false, - "reason": "RegNet and SqueezeBert are unrelated model implementations; they share a refactor theme but not the same underlying code-path problem." + "reason": "RegNet and SqueezeBert are unrelated model implementations; this is only a shared refactoring pattern, not the same underlying bug or change." }, { "left": "pull_request:44010", "right": "pull_request:44129", "accept": false, - "reason": "These are separate model migrations to the same output-capture style, not the same concrete fix and not something that would plausibly be merged into one PR." + "reason": "SqueezeBert and SpeechT5 both adjust output tracing, but they address different model code paths and SpeechT5 has its own explicit issue context." } ] }, @@ -17561,268 +17579,274 @@ "pull_request:44028|pull_request:44030", "pull_request:44029|pull_request:44030" ], - "prepared_review_unit_hash": "18df8742397e7118a6db5a5063e29e23080ad826207e3e80fa0150d2f749ef99", + "prepared_review_unit_hash": "e37b764106e27d2d7955ac1b6ff583f46ea3a8b926ae1e28b5fbe76e06bd880b", "component_budget": { "node_count": 13, "item_count": 13, "soft_pair_count": 40, - "serialized_chars": 40178, - "estimated_input_tokens": 10045, - "estimated_eval_tokens": 20346 + "serialized_chars": 40164, + "estimated_input_tokens": 10041, + "estimated_eval_tokens": 20338 }, "budget": { "node_count": 13, "item_count": 13, "soft_pair_count": 40, - "serialized_chars": 40178, - "estimated_input_tokens": 10045, - "estimated_eval_tokens": 20346 + "serialized_chars": 40164, + "estimated_input_tokens": 10041, + "estimated_eval_tokens": 20338 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These PRs share a common refactor theme around output tracing, but they target different model implementations and code paths, so they are not duplicates of one another.", - "confidence": 0.95, + "summary": "This cluster is mostly a set of similarly named, model-specific output-tracing refactors. They share implementation patterns, but each PR targets a different architecture and code path, so they are not duplicates of the same underlying change.", + "confidence": 0.93, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 44003 is the broadest representative of the theme, touching both mamba and falcon_mamba and adding the output-capturing plumbing.", + "canonical_pr_reason": "PR 44025 is the strongest representative: it spans two closely related files in the same family and clearly implements the newer tuple-capture/output-tracing pattern, making it the most central example in this loose cluster.", "best_issue_reason": null, - "best_pr_reason": "PR 44003 is the best single representative because it is the most substantial and central of the output-tracing refactors in this cluster.", + "best_pr_reason": "44025 is the best single PR to represent the cluster, but only as a theme anchor; it is not a safe duplicate target for the others because the actual model code paths differ.", "soft_edge_verdicts": [ { "left": "pull_request:43998", "right": "pull_request:44002", "accept": false, - "reason": "Different model families and file paths (`timm_backbone` vs `upernet`); same refactor pattern but not the same change or bug." + "reason": "Same refactor theme, but different model families and forward paths." }, { "left": "pull_request:43998", "right": "pull_request:44000", "accept": false, - "reason": "`timm_backbone` and `vision_text_dual_encoder` are unrelated code paths; this is only a shared refactor theme." + "reason": "Different architectures; the similarity is only in the output-tracing refactor pattern." }, { "left": "pull_request:43998", "right": "pull_request:44001", "accept": false, - "reason": "These are separate model implementations (`timm_backbone` vs `univnet`), so they are not one underlying fix." + "reason": "Separate model code paths, not the same concrete bug or change." }, { "left": "pull_request:43998", "right": "pull_request:44003", "accept": false, - "reason": "`timm_backbone` and `mamba/falcon_mamba` are distinct model stacks; no shared concrete bug or mergeable single PR." + "reason": "Different model implementations; too broad to treat as one mergeable PR." }, { "left": "pull_request:43998", "right": "pull_request:44029", "accept": false, - "reason": "`timm_backbone` and `rwkv` are different architectures and code paths; same general output-tracing refactor only." + "reason": "Unrelated model families despite similar tracing cleanup." }, { "left": "pull_request:43999", "right": "pull_request:44001", "accept": false, - "reason": "Different model files (`mobilenet_v1` vs `univnet`) and different tracing changes; not the same fix." + "reason": "Both are output-tracing refactors, but for different models and call stacks." }, { "left": "pull_request:43999", "right": "pull_request:44002", "accept": false, - "reason": "`mobilenet_v1` and `upernet` are separate implementations; similar refactor theme is not enough to duplicate them." + "reason": "Different architecture-specific implementations; not the same change." }, { "left": "pull_request:43999", "right": "pull_request:44025", "accept": false, - "reason": "These touch different depth models (`mobilenet_v1` vs `depth_anything/prompt_depth_anything`), so they are not one code-path issue." + "reason": "Different model families and files; only the refactor style matches." }, { "left": "pull_request:43999", "right": "pull_request:44026", "accept": false, - "reason": "`mobilenet_v1` and `vision_encoder_decoder` are unrelated models; this is only a shared tracing refactor." + "reason": "Separate encoder/decoder code paths; similarity is superficial." }, { "left": "pull_request:43999", "right": "pull_request:44027", "accept": false, - "reason": "Different model families (`mobilenet_v1` vs `speech_encoder_decoder`) and no shared concrete bug." + "reason": "Different model-specific tracing refactors, not one concrete fix." }, { "left": "pull_request:44000", "right": "pull_request:44001", "accept": false, - "reason": "`vision_text_dual_encoder` and `univnet` are distinct model paths; same refactor label does not make them duplicates." + "reason": "Same broad refactor theme, but distinct model code paths." }, { "left": "pull_request:44000", "right": "pull_request:44002", "accept": false, - "reason": "Separate implementations (`vision_text_dual_encoder` vs `upernet`) with no common underlying issue." + "reason": "Different models and outputs; not a duplicate change." }, { "left": "pull_request:44000", "right": "pull_request:44003", "accept": false, - "reason": "Different subsystems (`vision_text_dual_encoder` vs `mamba/falcon_mamba`); not plausibly one merged PR for a single bug." + "reason": "Too broad; these are separate model implementations." }, { "left": "pull_request:44000", "right": "pull_request:44025", "accept": false, - "reason": "These affect different model stacks (`vision_text_dual_encoder` vs `depth_anything`); only the refactor theme matches." + "reason": "Different model families; no shared concrete bug." }, { "left": "pull_request:44001", "right": "pull_request:44002", "accept": false, - "reason": "`univnet` and `upernet` are unrelated models, so they are not the same underlying change." + "reason": "Same tracing-refactor pattern, but unrelated model code." }, { "left": "pull_request:44001", "right": "pull_request:44003", "accept": false, - "reason": "`univnet` and `mamba/falcon_mamba` are different code paths with no shared bug fix." + "reason": "Different forward paths and modules; not mergeable as one PR." }, { "left": "pull_request:44001", "right": "pull_request:44025", "accept": false, - "reason": "These target different model families (`univnet` vs `depth_anything`), so they should not be merged as duplicates." + "reason": "Different architectures; only the refactor motif is shared." }, { "left": "pull_request:44002", "right": "pull_request:44029", "accept": false, - "reason": "`upernet` and `rwkv` are different models; similar output-tracing cleanup only." + "reason": "Different subsystems and model code paths." }, { "left": "pull_request:44002", "right": "pull_request:44003", "accept": false, - "reason": "`upernet` and `mamba/falcon_mamba` are separate code paths and not one concrete fix." + "reason": "Separate model families; same theme is not enough." }, { "left": "pull_request:44003", "right": "pull_request:44028", "accept": false, - "reason": "`mamba/falcon_mamba` and `superpoint` are unrelated implementations; same pattern, different changes." + "reason": "Different architectures and output formats; not the same bug/change." }, { "left": "pull_request:44003", "right": "pull_request:44004", "accept": false, - "reason": "`mamba` and `codegen` touch different model internals and cannot be treated as the same PR." + "reason": "Different concrete code paths; one is Mamba-related, the other CodeGen." }, { "left": "pull_request:43998", "right": "pull_request:44004", "accept": false, - "reason": "`timm_backbone` and `codegen` are different model paths with different output semantics; not a duplicate." + "reason": "Unrelated models; only broad refactoring similarity." }, { "left": "pull_request:44001", "right": "pull_request:44004", "accept": false, - "reason": "`univnet` and `codegen` are separate models; same refactor theme is insufficient." + "reason": "Different model code paths and behavior." }, { "left": "pull_request:44002", "right": "pull_request:44004", "accept": false, - "reason": "`upernet` and `codegen` are unrelated code paths, so these are not the same change." + "reason": "Different implementations; no shared underlying defect." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44004", + "accept": false, + "reason": "Both touch tracing/output handling, but in different models." }, { "left": "pull_request:44025", "right": "pull_request:44026", "accept": false, - "reason": "`depth_anything` and `vision_encoder_decoder` are different model families; no single underlying bug." + "reason": "Different model families; not the same concrete fix." }, { "left": "pull_request:44025", "right": "pull_request:44027", "accept": false, - "reason": "`depth_anything` and `speech_encoder_decoder` are distinct implementations, not one mergeable fix." + "reason": "Separate encoder-decoder and depth-model changes; too broad." }, { "left": "pull_request:44025", "right": "pull_request:44028", "accept": false, - "reason": "`depth_anything` and `superpoint` are separate model paths; only the refactor style is shared." + "reason": "Different architectures and output contracts." }, { "left": "pull_request:44025", "right": "pull_request:44029", "accept": false, - "reason": "`depth_anything` and `rwkv` are unrelated architectures, so this is not the same bug/change." + "reason": "Different model families and code paths." }, { "left": "pull_request:44025", "right": "pull_request:44030", "accept": false, - "reason": "`depth_anything` and `dpr` touch different code paths; similar output-tracing cleanup only." + "reason": "Different models; similar refactor wording only." }, { "left": "pull_request:44026", "right": "pull_request:44029", "accept": false, - "reason": "`vision_encoder_decoder` and `rwkv` are different model stacks and cannot be merged as one fix." + "reason": "Different model implementations and tracing logic." }, { "left": "pull_request:44026", "right": "pull_request:44027", "accept": false, - "reason": "`vision_encoder_decoder` and `speech_encoder_decoder` are related only by broad wrapper shape, not the same concrete code path." + "reason": "Separate model families; no single underlying bug." }, { "left": "pull_request:44026", "right": "pull_request:44028", "accept": false, - "reason": "`vision_encoder_decoder` and `superpoint` are different implementations with no shared underlying issue." + "reason": "Different architectures and output handling." }, { "left": "pull_request:44026", "right": "pull_request:44030", "accept": false, - "reason": "`vision_encoder_decoder` and `dpr` are separate model code paths; same theme, not same change." + "reason": "Different model code paths; not mergeable as one PR." }, { "left": "pull_request:44027", "right": "pull_request:44028", "accept": false, - "reason": "`speech_encoder_decoder` and `superpoint` are unrelated models and should not be deduped." + "reason": "Different model-specific refactors, not duplicates." }, { "left": "pull_request:44027", "right": "pull_request:44029", "accept": false, - "reason": "`speech_encoder_decoder` and `rwkv` affect different model architectures, so they are not one fix." + "reason": "Separate architectures and output contracts." }, { "left": "pull_request:44027", "right": "pull_request:44030", "accept": false, - "reason": "`speech_encoder_decoder` and `dpr` are different code paths; only the output-tracing motif matches." + "reason": "Different code paths; only the tracing refactor theme matches." }, { "left": "pull_request:44028", "right": "pull_request:44029", "accept": false, - "reason": "`superpoint` and `rwkv` are separate implementations with no shared concrete bug." + "reason": "Different model families and behavior." }, { "left": "pull_request:44028", "right": "pull_request:44030", "accept": false, - "reason": "`superpoint` and `dpr` are different model paths, so not the same underlying change." + "reason": "Separate models; no shared concrete bug or change." }, { "left": "pull_request:44029", "right": "pull_request:44030", "accept": false, - "reason": "`rwkv` and `dpr` are distinct model stacks; same refactor theme only, not a duplicate." + "reason": "Different model implementations; too broad for a duplicate." } ] }, @@ -17855,52 +17879,52 @@ "pull_request:44662|pull_request:45075", "pull_request:45075|pull_request:45401" ], - "prepared_review_unit_hash": "26a9d70375198efad7f3cce40a74f1b5ec4105e35ec8249f0cdc638de1a9d877", + "prepared_review_unit_hash": "2ba87f1db6b33bae991de182f8e5ac8815bbf8e9e34da8034b4d5b349e639d43", "component_budget": { "node_count": 3, "item_count": 3, "soft_pair_count": 3, - "serialized_chars": 10914, - "estimated_input_tokens": 2729, - "estimated_eval_tokens": 5714 + "serialized_chars": 10911, + "estimated_input_tokens": 2728, + "estimated_eval_tokens": 5712 }, "budget": { "node_count": 3, "item_count": 3, "soft_pair_count": 3, - "serialized_chars": 10914, - "estimated_input_tokens": 2729, - "estimated_eval_tokens": 5714 + "serialized_chars": 10911, + "estimated_input_tokens": 2728, + "estimated_eval_tokens": 5712 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are three separate model-integration PRs, not duplicates: PenguinVL, DeepSeek-OCR-2, and Voxtral TTS each add distinct model code, tests, and docs with only shared auto-registry boilerplate.", - "confidence": 0.98, + "summary": "All three pull requests add different model integrations (PenguinVL, DeepSeek-OCR-2, Voxtral TTS). They share boilerplate files because model additions in Transformers follow the same template, but they are not the same underlying change and should not be merged as duplicates.", + "confidence": 0.97, "canonical_issue_reason": null, - "canonical_pr_reason": "No canonical PR: the cluster does not represent one underlying fix/change. Each PR introduces a different model family and concrete codepath, so none should be collapsed into a single canonical PR.", + "canonical_pr_reason": "No single canonical PR: these are separate model-addition PRs for different architectures and codepaths, not duplicate fixes or the same feature work.", "best_issue_reason": null, - "best_pr_reason": "PR #45401 is the strongest standalone representative only because it has an explicit issue link and a clear scoped title, but it is still not duplicate-equivalent to the others.", + "best_pr_reason": "PR 45075 is the most mature/active implementation PR, but it only represents DeepSeek-OCR-2 and does not subsume the other model additions.", "soft_edge_verdicts": [ { "left": "pull_request:44662", "right": "pull_request:45401", "accept": false, - "reason": "Different model additions (PenguinVL vs Voxtral TTS). Shared registry/docs boilerplate is not enough to indicate the same underlying change or mergeable single PR." + "reason": "Different models and different implementation targets (PenguinVL vs Voxtral TTS); shared framework files are expected boilerplate, not evidence of the same change." }, { "left": "pull_request:44662", "right": "pull_request:45075", "accept": false, - "reason": "Different model additions (PenguinVL vs DeepSeek-OCR-2). They touch similar integration surfaces but implement unrelated models and codepaths." + "reason": "Both are model-add PRs, but for different architectures and codepaths (PenguinVL vs DeepSeek-OCR-2). Similar auto-registry/docs edits are generic template overlap." }, { "left": "pull_request:45075", "right": "pull_request:45401", "accept": false, - "reason": "Different model additions (DeepSeek-OCR-2 vs Voxtral TTS). Overlap is limited to common model-registration files, not the same bug or feature." + "reason": "Distinct model integrations for DeepSeek-OCR-2 and Voxtral TTS; they do not fix the same concrete bug or implement the same feature." } ] }, @@ -17930,7 +17954,7 @@ "soft_pairs": [ "pull_request:44979|pull_request:45363" ], - "prepared_review_unit_hash": "cab5bbb73f0047821d0397b1f59cedc42fce57429c1c5e45b383cd289fc89951", + "prepared_review_unit_hash": "06ec9e1a76dfbf95b769f72b6cb1bb318068b04ca92ed9ef03727052eb02cd61", "component_budget": { "node_count": 2, "item_count": 2, @@ -17952,18 +17976,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both PRs are fusion-related, but they target different layers and likely different change sets: #44979 introduces a standalone Module Fusion API, while #45363 adds end-to-end n-to-1 kernel fusion wiring via KernelConfig and model loading. They are related in theme, not the same concrete fix/change.", - "confidence": 0.86, + "summary": "These two pull requests are related only at a broad theme level (module/kernel fusion) but they do not look like the same underlying change. PR #44979 proposes a general Module Fusion API with a new `module_fusion.py` abstraction, while PR #45363 implements n-to-1 kernel fusion through `KernelConfig` and runtime integration points. They could coexist rather than be merged as duplicates.", + "confidence": 0.93, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #45363 is the stronger representative of the cluster because it spans the user-facing fusion flow end to end (KernelConfig, integrations, modeling_utils, and an example), whereas #44979 is a lower-level API addition for module fusion internals.", + "canonical_pr_reason": "PR #45363 is the better canonical PR for this cluster because it presents a concrete end-to-end fusion mechanism with integration into model loading/execution, whereas #44979 is a more generic API proposal with a narrower, earlier-looking abstraction.", "best_issue_reason": null, - "best_pr_reason": "PR #45363 is the best single artifact to keep as the cluster representative since it covers the broader, more actionable feature path and is closer to the practical mergeable change set.", + "best_pr_reason": "PR #45363 is the strongest standalone candidate: it has a specific feature goal, multiple integration files, and a clearer implementation path. PR #44979 is more of a foundational API sketch and does not appear to target the same concrete code path.", "soft_edge_verdicts": [ { "left": "pull_request:44979", "right": "pull_request:45363", "accept": false, - "reason": "They are not the same underlying code change: #44979 adds a general Module Fusion API in a new file, while #45363 implements kernel-based n-to-1 fusion plumbing through KernelConfig and model loading. Similar topic, but different abstraction level and merge target." + "reason": "Do not merge as duplicates: they address different fusion designs and different code paths. #44979 introduces a generic module-fusion API, while #45363 adds kernel-fusion via `KernelConfig` and related integration hooks." } ] }, @@ -17993,7 +18017,7 @@ "soft_pairs": [ "pull_request:45086|pull_request:45317" ], - "prepared_review_unit_hash": "1c322b94275d6648bf7092abf5664169d991dcf06c29be8d5f279c97763bbfe6", + "prepared_review_unit_hash": "2d118832712d15e2e1c24bf44c0540ee85adc704eea035846db7f0c9f9dc3922", "component_budget": { "node_count": 2, "item_count": 2, @@ -18015,18 +18039,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both PRs address the same Mistral regex patching AttributeError in the same code path. PR 45317 is the better canonical/best PR because it includes the code fix plus test coverage, while 45086 is the smaller earlier version of essentially the same change.", - "confidence": 0.97, + "summary": "Both pull requests implement the same Mistral regex fix in the same code path; the second PR adds test coverage, so they are duplicates with the later PR being the better canonical representative.", + "confidence": 0.98, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 45317 is the more complete instance of the fix: same underlying bug, same target file, and it adds tests in addition to the code change.", + "canonical_pr_reason": "PR 45317 is the better canonical PR because it contains the same core fix as 45086 plus a test file, making it the more complete and verifiable change.", "best_issue_reason": null, - "best_pr_reason": "PR 45317 is the best single artifact to keep because it subsumes the earlier patch and is more likely to be merge-ready due to added test coverage.", + "best_pr_reason": "PR 45317 is the strongest representative: same underlying AttributeError fix, same target issue, and additional tests for validation.", "soft_edge_verdicts": [ { "left": "pull_request:45086", "right": "pull_request:45317", "accept": true, - "reason": "Same concrete AttributeError fix in _patch_mistral_regex, same shared target/file, and the second PR is effectively an extended version of the first with tests; they could plausibly be merged into one PR." + "reason": "Same underlying bug fix in _patch_mistral_regex, same file, same issue target, and the later PR is just the fix plus tests." } ] }, @@ -18056,7 +18080,7 @@ "soft_pairs": [ "pull_request:45115|pull_request:45569" ], - "prepared_review_unit_hash": "7040051a56c35a46d47839ec33b3ff3871892a2faf417c855b69083881fa98a9", + "prepared_review_unit_hash": "a010488c94232a807cc2392fd56dca7dbcb9256503b4b633df6289ce3ff60fd8", "component_budget": { "node_count": 2, "item_count": 2, @@ -18078,18 +18102,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The two PRs are related NemotronH refactors, but they are not duplicates: PR#45115 is a narrow inheritance/weight-mapping cleanup, while PR#45569 is a broader model split that adds dedicated NemotronHDense/NemotronHSparse code, docs, and tests. The shared files reflect the same subsystem, not the same concrete change.", - "confidence": 0.84, + "summary": "These are related NemotronH refactors, but they are not the same change: one is a conversion-mapping/model-inheritance refactor, the other is a broader model split with docs, auto-mapping, and tests. They should not be merged as duplicates.", + "confidence": 0.88, "canonical_issue_reason": null, - "canonical_pr_reason": "PR#45569 is the more complete and forward-looking change: it introduces the dense/sparse model variants, updates auto-mappings and docs, and includes tests. PR#45115 is only a partial structural refactor focused on conversion mappings and inheritance cleanup.", + "canonical_pr_reason": "pull_request:45569 is the better canonical PR because it is the broader, more complete NemotronH architecture change, covering the new dense/sparse model split, auto mappings, docs, and tests; pull_request:45115 is a narrower conversion/mapping refactor.", "best_issue_reason": null, - "best_pr_reason": "PR#45569 best represents the final intended implementation because it covers the full NemotronH reorganization rather than just the conversion-mapping portion. It is the strongest candidate to keep as the cluster representative.", + "best_pr_reason": "pull_request:45569 is the best representative of the cluster since it touches the core model packaging and registration changes rather than only checkpoint conversion mappings.", "soft_edge_verdicts": [ { "left": "pull_request:45115", "right": "pull_request:45569", "accept": false, - "reason": "They overlap on NemotronH internals, but the concrete changes differ: one is a narrow refactor/mapping adjustment, the other is a broader architecture split with new model packages, docs, and tests. They are related work, not the same PR." + "reason": "Both are about NemotronH, but 45115 is mainly conversion-mapping and inheritance refactoring, while 45569 is a broader model split into dense/sparse variants with docs, auto-mapping, and tests. They do not appear to fix the same concrete code-path problem and would not plausibly merge into one PR." } ] }, @@ -18119,7 +18143,7 @@ "soft_pairs": [ "pull_request:45332|pull_request:45333" ], - "prepared_review_unit_hash": "e2fc1decbc8a4e2cc3769ae6a11b7009fa930005b952328e5b7c014f163307f9", + "prepared_review_unit_hash": "29d3bbe9ff0ca8230d72b8b8cedc208dfddbfdc9543a96ba9220ac7b1d82c809", "component_budget": { "node_count": 2, "item_count": 2, @@ -18141,18 +18165,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Two related open PRs are part of the same heterogeneous-model feature work, but they split along different layers: one adds per-layer config plumbing, the other adds modeling/runtime support. They overlap in infrastructure files, yet they are not the same concrete change and are not safe to merge as duplicates.", - "confidence": 0.88, + "summary": "The two PRs are related pieces of the same heterogeneous-model feature, but they are not duplicates: one adds per-layer config plumbing, while the other extends that plumbing into cache/masking/modeling paths.", + "confidence": 0.91, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #45332 is the broader umbrella change: it adds heterogeneous model support across cache, masking, modeling, and tests, so it is the most representative PR for the cluster.", + "canonical_pr_reason": "PR 45332 is the broader umbrella change: it implements heterogeneous model support across cache/masking/modeling and can incorporate the narrower per-layer config groundwork from 45333.", "best_issue_reason": null, - "best_pr_reason": "PR #45332 is the best single representative because it covers the wider end-to-end code path and subsumes the config-only groundwork from #45333.", + "best_pr_reason": "PR 45332 is the best representative of the cluster because it covers the end-to-end feature rather than only the config layer, making it the more comprehensive change to keep.", "soft_edge_verdicts": [ { "left": "pull_request:45332", "right": "pull_request:45333", "accept": false, - "reason": "They are closely related feature PRs, but one is config-layer support and the other is runtime/modeling support. They share groundwork files, yet they do not fix the same concrete bug or change the same code path, so they should not be merged as duplicates." + "reason": "Reject as a duplicate pair: 45333 adds per-layer configuration support, while 45332 uses that groundwork to change cache/modeling behavior. They are adjacent feature PRs, not the same concrete bug or change, and would not plausibly be merged into one PR without becoming a broader umbrella feature." } ] }, @@ -18182,12 +18206,12 @@ "soft_pairs": [ "pull_request:45415|pull_request:45425" ], - "prepared_review_unit_hash": "7f81b078333ede3d286e1eddc44240b2007d3a5de90114717aff3a3e71d7a719", + "prepared_review_unit_hash": "2ce4e6e36b81f041833bcaa1d30aef7a849fb0424cef41e080136cf23179395c", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 5514, + "serialized_chars": 5513, "estimated_input_tokens": 1379, "estimated_eval_tokens": 3014 }, @@ -18195,7 +18219,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 5514, + "serialized_chars": 5513, "estimated_input_tokens": 1379, "estimated_eval_tokens": 3014 }, @@ -18204,18 +18228,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are related only at a high level (typing cleanup), but they do not look like the same underlying change. PR 45415 is a broad type-checking sweep across many Transformers files plus pyproject ignores, while PR 45425 introduces shared typing protocols/aliases and updates a smaller set of typing consumers.", - "confidence": 0.92, + "summary": "Two typing-related pull requests, but they are different changes: one is a broad type-checking sweep across many transformers files, the other adds/adjusts typing protocols and helpers for specific models and integrations.", + "confidence": 0.87, "canonical_issue_reason": null, - "canonical_pr_reason": "45425 is the more self-contained change: it defines shared typing helpers and updates a limited set of consumers. 45415 is a much broader repo-wide typing pass, so it is less suitable as the representative PR for this cluster.", + "canonical_pr_reason": "PR #45415 is the clearest cluster anchor: it is a broad, repo-wide type-checking cleanup with many affected files and an explicit pyproject type-checking config change.", "best_issue_reason": null, - "best_pr_reason": "45425 is the best representative because it has a tighter scope and a clearer concrete code-path change, making it more coherent than the broad type-checking sweep in 45415.", + "best_pr_reason": "PR #45415 is the strongest representative of the cluster overall because it is the most comprehensive and clearly scoped typing-focused change.", "soft_edge_verdicts": [ { "left": "pull_request:45415", "right": "pull_request:45425", "accept": false, - "reason": "Both are typing-related, but they address different work: 45415 is a wide type-checking cleanup and config relaxation across many modules, while 45425 adds specific typing protocols/aliases for a few consumers. They do not appear mergeable as one PR or to fix the same concrete bug/change." + "reason": "Both are about typing, but they address different work: #45415 is a large type-checking pass, while #45425 adds specific typing protocols/helpers for modeling_utils and related code paths. They are not the same underlying bug/change and would not naturally merge into one duplicate PR." } ] }, @@ -18245,12 +18269,12 @@ "soft_pairs": [ "pull_request:45487|pull_request:45615" ], - "prepared_review_unit_hash": "3690261f39aeee742472cf34e1dbdabce18edc4b6b5eb85a3ccbf749ec78b47e", + "prepared_review_unit_hash": "52320edc30d152c884e1300110e352c4bb91f510be95feb3909f2886c0ea9dee", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 5160, + "serialized_chars": 5157, "estimated_input_tokens": 1290, "estimated_eval_tokens": 2836 }, @@ -18258,7 +18282,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 5160, + "serialized_chars": 5157, "estimated_input_tokens": 1290, "estimated_eval_tokens": 2836 }, @@ -18267,18 +18291,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These two pull requests are unrelated: one fixes a model-parallel device placement bug in AltCLIP/ChineseCLIP code, while the other only adds XPU test expectations for Qianfan OCR. They should not be merged or clustered as duplicates.", - "confidence": 0.98, + "summary": "45487 is a substantive cross-model device-placement bug fix; 45615 is only test expectation updates for XPU in qianfan_ocr. They are not the same underlying change.", + "confidence": 0.95, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #45487 is the stronger canonical PR because it fixes an actual runtime code-path bug in model parallel handling, with concrete source changes across affected model implementations.", + "canonical_pr_reason": "45487 fixes the concrete code-path bug: buffered token_type_ids are moved to the position_ids device before gather, addressing model-parallel/device mismatch across several text-embedding paths. It is the substantive implementation change here.", "best_issue_reason": null, - "best_pr_reason": "PR #45487 is the best representative of the cluster since it addresses the substantive bug fix; PR #45615 is test-only and does not represent the same underlying change.", + "best_pr_reason": "45487 is the best PR candidate because it contains the actual bug fix in model code, while 45615 only adjusts integration test expectations and does not fix the underlying runtime issue.", "soft_edge_verdicts": [ { "left": "pull_request:45487", "right": "pull_request:45615", "accept": false, - "reason": "Different problems and code paths: #45487 changes AltCLIP/ChineseCLIP model device handling for parallel execution, while #45615 only updates Qianfan OCR integration test outputs for XPU. They are not the same fix and would not plausibly be merged into one PR." + "reason": "Different artifacts and problems: 45487 changes model code to fix a device-placement bug in AltCLIP/related models, while 45615 only updates qianfan_ocr test outputs for XPU. They would not plausibly merge into one PR." } ] }, @@ -18308,7 +18332,7 @@ "soft_pairs": [ "pull_request:45548|pull_request:45662" ], - "prepared_review_unit_hash": "d5b6a8b4d5e394022172fd72147e6511f65c63aab321a8e243112e7f5c47b179", + "prepared_review_unit_hash": "3d4f19c0aaf88280f68877fdeeaaef1367a375f817d5dbef8716321c05cab30d", "component_budget": { "node_count": 2, "item_count": 2, @@ -18330,18 +18354,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The two PRs both touch expert-parallel loading, but they fix different concrete bugs: one is about DeepSpeed ZeRO-3 / accelerate launch serialization and loading, the other is about FSDP2 rank-0 broadcast overwriting experts. They are related in subsystem, not duplicates.", - "confidence": 0.92, + "summary": "Two PRs touch expert-parallel/modeling_utils, but they fix different loading failures: one is config serialization and EP detection for accelerate+ZeRO-3, the other is expert parameters being overwritten during FSDP2 loading.", + "confidence": 0.89, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #45662 is the better representative of the cluster because its title and diff describe a specific, concrete failure mode with a clear code-path fix across loading and MoE handling.", + "canonical_pr_reason": "No clear duplicate relationship between the PRs; if one must anchor the cluster, 45662 is the more concrete code-path fix for distributed expert loading, while 45548 is a narrower compatibility patch.", "best_issue_reason": null, - "best_pr_reason": "PR #45662 is the strongest standalone PR: it targets a well-defined bug, has a more explicit failure description, and its changes are directly tied to that bug rather than a broader loading workaround.", + "best_pr_reason": "45662 is the stronger representative because it addresses a specific runtime bug in expert parameter loading/wrapping across distributed execution, with broader code-path coverage.", "soft_edge_verdicts": [ { "left": "pull_request:45548", "right": "pull_request:45662", "accept": false, - "reason": "Both are about EP/loading, but #45548 addresses accelerate/DeepSpeed ZeRO-3 config serialization and a has_ep helper, while #45662 fixes FSDP2 expert tensors being overwritten by rank-0 broadcast. Different underlying bugs and not plausibly one merged PR." + "reason": "Different underlying bugs and fixes: 45548 removes distributed_config from serialization and adds has_ep for accelerate/ZeRO-3 compatibility, while 45662 fixes expert tensors being overwritten by rank-0 broadcast during FSDP2 loading. Shared EP context is too broad to treat as the same change." } ] }, @@ -18371,7 +18395,7 @@ "soft_pairs": [ "pull_request:45639|pull_request:45645" ], - "prepared_review_unit_hash": "5b3f5fe01c5886d8ec3a215cb301265219675c34e96128d9ce9bf8c36e98dcd3", + "prepared_review_unit_hash": "0b1a21c859ca0ae2e186591cca9b0b123c87c74e43762a5ccbd60a3b4186a745", "component_budget": { "node_count": 2, "item_count": 2, @@ -18393,18 +18417,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both PRs target the same xdist-related captured_info debug-log collision on issue 45561. PR 45645 is the broader, more complete version because it includes the core testing_utils fix plus CI/workflow handling to read multiple captured_info files; PR 45639 is a narrower subset of that same fix.", + "summary": "Both PRs address the same xdist-related captured_info logging problem in testing_utils, but 45645 is the broader, more complete change set because it also updates CI handling of multiple log artifacts. 45639 looks like a narrower precursor/slice of the same fix.", "confidence": 0.87, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 45645 is the best canonical PR because it covers the core bug fix and the supporting CI/log-collection changes, making it the most complete representative of the underlying xdist captured_info problem.", + "canonical_pr_reason": "PR 45645 is the better canonical representative because it contains the core xdist-safe testing_utils fix and the additional workflow changes needed to consume multiple captured_info files in CI, making it the more complete implementation.", "best_issue_reason": null, - "best_pr_reason": "PR 45645 is the strongest global representative: it addresses the same concrete xdist-safe captured_info behavior as 45639, but also handles downstream artifact collection so the fix is end-to-end.", + "best_pr_reason": "PR 45645 is the strongest overall candidate: it directly targets the xdist collision problem, includes the shared testing_utils change, and extends the fix to downstream CI/debug-log handling.", "soft_edge_verdicts": [ { "left": "pull_request:45639", "right": "pull_request:45645", "accept": true, - "reason": "Both PRs fix the same captured_info xdist collision problem in the same testing_utils code path; 45645 simply extends the fix with workflow changes to consume multiple files, so they could plausibly be merged into one PR." + "reason": "Same underlying xdist captured_info bug/change, with 45645 effectively subsuming 45639 by carrying the same testing_utils fix plus related CI artifact handling." } ] }, diff --git a/analysis/current/manifest.json b/analysis/current/manifest.json index 695c80a666c2ad4f58d9d3fdc85854a88a75d15a..a4787f0a41a627523863d3c7ced3082c3d9cddd7 100644 --- a/analysis/current/manifest.json +++ b/analysis/current/manifest.json @@ -1,8 +1,8 @@ { - "analysis_id": "hybrid-model-20260506t120019z", + "analysis_id": "hybrid-model-20260506t180045z", "archived_artifacts": { - "hybrid": "snapshots/20260506T120019Z/analysis-runs/hybrid-model-20260506t120019z/analysis-report-hybrid.json", - "hybrid_reviews": "snapshots/20260506T120019Z/analysis-runs/hybrid-model-20260506t120019z/analysis-report-hybrid.llm-reviews.json" + "hybrid": "snapshots/20260506T180045Z/analysis-runs/hybrid-model-20260506t180045z/analysis-report-hybrid.json", + "hybrid_reviews": "snapshots/20260506T180045Z/analysis-runs/hybrid-model-20260506t180045z/analysis-report-hybrid.llm-reviews.json" }, "artifacts": { "hybrid": "analysis/current/analysis-report-hybrid.json", @@ -10,9 +10,9 @@ }, "channel": "canonical", "model": null, - "published_at": "2026-05-06T12:08:33Z", + "published_at": "2026-05-06T18:15:25Z", "repo": "huggingface/transformers", "schema_version": 1, - "snapshot_id": "20260506T120019Z", + "snapshot_id": "20260506T180045Z", "variant": "hybrid" } diff --git a/snapshots/20260506T180045Z/analysis-runs/hybrid-model-20260506t180045z/analysis-report-hybrid.json b/snapshots/20260506T180045Z/analysis-runs/hybrid-model-20260506t180045z/analysis-report-hybrid.json new file mode 100644 index 0000000000000000000000000000000000000000..94ce5ee822ed2dd88041faad6b2a22d3bb3cafc7 --- /dev/null +++ b/snapshots/20260506T180045Z/analysis-runs/hybrid-model-20260506t180045z/analysis-report-hybrid.json @@ -0,0 +1,2076 @@ +{ + "schema_version": "1.0", + "repo": "huggingface/transformers", + "snapshot_id": "20260506T180045Z", + "generated_at": "2026-05-06T18:15:23Z", + "evidence_quality": "full", + "llm_enrichment": true, + "meta_bugs": [ + { + "cluster_id": "cluster-43979-11", + "summary": "Cluster of 1 issues and 10 PRs centered on issue #43979.", + "status": "open", + "confidence": 0.8, + "canonical_issue_number": 43979, + "canonical_pr_number": 44007, + "issue_numbers": [ + 43979 + ], + "pr_numbers": [ + 43996, + 44007, + 44013, + 44044, + 44066, + 44072, + 44085, + 44129, + 44154, + 44722 + ], + "evidence_types": [ + "closing_reference", + "shared_issue_target", + "soft_similarity" + ], + "pr_comparisons": [ + { + "left_pr_number": 43996, + "right_pr_number": 44007, + "code_similarity": 0.179, + "size_similarity": 0.576, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.429, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 43996, + "right_pr_number": 44013, + "code_similarity": 0.122, + "size_similarity": 0.318, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.392, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 43996, + "right_pr_number": 44044, + "code_similarity": 0.245, + "size_similarity": 0.864, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.479, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 43996, + "right_pr_number": 44066, + "code_similarity": 0.225, + "size_similarity": 0.818, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.408, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 43996, + "right_pr_number": 44072, + "code_similarity": 0.14, + "size_similarity": 0.303, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.528, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 43996, + "right_pr_number": 44085, + "code_similarity": 0.216, + "size_similarity": 0.783, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.398, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 43996, + "right_pr_number": 44129, + "code_similarity": 0.163, + "size_similarity": 0.643, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.229, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 43996, + "right_pr_number": 44154, + "code_similarity": 0.153, + "size_similarity": 0.535, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.31, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 43996, + "right_pr_number": 44722, + "code_similarity": 0.225, + "size_similarity": 0.848, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.368, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44007, + "right_pr_number": 44013, + "code_similarity": 0.19, + "size_similarity": 0.553, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.531, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44007, + "right_pr_number": 44044, + "code_similarity": 0.186, + "size_similarity": 0.667, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.354, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44007, + "right_pr_number": 44066, + "code_similarity": 0.188, + "size_similarity": 0.704, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.315, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44007, + "right_pr_number": 44072, + "code_similarity": 0.212, + "size_similarity": 0.526, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.708, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44007, + "right_pr_number": 44085, + "code_similarity": 0.195, + "size_similarity": 0.735, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.318, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44007, + "right_pr_number": 44129, + "code_similarity": 0.103, + "size_similarity": 0.37, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.191, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44007, + "right_pr_number": 44154, + "code_similarity": 0.238, + "size_similarity": 0.93, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.344, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44007, + "right_pr_number": 44722, + "code_similarity": 0.178, + "size_similarity": 0.679, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.28, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44013, + "right_pr_number": 44044, + "code_similarity": 0.126, + "size_similarity": 0.368, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.351, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44013, + "right_pr_number": 44066, + "code_similarity": 0.127, + "size_similarity": 0.389, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.325, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44013, + "right_pr_number": 44072, + "code_similarity": 0.29, + "size_similarity": 0.952, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.667, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44013, + "right_pr_number": 44085, + "code_similarity": 0.131, + "size_similarity": 0.406, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.329, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44013, + "right_pr_number": 44129, + "code_similarity": 0.07, + "size_similarity": 0.205, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.192, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44013, + "right_pr_number": 44154, + "code_similarity": 0.177, + "size_similarity": 0.594, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.389, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44013, + "right_pr_number": 44722, + "code_similarity": 0.118, + "size_similarity": 0.375, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.287, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44044, + "right_pr_number": 44066, + "code_similarity": 0.25, + "size_similarity": 0.947, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.404, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44044, + "right_pr_number": 44072, + "code_similarity": 0.136, + "size_similarity": 0.351, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.442, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44044, + "right_pr_number": 44085, + "code_similarity": 0.24, + "size_similarity": 0.906, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.394, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44044, + "right_pr_number": 44129, + "code_similarity": 0.147, + "size_similarity": 0.555, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.243, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44044, + "right_pr_number": 44154, + "code_similarity": 0.17, + "size_similarity": 0.62, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.306, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44044, + "right_pr_number": 44722, + "code_similarity": 0.257, + "size_similarity": 0.982, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.402, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44066, + "right_pr_number": 44072, + "code_similarity": 0.133, + "size_similarity": 0.37, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.393, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44066, + "right_pr_number": 44085, + "code_similarity": 0.763, + "size_similarity": 0.957, + "file_overlap": 0.5, + "area_overlap": 0.825, + "patch_similarity": 0.887, + "shared_filenames": [ + "src/transformers/models/gptj/modeling_gptj.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/models/gptj/modeling_gptj.py", + "left_ranges": [ + [ + 33, + 41 + ], + [ + 174, + 181 + ], + [ + 250, + 257 + ], + [ + 398, + 419 + ], + [ + 425, + 434 + ], + [ + 459, + 466 + ], + [ + 471, + 488 + ], + [ + 518, + 540 + ], + [ + 554, + 560 + ], + [ + 566, + 575 + ], + [ + 580, + 598 + ], + [ + 601, + 606 + ], + [ + 634, + 640 + ], + [ + 646, + 653 + ], + [ + 658, + 674 + ], + [ + 717, + 722 + ], + [ + 738, + 744 + ], + [ + 749, + 772 + ], + [ + 790, + 795 + ] + ], + "right_ranges": [ + [ + 33, + 40 + ], + [ + 173, + 180 + ], + [ + 249, + 256 + ], + [ + 397, + 405 + ], + [ + 408, + 419 + ], + [ + 425, + 434 + ], + [ + 459, + 465 + ], + [ + 470, + 489 + ], + [ + 517, + 539 + ], + [ + 553, + 559 + ], + [ + 565, + 574 + ], + [ + 579, + 597 + ], + [ + 600, + 611 + ], + [ + 633, + 639 + ], + [ + 645, + 652 + ], + [ + 657, + 673 + ], + [ + 716, + 728 + ], + [ + 737, + 743 + ], + [ + 748, + 771 + ], + [ + 789, + 794 + ] + ] + } + ] + }, + { + "left_pr_number": 44066, + "right_pr_number": 44129, + "code_similarity": 0.145, + "size_similarity": 0.526, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.263, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44066, + "right_pr_number": 44154, + "code_similarity": 0.174, + "size_similarity": 0.654, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.286, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44066, + "right_pr_number": 44722, + "code_similarity": 0.907, + "size_similarity": 0.964, + "file_overlap": 1.0, + "area_overlap": 0.808, + "patch_similarity": 0.874, + "shared_filenames": [ + "src/transformers/models/codegen/modeling_codegen.py", + "src/transformers/models/gptj/modeling_gptj.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/models/codegen/modeling_codegen.py", + "left_ranges": [ + [ + 245, + 266 + ] + ], + "right_ranges": [ + [ + 228, + 234 + ] + ] + }, + { + "filename": "src/transformers/models/gptj/modeling_gptj.py", + "left_ranges": [ + [ + 33, + 41 + ], + [ + 174, + 181 + ], + [ + 250, + 257 + ], + [ + 398, + 419 + ], + [ + 425, + 434 + ], + [ + 459, + 466 + ], + [ + 471, + 488 + ], + [ + 518, + 540 + ], + [ + 554, + 560 + ], + [ + 566, + 575 + ], + [ + 580, + 598 + ], + [ + 601, + 606 + ], + [ + 634, + 640 + ], + [ + 646, + 653 + ], + [ + 658, + 674 + ], + [ + 717, + 722 + ], + [ + 738, + 744 + ], + [ + 749, + 772 + ], + [ + 790, + 795 + ] + ], + "right_ranges": [ + [ + 33, + 42 + ], + [ + 175, + 182 + ], + [ + 251, + 258 + ], + [ + 399, + 420 + ], + [ + 426, + 435 + ], + [ + 460, + 467 + ], + [ + 472, + 489 + ], + [ + 519, + 540 + ], + [ + 554, + 560 + ], + [ + 566, + 575 + ], + [ + 580, + 598 + ], + [ + 601, + 612 + ], + [ + 634, + 640 + ], + [ + 646, + 653 + ], + [ + 658, + 663 + ], + [ + 666, + 674 + ], + [ + 717, + 722 + ], + [ + 738, + 744 + ], + [ + 749, + 772 + ], + [ + 790, + 795 + ] + ] + } + ] + }, + { + "left_pr_number": 44072, + "right_pr_number": 44085, + "code_similarity": 0.137, + "size_similarity": 0.387, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.398, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44072, + "right_pr_number": 44129, + "code_similarity": 0.074, + "size_similarity": 0.195, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.231, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44072, + "right_pr_number": 44154, + "code_similarity": 0.175, + "size_similarity": 0.566, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.414, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44072, + "right_pr_number": 44722, + "code_similarity": 0.124, + "size_similarity": 0.357, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.347, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44085, + "right_pr_number": 44129, + "code_similarity": 0.141, + "size_similarity": 0.503, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.272, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44085, + "right_pr_number": 44154, + "code_similarity": 0.18, + "size_similarity": 0.684, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.289, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44085, + "right_pr_number": 44722, + "code_similarity": 0.728, + "size_similarity": 0.923, + "file_overlap": 0.5, + "area_overlap": 0.791, + "patch_similarity": 0.78, + "shared_filenames": [ + "src/transformers/models/gptj/modeling_gptj.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/models/gptj/modeling_gptj.py", + "left_ranges": [ + [ + 33, + 40 + ], + [ + 173, + 180 + ], + [ + 249, + 256 + ], + [ + 397, + 405 + ], + [ + 408, + 419 + ], + [ + 425, + 434 + ], + [ + 459, + 465 + ], + [ + 470, + 489 + ], + [ + 517, + 539 + ], + [ + 553, + 559 + ], + [ + 565, + 574 + ], + [ + 579, + 597 + ], + [ + 600, + 611 + ], + [ + 633, + 639 + ], + [ + 645, + 652 + ], + [ + 657, + 673 + ], + [ + 716, + 728 + ], + [ + 737, + 743 + ], + [ + 748, + 771 + ], + [ + 789, + 794 + ] + ], + "right_ranges": [ + [ + 33, + 42 + ], + [ + 175, + 182 + ], + [ + 251, + 258 + ], + [ + 399, + 420 + ], + [ + 426, + 435 + ], + [ + 460, + 467 + ], + [ + 472, + 489 + ], + [ + 519, + 540 + ], + [ + 554, + 560 + ], + [ + 566, + 575 + ], + [ + 580, + 598 + ], + [ + 601, + 612 + ], + [ + 634, + 640 + ], + [ + 646, + 653 + ], + [ + 658, + 663 + ], + [ + 666, + 674 + ], + [ + 717, + 722 + ], + [ + 738, + 744 + ], + [ + 749, + 772 + ], + [ + 790, + 795 + ] + ] + } + ] + }, + { + "left_pr_number": 44129, + "right_pr_number": 44154, + "code_similarity": 0.099, + "size_similarity": 0.344, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.199, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44129, + "right_pr_number": 44722, + "code_similarity": 0.146, + "size_similarity": 0.545, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.247, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44154, + "right_pr_number": 44722, + "code_similarity": 0.164, + "size_similarity": 0.631, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.255, + "shared_filenames": [], + "shared_file_areas": [] + } + ] + }, + { + "cluster_id": "cluster-41211-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #41211.", + "status": "open", + "confidence": 0.55, + "canonical_issue_number": 41211, + "canonical_pr_number": 44339, + "issue_numbers": [ + 41211 + ], + "pr_numbers": [ + 41356, + 44339 + ], + "evidence_types": [ + "closing_reference" + ], + "pr_comparisons": [ + { + "left_pr_number": 41356, + "right_pr_number": 44339, + "code_similarity": 0.155, + "size_similarity": 0.094, + "file_overlap": 0.273, + "area_overlap": 0.078, + "patch_similarity": 0.183, + "shared_filenames": [ + "docs/source/en/model_doc/deimv2.md", + "src/transformers/models/deimv2/__init__.py", + "src/transformers/models/deimv2/configuration_deimv2.py", + "src/transformers/models/deimv2/modeling_deimv2.py", + "tests/models/deimv2/__init__.py", + "tests/models/deimv2/test_modeling_deimv2.py" + ], + "shared_file_areas": [ + { + "filename": "docs/source/en/model_doc/deimv2.md", + "left_ranges": [ + [ + 1, + 132 + ] + ], + "right_ranges": [ + [ + 1, + 65 + ] + ] + }, + { + "filename": "src/transformers/models/deimv2/__init__.py", + "left_ranges": [ + [ + 1, + 15 + ] + ], + "right_ranges": [ + [ + 1, + 29 + ] + ] + }, + { + "filename": "src/transformers/models/deimv2/configuration_deimv2.py", + "left_ranges": [ + [ + 1, + 74 + ] + ], + "right_ranges": [ + [ + 1, + 266 + ] + ] + }, + { + "filename": "tests/models/deimv2/test_modeling_deimv2.py", + "left_ranges": [ + [ + 1, + 15 + ] + ], + "right_ranges": [ + [ + 1, + 1734 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-43240-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #43240.", + "status": "open", + "confidence": 0.75, + "canonical_issue_number": 43240, + "canonical_pr_number": 43251, + "issue_numbers": [ + 43240 + ], + "pr_numbers": [ + 43251, + 43254 + ], + "evidence_types": [ + "closing_reference", + "shared_issue_target" + ], + "pr_comparisons": [ + { + "left_pr_number": 43251, + "right_pr_number": 43254, + "code_similarity": 0.794, + "size_similarity": 0.64, + "file_overlap": 1.0, + "area_overlap": 0.667, + "patch_similarity": 0.882, + "shared_filenames": [ + "src/transformers/loss/loss_utils.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/loss/loss_utils.py", + "left_ranges": [ + [ + 30, + 50 + ] + ], + "right_ranges": [ + [ + 30, + 43 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-41115-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #41115.", + "status": "open", + "confidence": 0.55, + "canonical_issue_number": 41115, + "canonical_pr_number": 45613, + "issue_numbers": [ + 41115 + ], + "pr_numbers": [ + 41116, + 45613 + ], + "evidence_types": [ + "closing_reference" + ], + "pr_comparisons": [ + { + "left_pr_number": 41116, + "right_pr_number": 45613, + "code_similarity": 0.805, + "size_similarity": 0.885, + "file_overlap": 0.75, + "area_overlap": 0.854, + "patch_similarity": 0.693, + "shared_filenames": [ + "docs/source/en/_toctree.yml", + "docs/source/en/model_doc/minicpm3.md", + "src/transformers/models/auto/modeling_auto.py", + "src/transformers/models/minicpm3/__init__.py", + "src/transformers/models/minicpm3/configuration_minicpm3.py", + "src/transformers/models/minicpm3/modeling_minicpm3.py", + "src/transformers/models/minicpm3/modular_minicpm3.py", + "tests/models/minicpm3/__init__.py", + "tests/models/minicpm3/test_modeling_minicpm3.py" + ], + "shared_file_areas": [ + { + "filename": "docs/source/en/_toctree.yml", + "left_ranges": [ + [ + 713, + 720 + ] + ], + "right_ranges": [ + [ + 713, + 720 + ] + ] + }, + { + "filename": "docs/source/en/model_doc/minicpm3.md", + "left_ranges": [ + [ + 1, + 67 + ] + ], + "right_ranges": [ + [ + 1, + 45 + ] + ] + }, + { + "filename": "src/transformers/models/auto/modeling_auto.py", + "left_ranges": [ + [ + 285, + 291 + ], + [ + 700, + 706 + ], + [ + 1298, + 1304 + ] + ], + "right_ranges": [ + [ + 699, + 705 + ], + [ + 1301, + 1307 + ] + ] + }, + { + "filename": "src/transformers/models/minicpm3/__init__.py", + "left_ranges": [ + [ + 1, + 29 + ] + ], + "right_ranges": [ + [ + 1, + 29 + ] + ] + }, + { + "filename": "src/transformers/models/minicpm3/configuration_minicpm3.py", + "left_ranges": [ + [ + 1, + 141 + ] + ], + "right_ranges": [ + [ + 1, + 126 + ] + ] + }, + { + "filename": "src/transformers/models/minicpm3/modeling_minicpm3.py", + "left_ranges": [ + [ + 1, + 544 + ] + ], + "right_ranges": [ + [ + 1, + 522 + ] + ] + }, + { + "filename": "src/transformers/models/minicpm3/modular_minicpm3.py", + "left_ranges": [ + [ + 1, + 444 + ] + ], + "right_ranges": [ + [ + 1, + 342 + ] + ] + }, + { + "filename": "tests/models/minicpm3/test_modeling_minicpm3.py", + "left_ranges": [ + [ + 1, + 109 + ] + ], + "right_ranges": [ + [ + 1, + 136 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-45647-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #45647.", + "status": "open", + "confidence": 0.75, + "canonical_issue_number": 45647, + "canonical_pr_number": 45738, + "issue_numbers": [ + 45647 + ], + "pr_numbers": [ + 45737, + 45738 + ], + "evidence_types": [ + "closing_reference", + "shared_issue_target" + ], + "pr_comparisons": [ + { + "left_pr_number": 45737, + "right_pr_number": 45738, + "code_similarity": 0.526, + "size_similarity": 0.11, + "file_overlap": 0.5, + "area_overlap": 1.0, + "patch_similarity": 0.024, + "shared_filenames": [ + "src/transformers/models/musicgen_melody/modeling_musicgen_melody.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/models/musicgen_melody/modeling_musicgen_melody.py", + "left_ranges": [ + [ + 491, + 497 + ] + ], + "right_ranges": [ + [ + 491, + 497 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-43824-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #43824.", + "status": "open", + "confidence": 0.75, + "canonical_issue_number": 43824, + "canonical_pr_number": 43836, + "issue_numbers": [ + 43824 + ], + "pr_numbers": [ + 43836, + 43842 + ], + "evidence_types": [ + "closing_reference", + "shared_issue_target" + ], + "pr_comparisons": [ + { + "left_pr_number": 43836, + "right_pr_number": 43842, + "code_similarity": 0.332, + "size_similarity": 0.088, + "file_overlap": 1.0, + "area_overlap": 0.033, + "patch_similarity": 0.017, + "shared_filenames": [ + "src/transformers/cli/serve.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/cli/serve.py", + "left_ranges": [ + [ + 11, + 18 + ], + [ + 359, + 429 + ], + [ + 584, + 590 + ], + [ + 1892, + 1910 + ], + [ + 1917, + 1923 + ] + ], + "right_ranges": [ + [ + 54, + 61 + ], + [ + 587, + 593 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-45081-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #45081.", + "status": "open", + "confidence": 0.8, + "canonical_issue_number": 45081, + "canonical_pr_number": 45317, + "issue_numbers": [ + 45081 + ], + "pr_numbers": [ + 45086, + 45317 + ], + "evidence_types": [ + "closing_reference", + "shared_issue_target", + "soft_similarity" + ], + "pr_comparisons": [ + { + "left_pr_number": 45086, + "right_pr_number": 45317, + "code_similarity": 0.589, + "size_similarity": 0.34, + "file_overlap": 0.5, + "area_overlap": 1.0, + "patch_similarity": 0.136, + "shared_filenames": [ + "src/transformers/tokenization_utils_tokenizers.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/tokenization_utils_tokenizers.py", + "left_ranges": [ + [ + 1360, + 1370 + ], + [ + 1374, + 1380 + ] + ], + "right_ranges": [ + [ + 1360, + 1370 + ], + [ + 1374, + 1380 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-45561-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #45561.", + "status": "open", + "confidence": 0.8, + "canonical_issue_number": 45561, + "canonical_pr_number": 45645, + "issue_numbers": [ + 45561 + ], + "pr_numbers": [ + 45639, + 45645 + ], + "evidence_types": [ + "closing_reference", + "shared_issue_target", + "soft_similarity" + ], + "pr_comparisons": [ + { + "left_pr_number": 45639, + "right_pr_number": 45645, + "code_similarity": 0.532, + "size_similarity": 0.619, + "file_overlap": 0.5, + "area_overlap": 0.543, + "patch_similarity": 0.451, + "shared_filenames": [ + "src/transformers/testing_utils.py", + "tests/utils/test_testing_utils.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/testing_utils.py", + "left_ranges": [ + [ + 3529, + 3536 + ], + [ + 3753, + 3787 + ] + ], + "right_ranges": [ + [ + 3525, + 3558 + ], + [ + 3782, + 3788 + ] + ] + }, + { + "filename": "tests/utils/test_testing_utils.py", + "left_ranges": [ + [ + 1, + 86 + ] + ], + "right_ranges": [ + [ + 1, + 114 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-43698-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #43698.", + "status": "open", + "confidence": 0.75, + "canonical_issue_number": 43698, + "canonical_pr_number": 43779, + "issue_numbers": [ + 43698 + ], + "pr_numbers": [ + 43779, + 43816 + ], + "evidence_types": [ + "closing_reference", + "shared_issue_target" + ], + "pr_comparisons": [ + { + "left_pr_number": 43779, + "right_pr_number": 43816, + "code_similarity": 0.418, + "size_similarity": 0.538, + "file_overlap": 1.0, + "area_overlap": 0.02, + "patch_similarity": 0.02, + "shared_filenames": [ + "src/transformers/integrations/integration_utils.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/integrations/integration_utils.py", + "left_ranges": [ + [ + 2227, + 2241 + ], + [ + 2303, + 2309 + ] + ], + "right_ranges": [ + [ + 2278, + 2291 + ], + [ + 2309, + 2322 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-44018-2", + "summary": "Cluster of 2 related pull requests linked by soft_similarity.", + "status": "open", + "confidence": 0.5, + "canonical_issue_number": null, + "canonical_pr_number": 44068, + "issue_numbers": [], + "pr_numbers": [ + 44018, + 44068 + ], + "evidence_types": [ + "soft_similarity" + ], + "pr_comparisons": [ + { + "left_pr_number": 44018, + "right_pr_number": 44068, + "code_similarity": 0.766, + "size_similarity": 0.939, + "file_overlap": 1.0, + "area_overlap": 0.425, + "patch_similarity": 0.866, + "shared_filenames": [ + "src/transformers/models/gpt_neo/modeling_gpt_neo.py", + "tests/models/gpt_neo/test_modeling_gpt_neo.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/models/gpt_neo/modeling_gpt_neo.py", + "left_ranges": [ + [ + 26, + 46 + ], + [ + 138, + 143 + ], + [ + 182, + 187 + ], + [ + 283, + 295 + ], + [ + 327, + 341 + ], + [ + 348, + 354 + ], + [ + 360, + 369 + ], + [ + 399, + 406 + ], + [ + 411, + 419 + ], + [ + 428, + 433 + ], + [ + 472, + 492 + ], + [ + 507, + 513 + ], + [ + 519, + 528 + ], + [ + 541, + 559 + ], + [ + 562, + 567 + ], + [ + 595, + 601 + ], + [ + 607, + 614 + ], + [ + 627, + 643 + ], + [ + 685, + 690 + ], + [ + 708, + 714 + ], + [ + 720, + 727 + ], + [ + 740, + 757 + ], + [ + 761, + 766 + ], + [ + 780, + 786 + ], + [ + 791, + 798 + ], + [ + 807, + 822 + ], + [ + 840, + 845 + ] + ], + "right_ranges": [ + [ + 26, + 31 + ], + [ + 34, + 44 + ], + [ + 136, + 141 + ], + [ + 180, + 185 + ], + [ + 281, + 293 + ], + [ + 325, + 339 + ], + [ + 346, + 352 + ], + [ + 358, + 367 + ], + [ + 397, + 404 + ], + [ + 409, + 417 + ], + [ + 426, + 434 + ], + [ + 463, + 483 + ], + [ + 498, + 504 + ], + [ + 510, + 519 + ], + [ + 532, + 550 + ], + [ + 553, + 564 + ], + [ + 586, + 592 + ], + [ + 598, + 605 + ], + [ + 618, + 634 + ], + [ + 676, + 688 + ], + [ + 699, + 705 + ], + [ + 711, + 718 + ], + [ + 731, + 748 + ], + [ + 752, + 762 + ], + [ + 771, + 777 + ], + [ + 782, + 789 + ], + [ + 798, + 813 + ], + [ + 831, + 836 + ] + ] + }, + { + "filename": "tests/models/gpt_neo/test_modeling_gpt_neo.py", + "left_ranges": [ + [ + 458, + 464 + ] + ], + "right_ranges": [ + [ + 458, + 464 + ] + ] + } + ] + } + ] + } + ], + "duplicate_issues": [], + "duplicate_prs": [ + { + "cluster_id": "cluster-41115-3", + "canonical_pr_number": 45613, + "duplicate_pr_numbers": [ + 41116 + ], + "target_issue_number": 41115, + "reason": "PRs in cluster-41115-3 are treated as duplicates because they converge on issue #41115 with closing_reference evidence." + }, + { + "cluster_id": "cluster-41211-3", + "canonical_pr_number": 44339, + "duplicate_pr_numbers": [ + 41356 + ], + "target_issue_number": 41211, + "reason": "PRs in cluster-41211-3 are treated as duplicates because they converge on issue #41211 with closing_reference evidence." + }, + { + "cluster_id": "cluster-43240-3", + "canonical_pr_number": 43251, + "duplicate_pr_numbers": [ + 43254 + ], + "target_issue_number": 43240, + "reason": "PRs in cluster-43240-3 are treated as duplicates because they converge on issue #43240 with closing_reference, shared_issue_target evidence." + }, + { + "cluster_id": "cluster-43698-3", + "canonical_pr_number": 43779, + "duplicate_pr_numbers": [ + 43816 + ], + "target_issue_number": 43698, + "reason": "PRs in cluster-43698-3 are treated as duplicates because they converge on issue #43698 with closing_reference, shared_issue_target evidence." + }, + { + "cluster_id": "cluster-43824-3", + "canonical_pr_number": 43836, + "duplicate_pr_numbers": [ + 43842 + ], + "target_issue_number": 43824, + "reason": "PRs in cluster-43824-3 are treated as duplicates because they converge on issue #43824 with closing_reference, shared_issue_target evidence." + }, + { + "cluster_id": "cluster-43979-11", + "canonical_pr_number": 44007, + "duplicate_pr_numbers": [ + 43996, + 44013, + 44044, + 44066, + 44072, + 44085, + 44129, + 44154, + 44722 + ], + "target_issue_number": 43979, + "reason": "PRs in cluster-43979-11 are treated as duplicates because they converge on issue #43979 with closing_reference, shared_issue_target, soft_similarity evidence." + }, + { + "cluster_id": "cluster-44018-2", + "canonical_pr_number": 44068, + "duplicate_pr_numbers": [ + 44018 + ], + "target_issue_number": null, + "reason": "PRs in cluster-44018-2 are treated as duplicates because they share soft_similarity evidence." + }, + { + "cluster_id": "cluster-45081-3", + "canonical_pr_number": 45317, + "duplicate_pr_numbers": [ + 45086 + ], + "target_issue_number": 45081, + "reason": "PRs in cluster-45081-3 are treated as duplicates because they converge on issue #45081 with closing_reference, shared_issue_target, soft_similarity evidence." + }, + { + "cluster_id": "cluster-45561-3", + "canonical_pr_number": 45645, + "duplicate_pr_numbers": [ + 45639 + ], + "target_issue_number": 45561, + "reason": "PRs in cluster-45561-3 are treated as duplicates because they converge on issue #45561 with closing_reference, shared_issue_target, soft_similarity evidence." + }, + { + "cluster_id": "cluster-45647-3", + "canonical_pr_number": 45738, + "duplicate_pr_numbers": [ + 45737 + ], + "target_issue_number": 45647, + "reason": "PRs in cluster-45647-3 are treated as duplicates because they converge on issue #45647 with closing_reference, shared_issue_target evidence." + } + ], + "best_issue": { + "cluster_id": "cluster-45647-3", + "issue_number": 45647, + "reason": "Issue #45647 is the strongest global issue candidate because it is open, belongs to a cluster with 3 artifacts, and carries 12 discussion comments plus 8 inbound references.", + "score": 162.0 + }, + "best_pr": { + "cluster_id": "cluster-43979-11", + "pr_number": 44007, + "reason": "PR #44007 is the strongest global PR candidate because it is open, belongs to a cluster with 11 artifacts, links to 1 issues, and carries 2 review/discussion events.", + "score": 238.886 + } +} diff --git a/snapshots/20260506T180045Z/analysis-runs/hybrid-model-20260506t180045z/analysis-report-hybrid.llm-reviews.json b/snapshots/20260506T180045Z/analysis-runs/hybrid-model-20260506t180045z/analysis-report-hybrid.llm-reviews.json new file mode 100644 index 0000000000000000000000000000000000000000..f79fecff12cfa0fdf7e433cc4edcb3b7cd8dee0c --- /dev/null +++ b/snapshots/20260506T180045Z/analysis-runs/hybrid-model-20260506t180045z/analysis-report-hybrid.llm-reviews.json @@ -0,0 +1,18572 @@ +{ + "schema_version": "1.0", + "repo": "huggingface/transformers", + "snapshot_id": "20260506T180045Z", + "generated_at": "2026-05-06T18:15:23Z", + "model": "gpt-5.4-mini?service_tier=flex", + "reviews": [ + { + "label": "issue", + "component_index": 1, + "component_count": 16, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-13244-3", + "nodes": [ + "issue:13244", + "issue:15980", + "issue:16618" + ], + "soft_pairs": [ + "issue:13244|issue:16618", + "issue:15980|issue:16618" + ], + "prepared_review_unit_hash": "285d43af9708d76b766e40051da8365659a39c2e42962ae4c4220e916293123f", + "component_budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 2478, + "estimated_input_tokens": 620, + "estimated_eval_tokens": 1496 + }, + "budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 2478, + "estimated_input_tokens": 620, + "estimated_eval_tokens": 1496 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "This cluster looks heterogeneous: one issue is about Tapas tokenization differences, one about private-model auth/error messaging, and one about failing to load bert-base-uncased. The only plausible anchor is the generic model-loading report, but the soft links do not justify deduping the issues together.", + "confidence": 0.92, + "canonical_issue_reason": "issue:16618 is the most central and broadly phrased report in the cluster, and it is the only node softly linked to both other issues, so it is the best anchor even though the topics are likely distinct.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:16618 is the best overall issue candidate because it is the most general failure description and sits at the center of the soft-similarity links.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:13244", + "right": "issue:16618", + "accept": false, + "reason": "Tokenization mismatch in Tapas and inability to load bert-base-uncased are different problem types with no clear shared code-path or fix." + }, + { + "left": "issue:15980", + "right": "issue:16618", + "accept": false, + "reason": "Private-model login/error-message handling and a generic model-loading failure are not the same underlying bug; the overlap is too broad to merge." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 1, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-387", + "nodes": [ + "issue:24643", + "issue:30277", + "issue:34634", + "issue:34928", + "issue:35545", + "issue:36010", + "issue:36331", + "issue:39290", + "issue:41093", + "issue:41720", + "issue:41762", + "issue:42915", + "issue:43039", + "issue:43404", + "issue:43425", + "issue:43531", + "issue:43541", + "issue:43596", + "issue:43638", + "issue:43716", + "issue:43792", + "issue:43827", + "issue:43828", + "issue:43854", + "issue:43866", + "issue:43901", + "issue:43975", + "issue:44292", + "issue:44322", + "issue:44457", + "issue:44509", + "issue:44512", + "issue:44560", + "issue:44661", + "issue:44805", + "issue:44841", + "issue:44863", + "issue:44918", + "issue:45070", + "issue:45081", + "issue:45084", + "issue:45161", + "issue:45237", + "issue:45362", + "issue:45464", + "issue:45507", + "issue:45685", + "issue:45800" + ], + "soft_pairs": [ + "issue:44322|issue:45464", + "issue:45237|issue:45685", + "issue:43716|issue:45237", + "issue:43828|issue:45237", + "issue:45237|issue:45507", + "issue:44292|issue:45237", + "issue:30277|issue:35545", + "issue:35545|issue:42915", + "issue:35545|issue:41720", + "issue:44509|issue:44512", + "issue:35545|issue:36010", + "issue:43638|issue:44805", + "issue:43596|issue:44805", + "issue:44805|issue:45161", + "issue:44805|issue:44918", + "issue:44661|issue:44805", + "issue:34634|issue:35545", + "issue:30277|issue:43638", + "issue:43596|issue:43638", + "issue:43975|issue:44457", + "issue:43039|issue:45161", + "issue:44292|issue:44457", + "issue:34928|issue:44805", + "issue:24643|issue:30277", + "issue:43638|issue:45161", + "issue:34928|issue:36331", + "issue:43596|issue:45161", + "issue:43716|issue:45507", + "issue:39290|issue:45081", + "issue:43828|issue:45161", + "issue:43425|issue:44292", + "issue:44918|issue:45161", + "issue:34928|issue:41720", + "issue:43716|issue:45685", + "issue:43541|issue:43716", + "issue:43828|issue:44292", + "issue:45507|issue:45685", + "issue:43716|issue:44292", + "issue:44292|issue:44918", + "issue:43039|issue:44918", + "issue:41762|issue:43638", + "issue:43827|issue:44512", + "issue:43716|issue:44560", + "issue:43828|issue:45685", + "issue:44560|issue:45685", + "issue:43039|issue:43828", + "issue:30277|issue:36331", + "issue:34928|issue:36010", + "issue:43039|issue:44805", + "issue:41762|issue:43596", + "issue:43901|issue:44512", + "issue:39290|issue:41720", + "issue:43039|issue:44560", + "issue:43866|issue:44863", + "issue:43854|issue:43866", + "issue:39290|issue:43531", + "issue:41093|issue:43638", + "issue:43425|issue:45070", + "issue:43596|issue:45800", + "issue:45362|issue:45464", + "issue:43792|issue:45070", + "issue:43404|issue:43866", + "issue:45464|issue:45685", + "issue:43541|issue:45685", + "issue:43541|issue:43828", + "issue:36010|issue:39290", + "issue:44841|issue:45084", + "issue:41093|issue:41762" + ], + "prepared_review_unit_hash": "05f15a0f5f953554c2204d90c90d285906aade22f4def3724a0442b01d248647", + "component_budget": { + "node_count": 387, + "item_count": 387, + "soft_pair_count": 1455, + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 68, + "serialized_chars": 47982, + "estimated_input_tokens": 11996, + "estimated_eval_tokens": 24248 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This set is mostly a loose mix of unrelated issues. The only clear duplicate pair is the two DeepSpeed ZeRO-3/BERT reports; the rest are at best broad theme matches (docs cleanup, MoE, dtype/config errors) but not the same concrete bug.", + "confidence": 0.85, + "canonical_issue_reason": "issue:43596 is the cleaner canonical report for the ZeRO-3/BERT empty-index failure: it states the general load/init error, while issue:43638 reads like the same bug in a narrower non-pretrained-BERT variant.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43596 is the best global issue anchor because it most clearly captures the shared underlying bug among the only true duplicate pair and is the broader, more reusable description.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44322", + "right": "issue:45464", + "accept": false, + "reason": "Different Qwen3.5 failures: one is a missing config attribute, the other is a streaming chat/completions runtime issue." + }, + { + "left": "issue:45237", + "right": "issue:45685", + "accept": false, + "reason": "Different model/hardware paths; GPT-OSS on AMD GPUs is unrelated to an MPS histogram kernel limitation." + }, + { + "left": "issue:43716", + "right": "issue:45237", + "accept": false, + "reason": "Different bug classes: image preprocessor dtype mismatch versus a GPT-OSS GPU compatibility failure." + }, + { + "left": "issue:43828", + "right": "issue:45237", + "accept": false, + "reason": "Autocast dtype mismatch and AMD GPU failure are not the same concrete problem." + }, + { + "left": "issue:45237", + "right": "issue:45507", + "accept": false, + "reason": "Different models and code paths; this is not one shared underlying defect." + }, + { + "left": "issue:44292", + "right": "issue:45237", + "accept": false, + "reason": "Qwen-3 NVFP4 runtime failure is unrelated to GPT-OSS AMD GPU support." + }, + { + "left": "issue:30277", + "right": "issue:35545", + "accept": false, + "reason": "Collective mismatch under ZeRO-3 and ONNX export failure are distinct issues." + }, + { + "left": "issue:35545", + "right": "issue:42915", + "accept": false, + "reason": "ModernBERT ONNX export and Qwen3 MoE FP8 loading are unrelated change requests/bugs." + }, + { + "left": "issue:35545", + "right": "issue:41720", + "accept": false, + "reason": "ONNX export and CUDA assert with auto device mapping are different failures." + }, + { + "left": "issue:44509", + "right": "issue:44512", + "accept": false, + "reason": "Both are v5 docs cleanups, but they refer to different removed documentation surfaces and are not the same bug." + }, + { + "left": "issue:35545", + "right": "issue:36010", + "accept": false, + "reason": "ModernBERT export error and GenerationMixin import error have no shared concrete code path." + }, + { + "left": "issue:43638", + "right": "issue:44805", + "accept": false, + "reason": "Both are shape/index errors, but the reports point to different symptoms and contexts; not enough to treat as the same bug." + }, + { + "left": "issue:43596", + "right": "issue:44805", + "accept": false, + "reason": "ZeRO-3/BERT empty-index loading and a generic mask-shape mismatch are too different to merge." + }, + { + "left": "issue:44805", + "right": "issue:45161", + "accept": false, + "reason": "A mask/tensor shape error is unrelated to GPT-OSS tensor-parallel failure." + }, + { + "left": "issue:44805", + "right": "issue:44918", + "accept": false, + "reason": "Mask shape mismatch and Qwen3.5 embedding unpacking under TRL are different problems." + }, + { + "left": "issue:44661", + "right": "issue:44805", + "accept": false, + "reason": "Tokenizer mapping helper failure is unrelated to the mask-shape index error." + }, + { + "left": "issue:34634", + "right": "issue:35545", + "accept": false, + "reason": "BarkProcessor voice preset behavior and ModernBERT ONNX export are unrelated." + }, + { + "left": "issue:30277", + "right": "issue:43638", + "accept": false, + "reason": "Both involve DeepSpeed, but one is a collective mismatch and the other is a BERT indexing failure." + }, + { + "left": "issue:43596", + "right": "issue:43638", + "accept": true, + "reason": "Same underlying ZeRO-3/BERT zero-sized indexing bug; 43638 is just a narrower description of the same failure mode." + }, + { + "left": "issue:43975", + "right": "issue:44457", + "accept": false, + "reason": "Detokenization in v5 and LoRA merge/save inconsistency are different model-behavior bugs." + }, + { + "left": "issue:43039", + "right": "issue:45161", + "accept": false, + "reason": "Liger Kernel cross-entropy invocation is unrelated to GPT-OSS MoE tensor-parallel behavior." + }, + { + "left": "issue:44292", + "right": "issue:44457", + "accept": false, + "reason": "Qwen NVFP4 runtime issues and LoRA merge/save output mismatch are not the same bug." + }, + { + "left": "issue:34928", + "right": "issue:44805", + "accept": false, + "reason": "Activation-checkpointing/FSDP recomputation mismatch is a different failure class from the mask-shape index error." + }, + { + "left": "issue:24643", + "right": "issue:30277", + "accept": false, + "reason": "Both mention DeepSpeed, but the runtime weight-dimension error and collective mismatch are different code-path failures." + }, + { + "left": "issue:43638", + "right": "issue:45161", + "accept": false, + "reason": "BERT ZeRO-3 loading failure and GPT-OSS tensor-parallel failure are unrelated." + }, + { + "left": "issue:34928", + "right": "issue:36331", + "accept": false, + "reason": "FSDP checkpointing mismatch and CustomTrainer signature change are unrelated issues." + }, + { + "left": "issue:43596", + "right": "issue:45161", + "accept": false, + "reason": "BERT ZeRO-3 index error is unrelated to GPT-OSS MoE TP behavior." + }, + { + "left": "issue:43716", + "right": "issue:45507", + "accept": false, + "reason": "Image preprocessor dtype mismatch and GraniteMoEHybrid invalid method are different problems." + }, + { + "left": "issue:39290", + "right": "issue:45081", + "accept": false, + "reason": "Gemma3/vLLM sliding-window config error is unrelated to tokenizer regex patch crashes." + }, + { + "left": "issue:43828", + "right": "issue:45161", + "accept": false, + "reason": "Autocast dtype mismatch and GPT-OSS TP failure are different concrete bugs." + }, + { + "left": "issue:44918", + "right": "issue:45161", + "accept": false, + "reason": "TRL embedding unpacking and GPT-OSS tensor-parallel failure do not share the same bug." + }, + { + "left": "issue:43039", + "right": "issue:43828", + "accept": false, + "reason": "Both are runtime issues, but one is Liger Kernel cross-entropy selection and the other is an autocast dtype mismatch." + }, + { + "left": "issue:41762", + "right": "issue:43638", + "accept": false, + "reason": "Gemma3 ZeRO-3 out-of-bounds and BERT ZeRO-3 out-of-bounds are similar in shape but not clearly the same bug." + }, + { + "left": "issue:43827", + "right": "issue:44512", + "accept": false, + "reason": "Both are docs regressions after v5 removals, but they target different documentation sections and are not the same change." + }, + { + "left": "issue:43716", + "right": "issue:44560", + "accept": false, + "reason": "Mistral image dtype mismatch and Qwen3-vl video StopIteration are unrelated multimodal failures." + }, + { + "left": "issue:43828", + "right": "issue:45685", + "accept": false, + "reason": "Autocast dtype mismatch is not the same as an unimplemented MPS histogram kernel." + }, + { + "left": "issue:44560", + "right": "issue:45685", + "accept": false, + "reason": "Video processing StopIteration and MPS Int histogram kernel error are unrelated." + }, + { + "left": "issue:43866", + "right": "issue:44863", + "accept": false, + "reason": "Checkpoint corruption and NemotronH checkpoint loading are different issues." + }, + { + "left": "issue:43854", + "right": "issue:43866", + "accept": false, + "reason": "Unit-test model loading failure and corrupted checkpoint are different root causes." + }, + { + "left": "issue:39290", + "right": "issue:43531", + "accept": false, + "reason": "Gemma3 sliding-window attribute error and Qwen3-MoE sliding_window issue are related only by name, not by the same bug." + }, + { + "left": "issue:41093", + "right": "issue:43638", + "accept": false, + "reason": "Mask/index shape mismatch during one workflow is not the same as ZeRO-3 BERT empty-index loading failure." + }, + { + "left": "issue:43425", + "right": "issue:45070", + "accept": false, + "reason": "Torch version incompatibility and pydantic PretrainedConfig field breakage are unrelated." + }, + { + "left": "issue:43596", + "right": "issue:45800", + "accept": false, + "reason": "Both mention DeepSpeed-style incompatibilities in a broad sense, but the underlying bugs are different." + }, + { + "left": "issue:45362", + "right": "issue:45464", + "accept": false, + "reason": "Qwen3.5 chat crash and streaming inference failure are different runtime issues." + }, + { + "left": "issue:43792", + "right": "issue:45070", + "accept": false, + "reason": "Whisper load failure and pydantic config breakage are unrelated." + }, + { + "left": "issue:43404", + "right": "issue:43866", + "accept": false, + "reason": "Mistral3 lm_head tying and Ovis2 checkpoint corruption are different defects." + }, + { + "left": "issue:45464", + "right": "issue:45685", + "accept": false, + "reason": "Qwen3.5 streaming inference failure and MPS histogram kernel limitation are unrelated." + }, + { + "left": "issue:43541", + "right": "issue:45685", + "accept": false, + "reason": "TorchDynamo grouped_mm tracing failure and MPS histogram kernel error are different problems." + }, + { + "left": "issue:43541", + "right": "issue:43828", + "accept": false, + "reason": "Grouped_mm tracing failure and autocast dtype mismatch are not the same code-path bug." + }, + { + "left": "issue:36010", + "right": "issue:39290", + "accept": false, + "reason": "GenerationMixin import failure and Gemma3 config attribute error are unrelated." + }, + { + "left": "issue:44841", + "right": "issue:45084", + "accept": false, + "reason": "Voxtral processor failure and non-template-node compile error do not share a concrete bug." + }, + { + "left": "issue:41093", + "right": "issue:41762", + "accept": false, + "reason": "Both are index/shape errors, but they involve different models and different failure points, so they should not be merged." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet, and the only accepted duplicate pair is handled conservatively. The broader rejection rationale for the remaining pairs stays within what the titles/evidence support." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 2, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-387", + "nodes": [ + "issue:39186", + "issue:39290", + "issue:39692", + "issue:40990", + "issue:41093", + "issue:42915", + "issue:43054", + "issue:43296", + "issue:43366", + "issue:43404", + "issue:43531", + "issue:43541", + "issue:43572", + "issue:43596", + "issue:43644", + "issue:43645", + "issue:43716", + "issue:43742", + "issue:43828", + "issue:43873", + "issue:43950", + "issue:43975", + "issue:43994", + "issue:44291", + "issue:44292", + "issue:44360", + "issue:44387", + "issue:44492", + "issue:44512", + "issue:44530", + "issue:44534", + "issue:44589", + "issue:44863", + "issue:44918", + "issue:44928", + "issue:44960", + "issue:45005", + "issue:45020", + "issue:45070", + "issue:45161", + "issue:45399", + "issue:45464", + "issue:45507", + "issue:45542", + "issue:45632", + "issue:45685", + "issue:45698", + "issue:45800" + ], + "soft_pairs": [ + "issue:43296|issue:43366", + "issue:43572|issue:45070", + "issue:44387|issue:45005", + "issue:43541|issue:45161", + "issue:43828|issue:45464", + "issue:44291|issue:45070", + "issue:45020|issue:45800", + "issue:43366|issue:45464", + "issue:43531|issue:45070", + "issue:39290|issue:43742", + "issue:43541|issue:44589", + "issue:41093|issue:43596", + "issue:43054|issue:43994", + "issue:45542|issue:45800", + "issue:43716|issue:44918", + "issue:44918|issue:45685", + "issue:45632|issue:45698", + "issue:43404|issue:45005", + "issue:43645|issue:45399", + "issue:40990|issue:44960", + "issue:44530|issue:45005", + "issue:42915|issue:45399", + "issue:44360|issue:44512", + "issue:44863|issue:45005", + "issue:43873|issue:45005", + "issue:39186|issue:44918", + "issue:44492|issue:44512", + "issue:44918|issue:45507", + "issue:43975|issue:44292", + "issue:43950|issue:44534", + "issue:43644|issue:43950", + "issue:44292|issue:45507", + "issue:39692|issue:43054", + "issue:43828|issue:44928" + ], + "prepared_review_unit_hash": "d3c933a61214168108948c3e21d557c1e869298c7e92e416aa804ae5edaceae1", + "component_budget": { + "node_count": 387, + "item_count": 387, + "soft_pair_count": 1455, + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 34, + "serialized_chars": 37554, + "estimated_input_tokens": 9389, + "estimated_eval_tokens": 19034 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Mostly a grab bag of Transformers/vLLM regressions, but a few true duplicate clusters stand out. The clearest one is the non-persistent-buffer regression around `from_pretrained()`/v5 (`43950`, `44534`, with `43644` feeding into the same bug). `43054`/`43994` also looks like the same SigLIP2 model-output problem. The `trust_remote_code`/custom-module pair `45632`/`45698` is related thematically, but not strong enough to treat as the same bug.", + "confidence": 0.7, + "canonical_issue_reason": "issue:43950 is the cleanest canonical issue: it describes the core v5 regression around non-persistent buffers clearly and has an obvious near-duplicate in issue:44534.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43950 is the strongest representative because it captures the concrete regression most directly and is supported by the closest duplicate evidence.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43296", + "right": "issue:43366", + "accept": false, + "reason": "Different problems: PaddleOCR-VL/vLLM loading vs GGUF gpt-oss architecture support." + }, + { + "left": "issue:43572", + "right": "issue:45070", + "accept": false, + "reason": "Both are config regressions, but missing pad_token_idx and pydantic PretrainedConfig breakage are different bugs." + }, + { + "left": "issue:44387", + "right": "issue:45005", + "accept": false, + "reason": "Different symptoms and code paths: int4 quantization memory growth vs tied-weights behavior." + }, + { + "left": "issue:43541", + "right": "issue:45161", + "accept": false, + "reason": "Both mention MoE/GPT-OSS-like models, but grouped_mm tracing failure is not the same as tensor-parallelism not working." + }, + { + "left": "issue:43828", + "right": "issue:45464", + "accept": false, + "reason": "Unrelated failures: autocast dtype mismatch in Phi-tiny-MoE vs streaming chat/completions failure in Qwen3.5-0.8B." + }, + { + "left": "issue:44291", + "right": "issue:45070", + "accept": false, + "reason": "Different regressions: init_empty_weights/_is_hf_initialized issue vs pydantic PretrainedConfig field break." + }, + { + "left": "issue:45020", + "right": "issue:45800", + "accept": false, + "reason": "Too broad and different: remote_code breakage vs torch 2.4.1/transformers 5.8.0 incompatibility." + }, + { + "left": "issue:43366", + "right": "issue:45464", + "accept": false, + "reason": "Different scopes: GGUF gpt-oss architecture support vs Qwen3.5 streaming API failure." + }, + { + "left": "issue:43531", + "right": "issue:45070", + "accept": false, + "reason": "Qwen3-MoE sliding_window issue is unrelated to the PretrainedConfig/pydantic regression." + }, + { + "left": "issue:39290", + "right": "issue:43742", + "accept": false, + "reason": "Both are model-loading errors, but Gemma3 sliding-window config and MobileLLM key error are distinct bugs." + }, + { + "left": "issue:43541", + "right": "issue:44589", + "accept": false, + "reason": "Different concrete failures: torch dynamo grouped_mm tracing vs missing Float8 storage object." + }, + { + "left": "issue:41093", + "right": "issue:43596", + "accept": false, + "reason": "Different indexing errors in different contexts; no shared code-path evidence." + }, + { + "left": "issue:43054", + "right": "issue:43994", + "accept": true, + "reason": "Both report SigLIP2 producing bad/nonsensical outputs, likely the same model/config bug." + }, + { + "left": "issue:45542", + "right": "issue:45800", + "accept": false, + "reason": "TensorBoard/TensorFlow backend detection bug is unrelated to torch/transformers version incompatibility." + }, + { + "left": "issue:43716", + "right": "issue:44918", + "accept": false, + "reason": "Different model families and failures: Mistral-3 image preprocessor dtype mismatch vs Qwen3.5/TRL embedding unpacking." + }, + { + "left": "issue:44918", + "right": "issue:45685", + "accept": false, + "reason": "Qwen3.5/TRL embedding issue and MPS histogram error are unrelated." + }, + { + "left": "issue:45632", + "right": "issue:45698", + "accept": false, + "reason": "Both are about custom/remote-code loading, but one is a cache-path collision for local models sharing a leaf directory and the other is a wrong custom module being loaded after save_pretrained; related area, not the same concrete bug." + }, + { + "left": "issue:43404", + "right": "issue:45005", + "accept": false, + "reason": "Both involve tied weights, but they target different models and likely different regressions." + }, + { + "left": "issue:43645", + "right": "issue:45399", + "accept": false, + "reason": "Jupyter/custom-model initialization bug is unrelated to flash-attn2 fallback gating." + }, + { + "left": "issue:40990", + "right": "issue:44960", + "accept": false, + "reason": "Perplexity anomaly on gpt-oss-20b is not the same as the generic GLM5 issue." + }, + { + "left": "issue:44530", + "right": "issue:45005", + "accept": false, + "reason": "PagedAttentionCache linear_attention failure is unrelated to tied-weight issues." + }, + { + "left": "issue:42915", + "right": "issue:45399", + "accept": false, + "reason": "Qwen3-MoE FineGrainedFP8Config failure is unrelated to flash-attn2 fallback checks." + }, + { + "left": "issue:44360", + "right": "issue:44512", + "accept": false, + "reason": "DSA indexer/ReLU discussion is unrelated to the removed run-command docs issue." + }, + { + "left": "issue:44863", + "right": "issue:45005", + "accept": false, + "reason": "NemotronH checkpoint loading is unrelated to tied weights on translation models." + }, + { + "left": "issue:43873", + "right": "issue:45005", + "accept": false, + "reason": "Quantization offloading behavior is different from tied-weight regressions." + }, + { + "left": "issue:39186", + "right": "issue:44918", + "accept": false, + "reason": "FSDP weight-dimension error is unrelated to Qwen3.5/TRL embedding unpacking." + }, + { + "left": "issue:44492", + "right": "issue:44512", + "accept": false, + "reason": "A cache-strategy typo is unrelated to docs mentioning a removed command." + }, + { + "left": "issue:44918", + "right": "issue:45507", + "accept": false, + "reason": "Different model families and failure modes; no common code-path evidence." + }, + { + "left": "issue:43975", + "right": "issue:44292", + "accept": false, + "reason": "Incorrect detokenization and NVFP4 loading errors are distinct issues." + }, + { + "left": "issue:43950", + "right": "issue:44534", + "accept": true, + "reason": "Near-exact duplicate regression: both say v5 fills/corrupts non-persistent buffers with junk." + }, + { + "left": "issue:43644", + "right": "issue:43950", + "accept": true, + "reason": "Same underlying regression in from_pretrained corrupting non-persistent buffers." + }, + { + "left": "issue:44292", + "right": "issue:45507", + "accept": false, + "reason": "Qwen-3-8B-NVFP4 loading failure is unrelated to GraniteMoEHybrid invalid method calls." + }, + { + "left": "issue:39692", + "right": "issue:43054", + "accept": false, + "reason": "SigLIP2 docs-example errors are different from the broader SigLIP2 embedding-quality complaint." + }, + { + "left": "issue:43828", + "right": "issue:44928", + "accept": false, + "reason": "Different dtype/NaN problems in different models and training setups; not the same bug." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "The overall summary is mostly grounded, but one accepted soft-edge verdict is too strong for the evidence in the packet. `issue:45632` and `issue:45698` are both about remote-code/custom-module loading, but the titles alone do not establish that they are the same bug or clearly mergeable into one fix. The other accepted duplicates (`43950`/`44534`, `43644`/`43950`, and likely `43054`/`43994`) are fine, but this custom-module pair makes the cluster too speculative as written." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 3, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-387", + "nodes": [ + "issue:36331", + "issue:38175", + "issue:41553", + "issue:41628", + "issue:41762", + "issue:42491", + "issue:42915", + "issue:42947", + "issue:43054", + "issue:43257", + "issue:43404", + "issue:43493", + "issue:43643", + "issue:43742", + "issue:43782", + "issue:43792", + "issue:43824", + "issue:43827", + "issue:43828", + "issue:43856", + "issue:43866", + "issue:43872", + "issue:43881", + "issue:43883", + "issue:43940", + "issue:43950", + "issue:43975", + "issue:44220", + "issue:44291", + "issue:44368", + "issue:44387", + "issue:44403", + "issue:44451", + "issue:44488", + "issue:44509", + "issue:44589", + "issue:44661", + "issue:44960", + "issue:44991", + "issue:45020", + "issue:45081", + "issue:45237", + "issue:45313", + "issue:45356", + "issue:45399", + "issue:45520", + "issue:45701", + "issue:45800" + ], + "soft_pairs": [ + "issue:43866|issue:44403", + "issue:36331|issue:41762", + "issue:41628|issue:45800", + "issue:41553|issue:43883", + "issue:43872|issue:44589", + "issue:38175|issue:43054", + "issue:41553|issue:43881", + "issue:43828|issue:45800", + "issue:43054|issue:45701", + "issue:42947|issue:44387", + "issue:44403|issue:44451", + "issue:43881|issue:43883", + "issue:43643|issue:45020", + "issue:43054|issue:43493", + "issue:43975|issue:45520", + "issue:43975|issue:45237", + "issue:44403|issue:44488", + "issue:42947|issue:43856", + "issue:42491|issue:44368", + "issue:41762|issue:42915", + "issue:43975|issue:44451", + "issue:44661|issue:45020", + "issue:43824|issue:43872", + "issue:44991|issue:45399", + "issue:45020|issue:45356", + "issue:43782|issue:44960", + "issue:45081|issue:45399", + "issue:43827|issue:44509", + "issue:43792|issue:44220", + "issue:43404|issue:43950", + "issue:43742|issue:45399", + "issue:42915|issue:45020", + "issue:43257|issue:43866", + "issue:41553|issue:44991", + "issue:43742|issue:45020", + "issue:43940|issue:45313", + "issue:43872|issue:44291" + ], + "prepared_review_unit_hash": "83e778eba049c578dbea81bf7bea6b6393572d3cb393e3e36f711f7ec1128f25", + "component_budget": { + "node_count": 387, + "item_count": 387, + "soft_pair_count": 1455, + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 37, + "serialized_chars": 38719, + "estimated_input_tokens": 9680, + "estimated_eval_tokens": 19616 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a mix of unrelated model-loading/runtime regressions and docs issues, not one duplicate set. I only see two clear duplicate pairs: the v5 pipeline docs cleanup, and the unexpected `_is_hf_initialized` TypeError during loading/init_empty_weights.", + "confidence": 0.82, + "canonical_issue_reason": "issue:44291 is the clearest concrete bug report in the cluster and the exact duplicate target for issue:43872; it precisely pins the shared `_is_hf_initialized` TypeError during model loading.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45020 is the strongest overall umbrella for the cluster because it captures the broader family of recent remote_code / model-loading regressions that many of the other reports resemble, even though most are not true duplicates.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43866", + "right": "issue:44403", + "accept": false, + "reason": "Different problems: corrupted checkpoint vs generic loading noise; no shared code-path evidence." + }, + { + "left": "issue:36331", + "right": "issue:41762", + "accept": false, + "reason": "Trainer API signature error is unrelated to a ZeRO-3 model-loading index failure." + }, + { + "left": "issue:41628", + "right": "issue:45800", + "accept": false, + "reason": "Import/export API issue is not the same as a torch/transformers version incompatibility report." + }, + { + "left": "issue:41553", + "right": "issue:43883", + "accept": false, + "reason": "Bad tokenizer error message and a missing model attribute are different concrete bugs." + }, + { + "left": "issue:43872", + "right": "issue:44589", + "accept": false, + "reason": "Unexpected `_is_hf_initialized` argument is unrelated to the float8 storage lookup failure." + }, + { + "left": "issue:38175", + "right": "issue:43054", + "accept": false, + "reason": "Both involve SigLIP2, but one is zero probabilities and the other is embedding quality; too little evidence of the same bug." + }, + { + "left": "issue:41553", + "right": "issue:43881", + "accept": false, + "reason": "Voxtral AutoTokenizer loading error and glm-4v-9b loading failure are model-specific and distinct." + }, + { + "left": "issue:43828", + "right": "issue:45800", + "accept": false, + "reason": "Autocast dtype mismatch is not the same as a torch/transformers version compatibility issue." + }, + { + "left": "issue:43054", + "right": "issue:45701", + "accept": false, + "reason": "Embedding-quality complaint and tokenization-change report are different failure modes." + }, + { + "left": "issue:42947", + "right": "issue:44387", + "accept": false, + "reason": "LoRA gradient-checkpointing behavior is unrelated to int4 CUDA memory growth." + }, + { + "left": "issue:44403", + "right": "issue:44451", + "accept": false, + "reason": "Generic loading noise and a specific ScandiBERT load failure are not the same bug." + }, + { + "left": "issue:43881", + "right": "issue:43883", + "accept": false, + "reason": "Different models and different failure symptoms; no concrete overlap." + }, + { + "left": "issue:43643", + "right": "issue:45020", + "accept": false, + "reason": "Both mention remote_code, but one is missing fields in AutoConfig and the other is a broad breakage report; not the same bug." + }, + { + "left": "issue:43054", + "right": "issue:43493", + "accept": false, + "reason": "Strongly related SigLIP2 reports, but one is a quality regression and the other is an HF-vs-JAX discrepancy; not proven identical." + }, + { + "left": "issue:43975", + "right": "issue:45520", + "accept": false, + "reason": "Detokenization regression is unrelated to a flash_attn import_utils KeyError." + }, + { + "left": "issue:43975", + "right": "issue:45237", + "accept": false, + "reason": "Tokenizer detokenization bug and AMD GPU runtime failure are different issues." + }, + { + "left": "issue:44403", + "right": "issue:44488", + "accept": false, + "reason": "Both are model-load complaints, but they point to different models and likely different root causes." + }, + { + "left": "issue:42947", + "right": "issue:43856", + "accept": false, + "reason": "Gradient checkpointing/LoRA behavior is not the same as Qwen3 MoE training memory usage." + }, + { + "left": "issue:42491", + "right": "issue:44368", + "accept": false, + "reason": "Cross-version LoRA compatibility issue and a tie_word_embeddings warning are not the same underlying bug." + }, + { + "left": "issue:41762", + "right": "issue:42915", + "accept": false, + "reason": "Both are Qwen-related loading failures, but the specific failure modes and model variants differ too much." + }, + { + "left": "issue:43975", + "right": "issue:44451", + "accept": false, + "reason": "Different model/tokenization regressions with no shared concrete code-path." + }, + { + "left": "issue:44661", + "right": "issue:45020", + "accept": false, + "reason": "Tokenizer-mapping failure in add-new-model-like is not the same as broad remote_code loading regressions." + }, + { + "left": "issue:43824", + "right": "issue:43872", + "accept": false, + "reason": "ImportError for Qwen2.5-VL and `_is_hf_initialized` TypeError are unrelated." + }, + { + "left": "issue:44991", + "right": "issue:45399", + "accept": false, + "reason": "Tokenizer loading regression and flash-attn fallback blocking are different issues." + }, + { + "left": "issue:45020", + "right": "issue:45356", + "accept": false, + "reason": "Broad remote_code breakage and a Kimi tokenizer codec/warning regression are not the same bug." + }, + { + "left": "issue:43782", + "right": "issue:44960", + "accept": false, + "reason": "Different model families and different loading errors." + }, + { + "left": "issue:45081", + "right": "issue:45399", + "accept": false, + "reason": "Mistral regex patch crash is unrelated to flash-attn fallback logic." + }, + { + "left": "issue:43827", + "right": "issue:44509", + "accept": true, + "reason": "Same docs regression: both say v5 removed pipeline tasks are still referenced in docs, just with different wording." + }, + { + "left": "issue:43792", + "right": "issue:44220", + "accept": false, + "reason": "Whisper runtime failure is too broad to equate with the specific `_torch_extract_fbank_features()` issue." + }, + { + "left": "issue:43404", + "right": "issue:43950", + "accept": false, + "reason": "Weight-tying bug in Mistral3 and non-persistent buffer corruption are distinct model/state-loading problems." + }, + { + "left": "issue:43742", + "right": "issue:45399", + "accept": false, + "reason": "MobileLLM-125M key error and flash-attn fallback blocking are not the same failure." + }, + { + "left": "issue:43940", + "right": "issue:45313", + "accept": false, + "reason": "Both are ZeRO-3 loading failures, but they affect different Qwen generations and the concrete symptoms differ." + }, + { + "left": "issue:43872", + "right": "issue:44291", + "accept": true, + "reason": "These describe the same `_is_hf_initialized`/init_empty_weights TypeError during loading; 44291 is the clearer full report." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "Grounded and conservative overall. The two accepted duplicate pairs are supported by the packet, and the other soft-edge rejections avoid overclaiming despite surface similarity. The cluster summary does not overstate duplication." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 4, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-387", + "nodes": [ + "issue:24643", + "issue:30064", + "issue:36010", + "issue:36331", + "issue:38175", + "issue:39692", + "issue:41093", + "issue:41553", + "issue:42617", + "issue:42915", + "issue:43257", + "issue:43317", + "issue:43366", + "issue:43531", + "issue:43550", + "issue:43596", + "issue:43638", + "issue:43645", + "issue:43646", + "issue:43824", + "issue:43828", + "issue:43854", + "issue:43872", + "issue:43950", + "issue:43957", + "issue:43994", + "issue:44155", + "issue:44479", + "issue:44488", + "issue:44530", + "issue:44560", + "issue:44589", + "issue:44661", + "issue:44756", + "issue:44805", + "issue:44863", + "issue:44877", + "issue:44910", + "issue:44918", + "issue:44928", + "issue:45030", + "issue:45305", + "issue:45313", + "issue:45325", + "issue:45357", + "issue:45538", + "issue:45584", + "issue:45588" + ], + "soft_pairs": [ + "issue:43854|issue:44863", + "issue:42915|issue:43872", + "issue:44928|issue:45588", + "issue:41553|issue:44488", + "issue:43257|issue:45305", + "issue:36331|issue:41093", + "issue:43872|issue:45538", + "issue:44530|issue:44863", + "issue:41093|issue:44805", + "issue:44479|issue:44560", + "issue:24643|issue:30064", + "issue:43366|issue:43828", + "issue:43366|issue:43531", + "issue:43550|issue:44910", + "issue:43645|issue:43646", + "issue:43366|issue:44589", + "issue:43950|issue:45325", + "issue:44560|issue:44918", + "issue:42617|issue:43366", + "issue:42915|issue:43957", + "issue:43824|issue:43957", + "issue:36010|issue:42915", + "issue:38175|issue:43994", + "issue:43646|issue:43950", + "issue:44155|issue:45584", + "issue:44877|issue:45030", + "issue:44560|issue:44805", + "issue:43638|issue:44661", + "issue:43596|issue:44661", + "issue:43317|issue:44756", + "issue:38175|issue:39692", + "issue:45313|issue:45357" + ], + "prepared_review_unit_hash": "cc27aef110d8c6c96dfb6199d1efa7db385bd17e45f4ad2ba4f1dc6a621c081f", + "component_budget": { + "node_count": 387, + "item_count": 387, + "soft_pair_count": 1455, + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 32, + "serialized_chars": 37467, + "estimated_input_tokens": 9367, + "estimated_eval_tokens": 18990 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This set is mostly a grab-bag of unrelated bug reports. Only three soft pairs look like true duplicates: the custom-model-init regression in Transformers 5.0.0, the SigLIP2 bad-output reports, and the mask/tensor shape mismatch error.", + "confidence": 0.83, + "canonical_issue_reason": "No single issue represents the whole list, but 43646 is the best canonical anchor among the confirmed duplicates: it states the broad Transformers 5.0 custom model initialization regression, while 43645 is just a notebook-specific repro.", + "canonical_pr_reason": null, + "best_issue_reason": "43646 is the strongest overall issue to keep as the representative issue: it is concise, root-cause oriented, and broad enough to subsume the narrower 43645 report.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43854", + "right": "issue:44863", + "accept": false, + "reason": "Different model families and failure modes; both are loading bugs, but not the same underlying defect." + }, + { + "left": "issue:42915", + "right": "issue:43872", + "accept": false, + "reason": "Different quantization-related regressions with different error surfaces and code paths." + }, + { + "left": "issue:44928", + "right": "issue:45588", + "accept": false, + "reason": "Both touch attention-related code, but the symptoms and underlying causes are different." + }, + { + "left": "issue:41553", + "right": "issue:44488", + "accept": false, + "reason": "Both are tokenizer-loading complaints, but they concern different models and likely different root causes." + }, + { + "left": "issue:43257", + "right": "issue:45305", + "accept": false, + "reason": "Distinct DeepSpeed issues: one is MOE weight conversion, the other is gradient averaging with GAS." + }, + { + "left": "issue:36331", + "right": "issue:41093", + "accept": false, + "reason": "Unrelated trainer API breakage versus tensor mask shape mismatch." + }, + { + "left": "issue:43872", + "right": "issue:45538", + "accept": false, + "reason": "Different bug classes entirely: bitsandbytes parameter incompatibility vs tokenizer max-length handling." + }, + { + "left": "issue:44530", + "right": "issue:44863", + "accept": false, + "reason": "Different models and different failures; no shared concrete bug." + }, + { + "left": "issue:41093", + "right": "issue:44805", + "accept": true, + "reason": "Same mask/index shape-mismatch error; 44805 is essentially the same bug with generalized placeholders." + }, + { + "left": "issue:44479", + "right": "issue:44560", + "accept": false, + "reason": "Both are video-related, but the reported failures and model scopes differ enough to avoid merging." + }, + { + "left": "issue:24643", + "right": "issue:30064", + "accept": false, + "reason": "Completely different subsystems and symptoms." + }, + { + "left": "issue:43366", + "right": "issue:43828", + "accept": false, + "reason": "GGUF architecture support and autocast dtype mismatch are unrelated problems." + }, + { + "left": "issue:43366", + "right": "issue:43531", + "accept": false, + "reason": "Different Qwen3-MoE and GGUF-related issues; same general area, not the same bug." + }, + { + "left": "issue:43550", + "right": "issue:44910", + "accept": false, + "reason": "Both involve attention kernels, but the concrete model, trigger, and failure mode differ." + }, + { + "left": "issue:43645", + "right": "issue:43646", + "accept": true, + "reason": "Same Transformers 5.0 custom-model-initialization regression; 43645 is a notebook-specific reproduction of 43646." + }, + { + "left": "issue:43366", + "right": "issue:44589", + "accept": false, + "reason": "Unrelated GGUF/model-architecture support issue vs missing Float8 storage type." + }, + { + "left": "issue:43950", + "right": "issue:45325", + "accept": false, + "reason": "Different regressions: non-persistent buffer corruption versus Qwen2.5-VL rope indexing." + }, + { + "left": "issue:44560", + "right": "issue:44918", + "accept": false, + "reason": "Different Qwen3.5 problems with different reproduction paths and symptoms." + }, + { + "left": "issue:42617", + "right": "issue:43366", + "accept": false, + "reason": "Unrelated runner failure versus GGUF architecture support." + }, + { + "left": "issue:42915", + "right": "issue:43957", + "accept": false, + "reason": "Different loading regressions: FineGrainedFP8Config versus meta-device initialization." + }, + { + "left": "issue:43824", + "right": "issue:43957", + "accept": false, + "reason": "Import error for a missing class is not the same as the meta-device loading regression." + }, + { + "left": "issue:36010", + "right": "issue:42915", + "accept": false, + "reason": "Completely different import-path regression versus quantization config failure." + }, + { + "left": "issue:38175", + "right": "issue:43994", + "accept": true, + "reason": "Same SigLIP2 model producing invalid outputs; these are two reports of the same underlying inference problem." + }, + { + "left": "issue:43646", + "right": "issue:43950", + "accept": false, + "reason": "Custom model initialization regression and non-persistent buffer corruption are separate issues." + }, + { + "left": "issue:44155", + "right": "issue:45584", + "accept": false, + "reason": "Different models, different workflows, and different failure mechanisms." + }, + { + "left": "issue:44877", + "right": "issue:45030", + "accept": false, + "reason": "Both are config-validation related, but they affect different models and are not the same bug." + }, + { + "left": "issue:44560", + "right": "issue:44805", + "accept": false, + "reason": "Unrelated video StopIteration bug versus mask/tensor shape mismatch." + }, + { + "left": "issue:43638", + "right": "issue:44661", + "accept": false, + "reason": "Different failure modes: DeepSpeed/BERT zero3 init versus tokenizer-mapping-name handling." + }, + { + "left": "issue:43596", + "right": "issue:44661", + "accept": false, + "reason": "Same as above; not the same underlying defect." + }, + { + "left": "issue:43317", + "right": "issue:44756", + "accept": false, + "reason": "Different memory/loading problems with different triggers and remedies." + }, + { + "left": "issue:38175", + "right": "issue:39692", + "accept": false, + "reason": "Same model family, but this is a docs/example and quantization failure, not the same bug as the bad-output report." + }, + { + "left": "issue:45313", + "right": "issue:45357", + "accept": false, + "reason": "Different Qwen3.5 regressions: ZeRO-3 weight loading versus save_pretrained key serialization." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "Grounded overall. The three accepted soft pairs are defensible from the packet, and the many rejections stay conservative by avoiding broad same-bug claims across different models/subsystems. The only slightly borderline call is the SigLIP2 pair, but it is still supportable from the titles alone, so the cluster summary remains precise." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 5, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-387", + "nodes": [ + "issue:36010", + "issue:39404", + "issue:39692", + "issue:41720", + "issue:42491", + "issue:42915", + "issue:42947", + "issue:43257", + "issue:43335", + "issue:43404", + "issue:43425", + "issue:43454", + "issue:43644", + "issue:43645", + "issue:43708", + "issue:43720", + "issue:43854", + "issue:43856", + "issue:43873", + "issue:43931", + "issue:43957", + "issue:43994", + "issue:44075", + "issue:44164", + "issue:44291", + "issue:44387", + "issue:44402", + "issue:44451", + "issue:44488", + "issue:44534", + "issue:44661", + "issue:44683", + "issue:44756", + "issue:44910", + "issue:44960", + "issue:44977", + "issue:44991", + "issue:45216", + "issue:45276", + "issue:45292", + "issue:45335", + "issue:45362", + "issue:45406", + "issue:45588", + "issue:45663", + "issue:45704", + "issue:45759" + ], + "soft_pairs": [ + "issue:43644|issue:44534", + "issue:44960|issue:44977", + "issue:44451|issue:44488", + "issue:42491|issue:43720", + "issue:43708|issue:44075", + "issue:43854|issue:43931", + "issue:41720|issue:42915", + "issue:42491|issue:45276", + "issue:43856|issue:43873", + "issue:43257|issue:43854", + "issue:39692|issue:43994", + "issue:42915|issue:44756", + "issue:42491|issue:43931", + "issue:43856|issue:44387", + "issue:44661|issue:44683", + "issue:43644|issue:43645", + "issue:44910|issue:45588", + "issue:45362|issue:45406", + "issue:42491|issue:45292", + "issue:44488|issue:44991", + "issue:36010|issue:39404", + "issue:43404|issue:43454", + "issue:45276|issue:45335", + "issue:44164|issue:44291", + "issue:44402|issue:44488", + "issue:39692|issue:44960", + "issue:43425|issue:43957", + "issue:43335|issue:45663", + "issue:42947|issue:45704", + "issue:45663|issue:45759", + "issue:43873|issue:44387", + "issue:43854|issue:45216", + "issue:44387|issue:44756", + "issue:44291|issue:44756" + ], + "prepared_review_unit_hash": "71bd6f313bb62a9786dddfbb9ef4945e54d03427c6c946d0476213906830759f", + "component_budget": { + "node_count": 387, + "item_count": 387, + "soft_pair_count": 1455, + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 34, + "serialized_chars": 37322, + "estimated_input_tokens": 9331, + "estimated_eval_tokens": 18918 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is mostly a set of unrelated Transformers issues. The strongest duplicate signal is issue:43644 \u2194 issue:44534, which appear to report the same Transformers v5 non-persistent-buffer regression; this should be treated as a likely/strong soft duplicate rather than a confirmed duplicate. The rest of the soft pairs are not safe to merge as duplicates.", + "confidence": 0.87, + "canonical_issue_reason": "issue:43644 is the best anchor because it describes a concrete Transformers 5.0.0 regression with clear symptoms, and issue:44534 is a strong soft duplicate candidate for the same non-persistent-buffer bug.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43644 is the most representative and actionable issue in the cluster: it is specific, reproducible, and central to the only clear duplicate pair.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43644", + "right": "issue:44534", + "accept": true, + "reason": "Strong soft duplicate candidate: both titles describe the same Transformers v5 bug where non-persistent buffers are filled with junk." + }, + { + "left": "issue:44960", + "right": "issue:44977", + "accept": false, + "reason": "Both are Qwen-related generation/flash-attention failures, but they describe different models and different failure modes." + }, + { + "left": "issue:44451", + "right": "issue:44488", + "accept": false, + "reason": "Both are model-loading problems, but they involve different checkpoints and no shared concrete bug is evident." + }, + { + "left": "issue:42491", + "right": "issue:43720", + "accept": false, + "reason": "Different subsystems and symptoms: Qwen3 MoE LoRA incompatibility vs BitNet packed-weight unpacking during accelerate loading." + }, + { + "left": "issue:43708", + "right": "issue:44075", + "accept": false, + "reason": "These are unrelated training/resume max_steps logic and SGD-argument usage bugs." + }, + { + "left": "issue:43854", + "right": "issue:43931", + "accept": false, + "reason": "Different model-loading failures on different models; no clear shared code path." + }, + { + "left": "issue:41720", + "right": "issue:42915", + "accept": false, + "reason": "Both mention Qwen3-family models, but one is device-mapping/cuda assert and the other is FineGrainedFP8Config; not the same bug." + }, + { + "left": "issue:42491", + "right": "issue:45276", + "accept": false, + "reason": "LoRA/model-compatibility issue vs resize_token_embeddings propagation bug; different code paths." + }, + { + "left": "issue:43856", + "right": "issue:43873", + "accept": false, + "reason": "Both are memory-related, but one is Qwen3 MoE training inefficiency and the other is offloading/quantization behavior." + }, + { + "left": "issue:43257", + "right": "issue:43854", + "accept": false, + "reason": "Different model families and different loading paths; not the same underlying bug." + }, + { + "left": "issue:39692", + "right": "issue:43994", + "accept": false, + "reason": "Both involve SigLIP2, but the reported failures are different: docs/example mismatch vs nonsensical outputs with AutoModel/pipeline." + }, + { + "left": "issue:42915", + "right": "issue:44756", + "accept": false, + "reason": "Unrelated issues: Qwen3Moe FP8 failure vs Strix Halo mmap/OOM." + }, + { + "left": "issue:42491", + "right": "issue:43931", + "accept": false, + "reason": "Different loading failures with different models and symptoms." + }, + { + "left": "issue:43856", + "right": "issue:44387", + "accept": false, + "reason": "One is Qwen3 MoE training memory inefficiency, the other is int4 reserved-memory growth causing OOM." + }, + { + "left": "issue:44661", + "right": "issue:44683", + "accept": false, + "reason": "Tokenizer-mapping bug vs compiled flex_attention failure on newer torch; unrelated." + }, + { + "left": "issue:43644", + "right": "issue:43645", + "accept": false, + "reason": "Both are Transformers 5.x regressions, but they affect different behaviors and code paths." + }, + { + "left": "issue:44910", + "right": "issue:45588", + "accept": false, + "reason": "Different flash-attention bugs with different triggers and failure modes." + }, + { + "left": "issue:45362", + "right": "issue:45406", + "accept": false, + "reason": "Different model-specific crashes affecting different components (chat vs serve/processor)." + }, + { + "left": "issue:42491", + "right": "issue:45292", + "accept": false, + "reason": "LoRA/model-compatibility issue vs resize_token_embeddings not updating output embeddings." + }, + { + "left": "issue:44488", + "right": "issue:44991", + "accept": false, + "reason": "Both are tokenizer/loading failures, but for different models and without evidence of a shared root cause." + }, + { + "left": "issue:36010", + "right": "issue:39404", + "accept": false, + "reason": "ImportError for GenerationMixin and Whisper return_language pipeline regression are unrelated." + }, + { + "left": "issue:43404", + "right": "issue:43454", + "accept": false, + "reason": "Both mention lm_head tying, but they affect different model classes/configs and are not the same code path." + }, + { + "left": "issue:45276", + "right": "issue:45335", + "accept": false, + "reason": "Similar resize_token_embeddings wording, but different architectures and affected submodules." + }, + { + "left": "issue:44164", + "right": "issue:44291", + "accept": false, + "reason": "save/from_pretrained extra_state handling and init_empty_weights argument handling are distinct bugs." + }, + { + "left": "issue:44402", + "right": "issue:44488", + "accept": false, + "reason": "Tokenizer vocab-size mismatch vs model loading failure; different symptoms and likely different fixes." + }, + { + "left": "issue:39692", + "right": "issue:44960", + "accept": false, + "reason": "Different models and different errors; no shared concrete bug." + }, + { + "left": "issue:43425", + "right": "issue:43957", + "accept": false, + "reason": "Torch version incompatibility is not the same as meta-device loading breakage in Transformers 5.x." + }, + { + "left": "issue:43335", + "right": "issue:45663", + "accept": false, + "reason": "Sparse-layer config bug vs Gemma-4 FSDP2 attention-state KeyError; unrelated." + }, + { + "left": "issue:42947", + "right": "issue:45704", + "accept": false, + "reason": "Gradient checkpointing with PEFT LoRA is unrelated to the apex RMSNorm memory leak." + }, + { + "left": "issue:45663", + "right": "issue:45759", + "accept": false, + "reason": "Different code paths: Gemma-4 FSDP2 attention state bug vs Qwen multimodal config unwrapping bug." + }, + { + "left": "issue:43873", + "right": "issue:44387", + "accept": false, + "reason": "Both touch memory usage, but the mechanisms and fixes are different." + }, + { + "left": "issue:43854", + "right": "issue:45216", + "accept": false, + "reason": "Different Qwen3-related issues: model loading in tests vs incorrect saved checkpoints after a version change." + }, + { + "left": "issue:44387", + "right": "issue:44756", + "accept": false, + "reason": "Unrelated OOM scenarios: int4 quantization reserved memory vs Strix Halo mmap." + }, + { + "left": "issue:44291", + "right": "issue:44756", + "accept": false, + "reason": "Init-empty-weights argument handling is unrelated to Strix Halo mmap/OOM." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "The cluster-level choice is mostly grounded, and the accepted duplicate (issue:43644 \u2194 issue:44534) looks reasonable from the titles. However, the writeup overstates the evidence by calling that duplicate \u201cconfirmed\u201d; the packet only shows a high-similarity soft pair, not an explicit confirmation. Please soften that language to something like \u201cstrong soft duplicate candidate\u201d or \u201clikely duplicate.\u201d" + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 6, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-387", + "nodes": [ + "issue:39404", + "issue:40990", + "issue:41720", + "issue:42947", + "issue:43257", + "issue:43299", + "issue:43381", + "issue:43425", + "issue:43611", + "issue:43645", + "issue:43646", + "issue:43708", + "issue:43716", + "issue:43828", + "issue:43854", + "issue:43856", + "issue:43901", + "issue:43906", + "issue:43931", + "issue:43957", + "issue:44155", + "issue:44164", + "issue:44265", + "issue:44292", + "issue:44402", + "issue:44451", + "issue:44479", + "issue:44509", + "issue:44530", + "issue:44589", + "issue:44756", + "issue:44805", + "issue:44863", + "issue:44877", + "issue:44898", + "issue:44912", + "issue:44918", + "issue:44928", + "issue:44960", + "issue:45084", + "issue:45276", + "issue:45292", + "issue:45335", + "issue:45588", + "issue:45663", + "issue:45684", + "issue:45698", + "issue:45701" + ], + "soft_pairs": [ + "issue:44292|issue:44912", + "issue:42947|issue:43381", + "issue:45684|issue:45698", + "issue:45276|issue:45292", + "issue:43381|issue:44928", + "issue:44164|issue:44756", + "issue:45588|issue:45663", + "issue:41720|issue:44155", + "issue:43646|issue:43708", + "issue:44479|issue:44877", + "issue:43854|issue:44960", + "issue:43957|issue:44756", + "issue:45292|issue:45335", + "issue:44402|issue:44451", + "issue:43645|issue:44164", + "issue:43828|issue:43957", + "issue:43257|issue:44863", + "issue:43425|issue:43828", + "issue:44589|issue:44756", + "issue:44479|issue:45663", + "issue:39404|issue:45701", + "issue:43906|issue:44530", + "issue:43611|issue:43645", + "issue:43425|issue:43716", + "issue:43931|issue:44863", + "issue:44530|issue:45084", + "issue:43901|issue:44509", + "issue:39404|issue:44912", + "issue:44863|issue:45663", + "issue:44265|issue:44898", + "issue:43299|issue:43931", + "issue:44479|issue:44805", + "issue:43425|issue:44918", + "issue:43856|issue:44756", + "issue:44589|issue:45084", + "issue:40990|issue:42947" + ], + "prepared_review_unit_hash": "7c0f34728ef80cc4f31b35faa8675901804b63deefd9bc39ff6125fe320ea4d6", + "component_budget": { + "node_count": 387, + "item_count": 387, + "soft_pair_count": 1455, + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 36, + "serialized_chars": 38397, + "estimated_input_tokens": 9600, + "estimated_eval_tokens": 19456 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Mostly a heterogeneous issue bucket with a few nearby Transformers v5 loading regressions. Only one soft edge looks like a true duplicate-style pair; the rest are related by subsystem or model family but not the same concrete bug.", + "confidence": 0.57, + "canonical_issue_reason": "Issue 43611 is the best representative of the broad Transformers 5 loading-regression theme: it is generic, concrete, and closer to several nearby loading/init reports than the more model-specific outliers.", + "canonical_pr_reason": null, + "best_issue_reason": "43611 is the strongest single anchor for the cluster because it captures the most reusable underlying failure mode (v5 model-loading breakage) without being overly narrow to one model or environment.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44292", + "right": "issue:44912", + "accept": false, + "reason": "Different models and quantization paths; both are load/quantization regressions, but not the same concrete bug." + }, + { + "left": "issue:42947", + "right": "issue:43381", + "accept": false, + "reason": "Both concern gradient checkpointing, but one is PEFT/LoRA ineffectiveness and the other is eval-mode incompatibility." + }, + { + "left": "issue:45684", + "right": "issue:45698", + "accept": false, + "reason": "Both involve save/from_pretrained workflows, but one is read-only permission propagation and the other is loading the wrong custom module." + }, + { + "left": "issue:45276", + "right": "issue:45292", + "accept": true, + "reason": "Same resize_token_embeddings bug family: both report missing propagation to output embeddings; 45276 is the broader Gemma4 manifestation." + }, + { + "left": "issue:43381", + "right": "issue:44928", + "accept": false, + "reason": "Unrelated failures: eval-mode checkpointing restriction versus Qwen3.5 RLHF NaNs from 3D position_ids/SDPA fallback." + }, + { + "left": "issue:44164", + "right": "issue:44756", + "accept": false, + "reason": "Different areas entirely: extra_state save/load handling versus mmap OOM on Strix Halo." + }, + { + "left": "issue:45588", + "right": "issue:45663", + "accept": false, + "reason": "Both are attention-related, but one is a flash_attention null-aux crash and the other is a Gemma4 FSDP2 shared_kv_states bug." + }, + { + "left": "issue:41720", + "right": "issue:44155", + "accept": false, + "reason": "Different model families and symptoms: Qwen3 auto device map assert versus AudioFlamingo3 batched inference token leakage." + }, + { + "left": "issue:43646", + "right": "issue:43708", + "accept": false, + "reason": "Both mention custom model behavior, but one is initialization breakage and the other is Trainer checkpoint-resume step accounting." + }, + { + "left": "issue:44479", + "right": "issue:44877", + "accept": false, + "reason": "Video-input regression versus strict config loading; different code paths and failure modes." + }, + { + "left": "issue:43854", + "right": "issue:44960", + "accept": false, + "reason": "Model loading in tests versus a generic GLM5 issue; too little evidence of the same underlying bug." + }, + { + "left": "issue:43957", + "right": "issue:44756", + "accept": false, + "reason": "Meta-device loading breakage and mmap OOM are both loading-related but clearly distinct problems." + }, + { + "left": "issue:45292", + "right": "issue:45335", + "accept": false, + "reason": "Related resize_token_embeddings theme, but the affected modules and model families differ enough to avoid merging as one bug." + }, + { + "left": "issue:44402", + "right": "issue:44451", + "accept": false, + "reason": "Different models with similar load failures; no clear shared code-path from the titles alone." + }, + { + "left": "issue:43645", + "right": "issue:44164", + "accept": false, + "reason": "Custom-model Jupyter initialization versus extra_state save/load handling are different bugs." + }, + { + "left": "issue:43828", + "right": "issue:43957", + "accept": false, + "reason": "Different dtype/autocast and meta-device loading issues; both are compatibility bugs but not the same one." + }, + { + "left": "issue:43257", + "right": "issue:44863", + "accept": false, + "reason": "Different model-loading failures: Qwen3 MoE with accelerate+deepspeed versus NemotronH checkpoint loading." + }, + { + "left": "issue:43425", + "right": "issue:43828", + "accept": false, + "reason": "Torch version incompatibility versus autocast dtype mismatch; related only at a very high level." + }, + { + "left": "issue:44589", + "right": "issue:44756", + "accept": false, + "reason": "Float8 storage lookup error is unrelated to the Strix Halo mmap/OOM report." + }, + { + "left": "issue:44479", + "right": "issue:45663", + "accept": false, + "reason": "Different regressions: video input handling versus Gemma4 FSDP2 shared_kv_states rebuild." + }, + { + "left": "issue:39404", + "right": "issue:45701", + "accept": false, + "reason": "Whisper pipeline return_language regression and tokenizer-version behavior changes are unrelated." + }, + { + "left": "issue:43906", + "right": "issue:44530", + "accept": false, + "reason": "Isolated reproduction of an earlier bug versus a Qwen3.5 cache/group-type crash; not the same issue." + }, + { + "left": "issue:43611", + "right": "issue:43645", + "accept": false, + "reason": "Both are Transformers 5 loading regressions, but base_model_prefix loading and notebook-defined custom model init are different failures." + }, + { + "left": "issue:43425", + "right": "issue:43716", + "accept": false, + "reason": "Torch compatibility issue versus a Mistral-3 image preprocessor/model dtype mismatch." + }, + { + "left": "issue:43931", + "right": "issue:44863", + "accept": false, + "reason": "Both are model loading errors, but Qwen3-VL shape mismatch and NemotronH checkpoint loading are distinct." + }, + { + "left": "issue:44530", + "right": "issue:45084", + "accept": false, + "reason": "PagedAttentionCache linear_attention crash and non-template-node compile error do not share the same bug." + }, + { + "left": "issue:43901", + "right": "issue:44509", + "accept": false, + "reason": "Docs/API mismatch around return_all_scores is unrelated to removed pipeline-task documentation." + }, + { + "left": "issue:39404", + "right": "issue:44912", + "accept": false, + "reason": "Different models and APIs: Whisper pipeline language return versus MXFP4 quantization fallback." + }, + { + "left": "issue:44863", + "right": "issue:45663", + "accept": false, + "reason": "Different loading paths and symptoms; no evidence they are the same underlying bug." + }, + { + "left": "issue:44265", + "right": "issue:44898", + "accept": false, + "reason": "torch.export/torch_compilable_check failure is unrelated to Perceiver non-default-resolution resizing." + }, + { + "left": "issue:43299", + "right": "issue:43931", + "accept": false, + "reason": "Same model family, but one is a broad Transformers 5.0 loading regression and the other is a specific weight-shape mismatch; not enough to treat as one bug." + }, + { + "left": "issue:44479", + "right": "issue:44805", + "accept": false, + "reason": "Video-input regression versus a mask/tensor shape IndexError are not the same failure." + }, + { + "left": "issue:43425", + "right": "issue:44918", + "accept": false, + "reason": "Torch 2.10 compatibility and Qwen3.5 embedding unpacking are unrelated." + }, + { + "left": "issue:43856", + "right": "issue:44756", + "accept": false, + "reason": "Memory inefficiency in Qwen3 MoE training is different from mmap-triggered OOM on Strix Halo." + }, + { + "left": "issue:44589", + "right": "issue:45084", + "accept": false, + "reason": "Float8 storage lookup and template-node compilation errors are unrelated." + }, + { + "left": "issue:40990", + "right": "issue:42947", + "accept": false, + "reason": "High perplexity on gpt-oss-20b is unrelated to gradient checkpointing with PEFT LoRA." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet and stays conservative overall. The single accepted soft edge (45276/45292) is well supported by the titles and appears mergeable as the same resize_token_embeddings bug family. The other pairwise rejections are cautious and do not overclaim duplicate status. The canonical/best-issue framing is subjective but reasonable for this heterogeneous cluster." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 7, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-387", + "nodes": [ + "issue:32090", + "issue:34567", + "issue:36032", + "issue:39404", + "issue:40444", + "issue:40990", + "issue:42915", + "issue:43317", + "issue:43404", + "issue:43421", + "issue:43425", + "issue:43452", + "issue:43482", + "issue:43541", + "issue:43582", + "issue:43596", + "issue:43611", + "issue:43632", + "issue:43643", + "issue:43644", + "issue:43645", + "issue:43646", + "issue:43716", + "issue:43883", + "issue:43994", + "issue:44079", + "issue:44164", + "issue:44206", + "issue:44220", + "issue:44451", + "issue:44466", + "issue:44479", + "issue:44488", + "issue:44530", + "issue:44671", + "issue:44683", + "issue:44811", + "issue:44849", + "issue:44987", + "issue:44991", + "issue:45081", + "issue:45084", + "issue:45325", + "issue:45507", + "issue:45584", + "issue:45685", + "issue:45759" + ], + "soft_pairs": [ + "issue:43425|issue:45685", + "issue:39404|issue:44671", + "issue:44671|issue:44991", + "issue:43994|issue:44079", + "issue:36032|issue:43452", + "issue:44164|issue:45759", + "issue:43596|issue:44683", + "issue:44991|issue:45081", + "issue:32090|issue:39404", + "issue:44164|issue:44479", + "issue:44488|issue:45081", + "issue:40990|issue:43421", + "issue:43582|issue:43632", + "issue:43883|issue:44488", + "issue:43611|issue:43644", + "issue:44206|issue:44220", + "issue:43716|issue:45084", + "issue:44488|issue:44987", + "issue:34567|issue:40444", + "issue:43643|issue:44530", + "issue:44811|issue:45584", + "issue:43644|issue:43646", + "issue:43883|issue:44991", + "issue:36032|issue:42915", + "issue:44451|issue:44987", + "issue:43541|issue:45084", + "issue:43452|issue:43482", + "issue:42915|issue:44530", + "issue:43317|issue:43482", + "issue:44849|issue:45325", + "issue:43404|issue:44466", + "issue:36032|issue:43645", + "issue:45084|issue:45507" + ], + "prepared_review_unit_hash": "de8280b2fb7c09912c0e1c31e9b806256111b844c9b17f9287be5e1143152ee7", + "component_budget": { + "node_count": 387, + "item_count": 387, + "soft_pair_count": 1455, + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 33, + "serialized_chars": 37023, + "estimated_input_tokens": 9256, + "estimated_eval_tokens": 18768 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The set is mostly a collection of unrelated closed issue reports across many subsystems. I only see one plausible duplicate family: the GGUF loading regressions. No PRs are present.", + "confidence": 0.66, + "canonical_issue_reason": "issue:43452 is the best anchor because it describes the shared broken `gguf_file`/`from_pretrained` loading path directly and covers both tokenizer and CausalLM loading, making it the most general and actionable report in the clear duplicate family.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43452 is the strongest standalone issue here: it names the concrete loader regression rather than a single model-specific symptom, and the Qwen2.5 GGUF report can naturally attach to it.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43425", + "right": "issue:45685", + "accept": false, + "reason": "Both are v5-era failures, but one is Torch-version incompatibility and the other is an MPS histogram/int-type bug; different code paths." + }, + { + "left": "issue:39404", + "right": "issue:44671", + "accept": false, + "reason": "Whisper `return_language` pipeline behavior and CamemBERT masked-LM predictions are unrelated model-specific bugs." + }, + { + "left": "issue:44671", + "right": "issue:44991", + "accept": false, + "reason": "Incorrect CamemBERT predictions and tokenizer loading failure for `EMBIDDIA/est-roberta` are different problems." + }, + { + "left": "issue:43994", + "right": "issue:44079", + "accept": false, + "reason": "SigLIP2 nonsensical outputs and `ModelOutput` key assignment with prior `None` are distinct issues." + }, + { + "left": "issue:36032", + "right": "issue:43452", + "accept": false, + "reason": "T5 tokenizer special-token method conflict and GGUF loader breakage are different bug classes." + }, + { + "left": "issue:44164", + "right": "issue:45759", + "accept": false, + "reason": "`extra_state` save/load handling and `AutoModelForCausalLM.from_config` text_config unwrapping are unrelated serialization/config bugs." + }, + { + "left": "issue:43596", + "right": "issue:44683", + "accept": false, + "reason": "Deepspeed ZeRO3 init index error and compiled flex_attention failing on newer Torch versions are different failures." + }, + { + "left": "issue:44991", + "right": "issue:45081", + "accept": false, + "reason": "Both are loading regressions, but one is a tokenizer load failure and the other is a Mistral regex patch crash; not the same code path." + }, + { + "left": "issue:32090", + "right": "issue:39404", + "accept": false, + "reason": "Trainer GPU broadcast NoneType error and Whisper pipeline `return_language` regression are unrelated." + }, + { + "left": "issue:44164", + "right": "issue:44479", + "accept": false, + "reason": "`extra_state` serialization and Qwen video-input regression do not share a concrete bug." + }, + { + "left": "issue:44488", + "right": "issue:45081", + "accept": false, + "reason": "Both involve model/tokenizer loading, but they affect different models and different failure mechanisms." + }, + { + "left": "issue:40990", + "right": "issue:43421", + "accept": false, + "reason": "High perplexity on GPT-OSS and runtime special-token post-processor updates are not the same issue." + }, + { + "left": "issue:43582", + "right": "issue:43632", + "accept": false, + "reason": "Apple Silicon allocator warmup TypeError and `_is_hf_initialized` regression are different problems." + }, + { + "left": "issue:43883", + "right": "issue:44488", + "accept": false, + "reason": "Missing `all_tied_weights_keys` on Molmo and failure to load `cjvt/sleng-bert` are unrelated." + }, + { + "left": "issue:43611", + "right": "issue:43644", + "accept": false, + "reason": "Both are v5 regression reports, but one is about `base_model_prefix` loading and the other about non-persistent buffer serialization; not the same bug." + }, + { + "left": "issue:44206", + "right": "issue:44220", + "accept": false, + "reason": "Unsupported `center` arg in LasrFeatureExtractor and `_torch_extract_fbank_features()` are too different to call duplicate from titles alone." + }, + { + "left": "issue:43716", + "right": "issue:45084", + "accept": false, + "reason": "Mistral image-preprocessor dtype mismatch and template compilation TypeError are unrelated." + }, + { + "left": "issue:44488", + "right": "issue:44987", + "accept": false, + "reason": "Different model-loading failures for different models; no clear shared code-path." + }, + { + "left": "issue:34567", + "right": "issue:40444", + "accept": false, + "reason": "TrainerState token-count tracking and multi-image IterableDataset finetuning failure are unrelated." + }, + { + "left": "issue:43643", + "right": "issue:44530", + "accept": false, + "reason": "`trust_remote_code` missing fields and Qwen PagedAttentionCache group-type crash are different bugs." + }, + { + "left": "issue:44811", + "right": "issue:45584", + "accept": false, + "reason": "Whisper skip-special-tokens decode behavior and empty-transcription generation failure are both Whisper-related but not the same concrete defect." + }, + { + "left": "issue:43644", + "right": "issue:43646", + "accept": false, + "reason": "Both are v5 regressions in model initialization/serialization, but they describe different root causes and fixes." + }, + { + "left": "issue:43883", + "right": "issue:44991", + "accept": false, + "reason": "Molmo tied-weights attribute missing and tokenizer loading failure for `est-roberta` are unrelated." + }, + { + "left": "issue:36032", + "right": "issue:42915", + "accept": false, + "reason": "T5 tokenizer special-token conflict and Qwen3Moe FP8 config failure are different issues." + }, + { + "left": "issue:44451", + "right": "issue:44987", + "accept": false, + "reason": "Separate model-loading regressions affecting different repositories/models." + }, + { + "left": "issue:43541", + "right": "issue:45084", + "accept": false, + "reason": "Mixtral Dynamo tracing error and template-node compilation error are distinct code-path failures." + }, + { + "left": "issue:43452", + "right": "issue:43482", + "accept": true, + "reason": "Both point to the same GGUF loading regression in the `gguf_file`/`from_pretrained` path; the Qwen2.5 report looks like a specific instance of the broader breakage." + }, + { + "left": "issue:42915", + "right": "issue:44530", + "accept": false, + "reason": "Qwen3Moe FP8 config failure and Qwen3.5 PagedAttentionCache crash are different problems." + }, + { + "left": "issue:43317", + "right": "issue:43482", + "accept": false, + "reason": "GPU+CPU offload loading failure and GGUF loading failure are different loader paths." + }, + { + "left": "issue:44849", + "right": "issue:45325", + "accept": false, + "reason": "Qwen3.5 hidden-state output bug and Qwen2.5-VL rope-index scaling bug are separate model-specific regressions." + }, + { + "left": "issue:43404", + "right": "issue:44466", + "accept": false, + "reason": "Both involve tied `lm_head.weight`, but one is a model-class tying bug and the other is device-dependent serialization inconsistency; not the same issue." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet and stays conservative: most pairs are clearly unrelated from the titles, and the only accepted soft edge (43452/43482) is a plausible shared GGUF loading regression. No PRs are present, and nothing is overstated." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 8, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-387", + "nodes": [ + "issue:36032", + "issue:40990", + "issue:41553", + "issue:42175", + "issue:42915", + "issue:43262", + "issue:43296", + "issue:43317", + "issue:43322", + "issue:43377", + "issue:43404", + "issue:43425", + "issue:43531", + "issue:43540", + "issue:43582", + "issue:43611", + "issue:43632", + "issue:43645", + "issue:43646", + "issue:43742", + "issue:43819", + "issue:43856", + "issue:43994", + "issue:44062", + "issue:44079", + "issue:44220", + "issue:44265", + "issue:44291", + "issue:44403", + "issue:44451", + "issue:44479", + "issue:44534", + "issue:44589", + "issue:44610", + "issue:44683", + "issue:44756", + "issue:44811", + "issue:44877", + "issue:44987", + "issue:44991", + "issue:45081", + "issue:45084", + "issue:45127", + "issue:45216", + "issue:45325", + "issue:45357", + "issue:45542", + "issue:45663" + ], + "soft_pairs": [ + "issue:43262|issue:43377", + "issue:43646|issue:44534", + "issue:44987|issue:44991", + "issue:43994|issue:44610", + "issue:43322|issue:43540", + "issue:44403|issue:45081", + "issue:43377|issue:43819", + "issue:45216|issue:45357", + "issue:43645|issue:44534", + "issue:44451|issue:45081", + "issue:42915|issue:45084", + "issue:44265|issue:44610", + "issue:44220|issue:44683", + "issue:44987|issue:45081", + "issue:43404|issue:45127", + "issue:41553|issue:43582", + "issue:43632|issue:44756", + "issue:43296|issue:43531", + "issue:44062|issue:44987", + "issue:36032|issue:43742", + "issue:42915|issue:44291", + "issue:43531|issue:44877", + "issue:43611|issue:44534", + "issue:42915|issue:44589", + "issue:40990|issue:44811", + "issue:44479|issue:45325", + "issue:43317|issue:43856", + "issue:43425|issue:44403", + "issue:42175|issue:45542", + "issue:44079|issue:45663" + ], + "prepared_review_unit_hash": "a3188fa4952733c47c24bfe3ac309303f95669c26cae32d6c72e54b28ae3d1c3", + "component_budget": { + "node_count": 387, + "item_count": 387, + "soft_pair_count": 1455, + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 36624, + "estimated_input_tokens": 9156, + "estimated_eval_tokens": 18568 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly heterogeneous model-loading/regression reports, so most soft edges are false positives. I only see two plausible duplicate families: Qwen save_pretrained checkpoint corruption and Qwen video/rope-index regression. No PRs are present.", + "confidence": 0.71, + "canonical_issue_reason": "Issue 44479 is the broadest concrete regression here: it captures a shared Qwen video-input failure across multiple variants, and it cleanly represents the narrower Qwen2.5-VL temporal-position bug in 45325.", + "canonical_pr_reason": null, + "best_issue_reason": "44479 is the best standalone representative because it describes a clear versioned regression with a shared code path and multiple affected models, making it the most central issue among the few genuinely related reports.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43262", + "right": "issue:43377", + "accept": false, + "reason": "Both are audio-related, but one is a processor chat-template sampling-rate default and the other is a MIMI padding-mask batching bug; different code paths." + }, + { + "left": "issue:43646", + "right": "issue:44534", + "accept": false, + "reason": "One is custom model initialization in v5, the other is non-persistent buffers being filled with junk; related release, different bug." + }, + { + "left": "issue:44987", + "right": "issue:44991", + "accept": false, + "reason": "Both fail on transformers>=5, but one is model loading for `physical-intelligence/fast` and the other is tokenizer loading for EMBEDDIA/est-roberta; not the same failure." + }, + { + "left": "issue:43994", + "right": "issue:44610", + "accept": false, + "reason": "Different multimodal models and different symptoms: nonsensical outputs vs processor/model input-size mismatch." + }, + { + "left": "issue:43322", + "right": "issue:43540", + "accept": false, + "reason": "Both are multimodal, but one is a Llava Next load segfault and the other is a Qwen3OmniMoe video ValueError; no shared bug evidence." + }, + { + "left": "issue:44403", + "right": "issue:45081", + "accept": false, + "reason": "Noise on transformer loading versus a Mistral tokenizer regex crash; unrelated code paths." + }, + { + "left": "issue:43377", + "right": "issue:43819", + "accept": false, + "reason": "Both involve audio model outputs, but one is batched-vs-single padding behavior and the other is DAC STE mismatch; distinct root causes." + }, + { + "left": "issue:45216", + "right": "issue:45357", + "accept": true, + "reason": "Both report the same Qwen3.5 `save_pretrained` checkpoint corruption/incorrect visual-encoder key saving across versions; one is the broader regression, the other a specific manifestation." + }, + { + "left": "issue:43645", + "right": "issue:44534", + "accept": false, + "reason": "Custom model initialization in notebooks is unrelated to filling non-persistent buffers with junk, despite both being v5 regressions." + }, + { + "left": "issue:44451", + "right": "issue:45081", + "accept": false, + "reason": "Different model families and failures: a model load failure for ScandiBERT versus a Mistral tokenizer regex patch crash." + }, + { + "left": "issue:42915", + "right": "issue:45084", + "accept": false, + "reason": "FineGrainedFP8Config load failure and a template-node compile TypeError are separate problems." + }, + { + "left": "issue:44265", + "right": "issue:44610", + "accept": false, + "reason": "One is `torch.export` with `torch_compilable_check`, the other is an OmDet-Turbo processor size mismatch; different layers of the stack." + }, + { + "left": "issue:44220", + "right": "issue:44683", + "accept": false, + "reason": "Audio feature extraction vs compiled flex_attention on newer torch; no shared code path." + }, + { + "left": "issue:44987", + "right": "issue:45081", + "accept": false, + "reason": "A model-loading regression and a tokenizer regex patch crash are unrelated." + }, + { + "left": "issue:43404", + "right": "issue:45127", + "accept": false, + "reason": "Both touch weight tying, but one is Mistral3 `lm_head` tying and the other is LoRA merge collapse with extended vocab; not the same bug." + }, + { + "left": "issue:41553", + "right": "issue:43582", + "accept": false, + "reason": "Tokenizer-loading error for Voxtral versus Apple Silicon allocator warmup TypeError; unrelated." + }, + { + "left": "issue:43632", + "right": "issue:44756", + "accept": false, + "reason": "Both are v5/runtime issues, but `_is_hf_initialized` flag breakage and mmap OOM on Strix Halo are different bugs." + }, + { + "left": "issue:43296", + "right": "issue:43531", + "accept": false, + "reason": "PaddleOCR-VL loading in vLLM and Qwen3-MoE sliding-window behavior are unrelated." + }, + { + "left": "issue:44062", + "right": "issue:44987", + "accept": false, + "reason": "Tokenizer AddedToken `special` keyword conflict is not the same as a model load failure for `physical-intelligence/fast`." + }, + { + "left": "issue:36032", + "right": "issue:43742", + "accept": false, + "reason": "Both are model/tokenizer load errors, but T5 `add_special_tokens` conflict and MobileLLM key error have different causes." + }, + { + "left": "issue:42915", + "right": "issue:44291", + "accept": false, + "reason": "Both involve model loading under v5 FP8/empty-init scenarios, but one is Qwen3Moe FineGrainedFP8Config and the other is unexpected `_is_hf_initialized` arg; distinct issues." + }, + { + "left": "issue:43531", + "right": "issue:44877", + "accept": false, + "reason": "Qwen3-MoE sliding-window behavior and strict config rejection for granite_speech are not the same bug." + }, + { + "left": "issue:43611", + "right": "issue:44534", + "accept": false, + "reason": "Base-model-prefix loading regression and junk-filled non-persistent buffers are separate v5 breakages." + }, + { + "left": "issue:42915", + "right": "issue:44589", + "accept": false, + "reason": "FP8 config load failure and missing Float8 storage type are related only loosely by dtype support, not by the same failure mode." + }, + { + "left": "issue:40990", + "right": "issue:44811", + "accept": false, + "reason": "High perplexity on gpt-oss-20b and Whisper `batch_decode` ignoring `skip_special_tokens` are unrelated." + }, + { + "left": "issue:44479", + "right": "issue:45325", + "accept": true, + "reason": "45325 is a precise Qwen2.5-VL rope-index/temporal-position regression that matches the broader Qwen video-input regression in 44479; same shared code path, one broader and one specific report." + }, + { + "left": "issue:43317", + "right": "issue:43856", + "accept": false, + "reason": "Dequantized model offload failure and Qwen3 MoE training memory inefficiency are different performance/load bugs." + }, + { + "left": "issue:43425", + "right": "issue:44403", + "accept": false, + "reason": "Torch 2.10 incompatibility is a dependency-compat issue, while unnecessary loading noise is a logging/UX issue." + }, + { + "left": "issue:42175", + "right": "issue:45542", + "accept": false, + "reason": "Different packaging/backend problems: TensorFlow missing from `.[torch]` versus tensorboard installed without TensorFlow causing backend errors." + }, + { + "left": "issue:44079", + "right": "issue:45663", + "accept": false, + "reason": "`ModelOutput` key assignment bug and Gemma4 FSDP2 shared_kv_states `KeyError` are unrelated." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet: it correctly characterizes the cluster as mostly heterogeneous and only identifies two plausible duplicate families that are supported by the issue titles. The accepted soft-edge pairs are conservative enough based on the available evidence, and the rejected pairs do not overclaim shared root causes." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 9, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-387", + "nodes": [ + "issue:33666", + "issue:36032", + "issue:40444", + "issue:41720", + "issue:42175", + "issue:42915", + "issue:43296", + "issue:43377", + "issue:43421", + "issue:43454", + "issue:43482", + "issue:43525", + "issue:43531", + "issue:43611", + "issue:43632", + "issue:43644", + "issue:43646", + "issue:43792", + "issue:43818", + "issue:43828", + "issue:43881", + "issue:43957", + "issue:43994", + "issue:44079", + "issue:44265", + "issue:44291", + "issue:44387", + "issue:44403", + "issue:44466", + "issue:44488", + "issue:44534", + "issue:44568", + "issue:44589", + "issue:44617", + "issue:44683", + "issue:44829", + "issue:44898", + "issue:44991", + "issue:45020", + "issue:45072", + "issue:45081", + "issue:45325", + "issue:45357", + "issue:45381", + "issue:45479", + "issue:45491", + "issue:45538", + "issue:45800" + ], + "soft_pairs": [ + "issue:43421|issue:44568", + "issue:36032|issue:41720", + "issue:43818|issue:45357", + "issue:43632|issue:44534", + "issue:43792|issue:44589", + "issue:44683|issue:45020", + "issue:44403|issue:44991", + "issue:33666|issue:40444", + "issue:43482|issue:45538", + "issue:44829|issue:45479", + "issue:40444|issue:45325", + "issue:43482|issue:43525", + "issue:44291|issue:44589", + "issue:43377|issue:45491", + "issue:41720|issue:45081", + "issue:42175|issue:43296", + "issue:44265|issue:45072", + "issue:43632|issue:45072", + "issue:43611|issue:43646", + "issue:43531|issue:44291", + "issue:44079|issue:44534", + "issue:43994|issue:44898", + "issue:43454|issue:44466", + "issue:43881|issue:44488", + "issue:43828|issue:44683", + "issue:42915|issue:43482", + "issue:43792|issue:43957", + "issue:44079|issue:45072", + "issue:43644|issue:44079", + "issue:40444|issue:45381", + "issue:45325|issue:45381", + "issue:43296|issue:45800", + "issue:44387|issue:44617" + ], + "prepared_review_unit_hash": "d193d5342c8ec7b6520225a37ead9b2975679b1d4c6cc94421caebfa075e4c27", + "component_budget": { + "node_count": 387, + "item_count": 387, + "soft_pair_count": 1455, + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 33, + "serialized_chars": 37774, + "estimated_input_tokens": 9444, + "estimated_eval_tokens": 19144 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is heterogeneous: most items are unrelated bug reports spanning different models, regressions, and subsystems. I found no soft-edge pair that clearly describes the same underlying bug or change, so all soft links should stay unmerged.", + "confidence": 0.92, + "canonical_issue_reason": "Issue 45020 is the broadest umbrella-style report in the set (general regressions in recent transformers versions affecting remote_code models), so it is the best representative issue if one must be chosen.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 45020 is also the most suitable global issue representative because it is open, broadly scoped, and closer to a cross-cutting regression than the more model-specific reports.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43421", + "right": "issue:44568", + "accept": false, + "reason": "Both involve special tokens, but one is about TokenizersBackend post-processor updates while the other is a specific mdeberta-v3 add_special_tokens regression." + }, + { + "left": "issue:36032", + "right": "issue:41720", + "accept": false, + "reason": "Completely different failures: a T5 tokenizer method conflict versus a Qwen3 auto device mapping CUDA assert." + }, + { + "left": "issue:43818", + "right": "issue:45357", + "accept": false, + "reason": "Both touch vision models, but one is a Video-LLaVA tower/weight issue and the other is Qwen3.5 save_pretrained visual key serialization." + }, + { + "left": "issue:43632", + "right": "issue:44534", + "accept": false, + "reason": "Both are Transformers v5 regressions, but they describe different code paths: _is_hf_initialized versus non-persistent buffer initialization." + }, + { + "left": "issue:43792", + "right": "issue:44589", + "accept": false, + "reason": "Unrelated runtime failures: Whisper loading versus a Float8 storage lookup error." + }, + { + "left": "issue:44683", + "right": "issue:45020", + "accept": false, + "reason": "Different problem areas: torch 2.9 flex_attention compilation versus remote_code model loading regressions." + }, + { + "left": "issue:44403", + "right": "issue:44991", + "accept": false, + "reason": "One is about noisy logging during model loading; the other is a tokenizer loading failure for a specific model." + }, + { + "left": "issue:33666", + "right": "issue:40444", + "accept": false, + "reason": "Both are Qwen2-VL-family training issues, but they concern different model generations and different failure modes." + }, + { + "left": "issue:43482", + "right": "issue:45538", + "accept": false, + "reason": "GGUF loading under Transformers v5 is unrelated to CLIPTokenizer model_max_length behavior." + }, + { + "left": "issue:44829", + "right": "issue:45479", + "accept": false, + "reason": "Different bugs in different heads: Perceiver image resolution handling versus sequence-classification zero-loss behavior." + }, + { + "left": "issue:40444", + "right": "issue:45325", + "accept": false, + "reason": "Both are Qwen2.5-VL-related, but one is an iterable-dataset multi-image training failure and the other is a still-image temporal position-id scaling bug." + }, + { + "left": "issue:43482", + "right": "issue:43525", + "accept": false, + "reason": "Unrelated: GGUF model loading failure versus missing pad_token_id on Llama4Config." + }, + { + "left": "issue:44291", + "right": "issue:44589", + "accept": false, + "reason": "Different initialization issues: unexpected _is_hf_initialized argument versus missing Float8 storage object." + }, + { + "left": "issue:43377", + "right": "issue:45491", + "accept": false, + "reason": "Both involve batching/masking, but they are different models and distinct failure mechanisms." + }, + { + "left": "issue:41720", + "right": "issue:45081", + "accept": false, + "reason": "A CUDA assert during Qwen3 device mapping is unrelated to the Mistral tokenizer backend_tokenizer crash." + }, + { + "left": "issue:42175", + "right": "issue:43296", + "accept": false, + "reason": "Packaging/backend inclusion differs from a model-loading failure in vLLM." + }, + { + "left": "issue:44265", + "right": "issue:45072", + "accept": false, + "reason": "torch.export/torch_compilable_check is unrelated to bf16 dtype mismatches in inference." + }, + { + "left": "issue:43632", + "right": "issue:45072", + "accept": false, + "reason": "Different Transformers v5 regressions with no shared concrete code path." + }, + { + "left": "issue:43611", + "right": "issue:43646", + "accept": false, + "reason": "Both are model-init/load regressions in v5, but one is about base_model_prefix handling and the other about custom initialization; not the same bug." + }, + { + "left": "issue:43531", + "right": "issue:44291", + "accept": false, + "reason": "Sliding-window behavior in Qwen3-MoE is unrelated to init_empty_weights/_is_hf_initialized." + }, + { + "left": "issue:44079", + "right": "issue:44534", + "accept": false, + "reason": "ModelOutput key assignment and non-persistent buffer junk are separate internal state bugs." + }, + { + "left": "issue:43994", + "right": "issue:44898", + "accept": false, + "reason": "Different models and symptoms: SigLIP2 bad outputs versus Perceiver resolution failure." + }, + { + "left": "issue:43454", + "right": "issue:44466", + "accept": false, + "reason": "Both mention lm_head, but one is failed weight tying causing bad generation and the other is serialization inconsistency depending on device." + }, + { + "left": "issue:43881", + "right": "issue:44488", + "accept": false, + "reason": "Unrelated loading failures for different models." + }, + { + "left": "issue:43828", + "right": "issue:44683", + "accept": false, + "reason": "Autocast dtype mismatch is a different issue from compiled flex_attention failing on torch 2.9." + }, + { + "left": "issue:42915", + "right": "issue:43482", + "accept": false, + "reason": "Qwen3MoE FP8 configuration failure is unrelated to GGUF loading under Transformers v5." + }, + { + "left": "issue:43792", + "right": "issue:43957", + "accept": false, + "reason": "Whisper loading failure and meta-device model loading breakage are different problems." + }, + { + "left": "issue:44079", + "right": "issue:45072", + "accept": false, + "reason": "ModelOutput key handling is unrelated to dtype mismatches in bf16 inference." + }, + { + "left": "issue:43644", + "right": "issue:44079", + "accept": false, + "reason": "Both are v5 internals issues, but one is buffer serialization and the other is ModelOutput key bookkeeping." + }, + { + "left": "issue:40444", + "right": "issue:45381", + "accept": false, + "reason": "Both are Qwen2.5-VL issues, but one is multi-image iterable-dataset training and the other is video vision_position_ids ordering." + }, + { + "left": "issue:45325", + "right": "issue:45381", + "accept": false, + "reason": "Both concern Qwen2.5-VL position IDs, but one is still-image temporal scaling and the other is video input position IDs; not the same concrete bug." + }, + { + "left": "issue:43296", + "right": "issue:45800", + "accept": false, + "reason": "Different compatibility issues: PaddleOCR-VL loading in vLLM versus torch 2.4.1/transformers 5.8.0 incompatibility." + }, + { + "left": "issue:44387", + "right": "issue:44617", + "accept": false, + "reason": "Both mention OOM, but one is an int4/Transformers v5 memory regression and the other is a model-specific CUDA OOM." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 10, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-387", + "nodes": [ + "issue:34634", + "issue:35707", + "issue:39692", + "issue:40444", + "issue:40990", + "issue:41720", + "issue:43262", + "issue:43296", + "issue:43317", + "issue:43377", + "issue:43404", + "issue:43454", + "issue:43526", + "issue:43550", + "issue:43572", + "issue:43577", + "issue:43606", + "issue:43632", + "issue:43701", + "issue:43708", + "issue:43720", + "issue:43746", + "issue:43792", + "issue:43819", + "issue:43873", + "issue:43881", + "issue:43883", + "issue:44186", + "issue:44265", + "issue:44291", + "issue:44423", + "issue:44466", + "issue:44610", + "issue:44683", + "issue:44877", + "issue:44898", + "issue:44977", + "issue:44991", + "issue:45072", + "issue:45325", + "issue:45357", + "issue:45406", + "issue:45563", + "issue:45584", + "issue:45632", + "issue:45684", + "issue:45701", + "issue:45759" + ], + "soft_pairs": [ + "issue:35707|issue:40444", + "issue:44291|issue:44877", + "issue:43526|issue:43883", + "issue:43606|issue:44265", + "issue:43317|issue:43873", + "issue:43881|issue:44991", + "issue:43701|issue:45325", + "issue:44186|issue:44265", + "issue:43632|issue:45563", + "issue:43606|issue:43819", + "issue:43262|issue:43746", + "issue:44610|issue:45325", + "issue:43454|issue:43550", + "issue:43377|issue:44610", + "issue:43454|issue:45072", + "issue:43377|issue:44898", + "issue:40990|issue:45584", + "issue:43572|issue:45759", + "issue:39692|issue:40444", + "issue:44466|issue:45357", + "issue:43296|issue:44991", + "issue:43454|issue:43720", + "issue:43792|issue:44683", + "issue:43873|issue:45701", + "issue:41720|issue:44977", + "issue:45632|issue:45684", + "issue:43404|issue:45357", + "issue:44423|issue:45406", + "issue:44186|issue:45072", + "issue:43577|issue:43708", + "issue:34634|issue:35707", + "issue:35707|issue:41720", + "issue:43526|issue:43881" + ], + "prepared_review_unit_hash": "4f08226759877c8617e9c37a39e48627232ec9b9057bd2a4c17d39841a942d02", + "component_budget": { + "node_count": 387, + "item_count": 387, + "soft_pair_count": 1455, + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 33, + "serialized_chars": 37933, + "estimated_input_tokens": 9484, + "estimated_eval_tokens": 19224 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is not a true duplicate set; it mixes unrelated issues across offloading/loading, multimodal preprocessing, tokenizer/config regressions, and training/generation bugs. If one issue must anchor it, 43873 is the broadest offloading/quantization report, but the soft edges do not describe the same underlying bug.", + "confidence": 0.83, + "canonical_issue_reason": "Issue 43873 is the closest umbrella because it is the broadest offloading/quantization failure and overlaps the nearby device/offload-loading reports more than any other item; the rest are mostly narrower, model-specific, or unrelated bugs.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 43873 is the best global representative only in a loose triage sense: it is the most general device/offloading problem in the set and is the only one that can plausibly subsume related offload/quantization failures, though it is not a duplicate of the others.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:35707", + "right": "issue:40444", + "accept": false, + "reason": "Different bugs: generation with inputs_embeds/past_key_values versus multimodal finetuning on multiple images per prompt." + }, + { + "left": "issue:44291", + "right": "issue:44877", + "accept": false, + "reason": "Constructor/init_empty_weights incompatibility is unrelated to strict config loading for granite_speech." + }, + { + "left": "issue:43526", + "right": "issue:43883", + "accept": false, + "reason": "Image label reduction bug in a processor is unrelated to a model missing all_tied_weights_keys metadata." + }, + { + "left": "issue:43606", + "right": "issue:44265", + "accept": false, + "reason": "CPU offload device mismatch is not the same code path as torch.export failures with torch_compilable_check." + }, + { + "left": "issue:43317", + "right": "issue:43873", + "accept": false, + "reason": "Both involve offloading/quantization, but one is a dequantized model load failure and the other is a broader offloading regression; not the same concrete bug." + }, + { + "left": "issue:43881", + "right": "issue:44991", + "accept": false, + "reason": "A model loading failure for glm-4v-9b is unrelated to a tokenizer-loading regression for EMBEDDIA/est-roberta." + }, + { + "left": "issue:43701", + "right": "issue:45325", + "accept": false, + "reason": "Resume-from-checkpoint key mismatch is unrelated to Qwen2.5-VL rope index temporal scaling." + }, + { + "left": "issue:44186", + "right": "issue:44265", + "accept": false, + "reason": "Tokenizer crashes on NER/padding are unrelated to export-time failures in torch_compilable_check." + }, + { + "left": "issue:43632", + "right": "issue:45563", + "accept": false, + "reason": "A v5 HF initialization flag regression is unrelated to a stale warning in paged generate()." + }, + { + "left": "issue:43606", + "right": "issue:43819", + "accept": false, + "reason": "Device mismatch during CPU offload is not the same as DAC.from_latents diverging from forward due to missing STE." + }, + { + "left": "issue:43262", + "right": "issue:43746", + "accept": false, + "reason": "Audio chat-template sampling-rate defaulting is unrelated to GraniteSpeech PEFT checkpoint loading." + }, + { + "left": "issue:44610", + "right": "issue:45325", + "accept": false, + "reason": "Processor image-size mismatch is unrelated to Qwen2.5-VL temporal position_id scaling." + }, + { + "left": "issue:43454", + "right": "issue:43550", + "accept": false, + "reason": "Tied lm_head weights causing bad text generation is unrelated to a torch.compile/SDPA failure." + }, + { + "left": "issue:43377", + "right": "issue:44610", + "accept": false, + "reason": "Batch-vs-single encoder output mismatch from padding-mask handling is unrelated to the OmDet-Turbo processor image-size bug." + }, + { + "left": "issue:43454", + "right": "issue:45072", + "accept": false, + "reason": "Weight-tying serialization bugs are not the same as dtype mismatches in bfloat16 inference." + }, + { + "left": "issue:43377", + "right": "issue:44898", + "accept": false, + "reason": "Missing padding-mask support in MIMI encoder is unrelated to Perceiver non-default resolution failures." + }, + { + "left": "issue:40990", + "right": "issue:45584", + "accept": false, + "reason": "High perplexity on gpt-oss-20b is unrelated to Whisper empty-transcription generation after token alignment." + }, + { + "left": "issue:43572", + "right": "issue:45759", + "accept": false, + "reason": "A missing config field in StableLmConfig is unrelated to unwrapping text_config for composite Qwen multimodal configs." + }, + { + "left": "issue:39692", + "right": "issue:40444", + "accept": false, + "reason": "Broken docs/example mismatch for SigLIP2 is unrelated to Qwen2.5-VL finetuning with multiple images per prompt." + }, + { + "left": "issue:44466", + "right": "issue:45357", + "accept": false, + "reason": "Device-dependent lm_head serialization is not the same bug as Qwen3.5 saving incorrect visual encoder keys." + }, + { + "left": "issue:43296", + "right": "issue:44991", + "accept": false, + "reason": "vLLM/PaddleOCR-VL model load failure is unrelated to tokenizer loading for EMBEDDIA/est-roberta." + }, + { + "left": "issue:43454", + "right": "issue:43720", + "accept": false, + "reason": "lm_head weight tying is unrelated to BitNet packed-weight unpacking during accelerate loading." + }, + { + "left": "issue:43792", + "right": "issue:44683", + "accept": false, + "reason": "Whisper runtime failure is unrelated to compiled flex_attention breakage on torch >= 2.9." + }, + { + "left": "issue:43873", + "right": "issue:45701", + "accept": false, + "reason": "Quantized offloading problems are unrelated to a version-dependent tokenization change." + }, + { + "left": "issue:41720", + "right": "issue:44977", + "accept": false, + "reason": "Qwen3 auto device mapping cuda asserts are unrelated to Qwen3.5 flash-attention generation failures." + }, + { + "left": "issue:45632", + "right": "issue:45684", + "accept": false, + "reason": "Cache-path collisions for trust_remote_code are unrelated to read-only permission propagation in save_pretrained." + }, + { + "left": "issue:43404", + "right": "issue:45357", + "accept": false, + "reason": "Mistral3 lm_head tying is unrelated to Qwen3.5 visual encoder key serialization." + }, + { + "left": "issue:44423", + "right": "issue:45406", + "accept": false, + "reason": "Continuous-batching multimodal crash from a string .to() call is unrelated to Gemma4Processor missing _tokenizer." + }, + { + "left": "issue:44186", + "right": "issue:45072", + "accept": false, + "reason": "Tokenizer padding/truncation crashes are unrelated to dtype mismatches in SwitchTransformers/TimmWrapperModel." + }, + { + "left": "issue:43577", + "right": "issue:43708", + "accept": false, + "reason": "dtype remaining float32 in Blip2 is unrelated to resume_from_checkpoint max_steps calculation." + }, + { + "left": "issue:34634", + "right": "issue:35707", + "accept": false, + "reason": "Bark voice_preset not working is unrelated to progressive generation with inputs_embeds and past_key_values." + }, + { + "left": "issue:35707", + "right": "issue:41720", + "accept": false, + "reason": "Progressive generation issues are unrelated to Qwen3 auto device mapping cuda asserts." + }, + { + "left": "issue:43526", + "right": "issue:43881", + "accept": false, + "reason": "BeitImageProcessorFast label reduction bug is unrelated to glm-4v-9b model loading failure." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 11, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-387", + "nodes": [ + "issue:30064", + "issue:34634", + "issue:36010", + "issue:36331", + "issue:40990", + "issue:41720", + "issue:42617", + "issue:42915", + "issue:43295", + "issue:43316", + "issue:43317", + "issue:43322", + "issue:43388", + "issue:43421", + "issue:43450", + "issue:43531", + "issue:43540", + "issue:43550", + "issue:43572", + "issue:43606", + "issue:43653", + "issue:43720", + "issue:43749", + "issue:43782", + "issue:43873", + "issue:44186", + "issue:44292", + "issue:44361", + "issue:44387", + "issue:44479", + "issue:44492", + "issue:44509", + "issue:44534", + "issue:44610", + "issue:44671", + "issue:44743", + "issue:44756", + "issue:44877", + "issue:44898", + "issue:44913", + "issue:44918", + "issue:44991", + "issue:45059", + "issue:45161", + "issue:45406", + "issue:45464", + "issue:45701", + "issue:45704" + ], + "soft_pairs": [ + "issue:43873|issue:44756", + "issue:43749|issue:44534", + "issue:44877|issue:44991", + "issue:42617|issue:42915", + "issue:43450|issue:45701", + "issue:43550|issue:43720", + "issue:36331|issue:42915", + "issue:44610|issue:44898", + "issue:44186|issue:44610", + "issue:36331|issue:42617", + "issue:44387|issue:45704", + "issue:44479|issue:44918", + "issue:43316|issue:44913", + "issue:42915|issue:43317", + "issue:43550|issue:43606", + "issue:40990|issue:45701", + "issue:43295|issue:45701", + "issue:41720|issue:43322", + "issue:41720|issue:45161", + "issue:44671|issue:44743", + "issue:43531|issue:43572", + "issue:44292|issue:44479", + "issue:41720|issue:42617", + "issue:36010|issue:42617", + "issue:44186|issue:44361", + "issue:43540|issue:43782", + "issue:45406|issue:45464", + "issue:41720|issue:43873", + "issue:43421|issue:45701", + "issue:30064|issue:34634", + "issue:44492|issue:44509", + "issue:43606|issue:43720", + "issue:43388|issue:45059", + "issue:42915|issue:43322", + "issue:43550|issue:43653" + ], + "prepared_review_unit_hash": "65990d0226dd1208b38b64868d7fd9382c178abaa883219793d9cf21ca844659", + "component_budget": { + "node_count": 387, + "item_count": 387, + "soft_pair_count": 1455, + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 35, + "serialized_chars": 37799, + "estimated_input_tokens": 9450, + "estimated_eval_tokens": 19156 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These items are mostly unrelated issues from many different subsystems. None of the soft pairs look like true duplicates, so I rejected them all. If forced to anchor the cluster, the broadest loading/offloading-related report is the best fit, but there is no strong canonical match overall.", + "confidence": 0.81, + "canonical_issue_reason": "No true canonical duplicate cluster is evident; issue 43873 is the broadest open report and the closest thing to an anchor because it covers the general offloading/quantization failure space.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 43873 is the best global representative of the only recurring theme in the set (loading/offloading/memory behavior), even though most items are still distinct.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43873", + "right": "issue:44756", + "accept": false, + "reason": "Both mention offloading/memory, but one is quantization offload behavior and the other is mmap/OOM on Strix Halo; different root causes and code paths." + }, + { + "left": "issue:43749", + "right": "issue:44534", + "accept": false, + "reason": "Both touch model loading/memory, but FSDP CPU RAM-efficient loading and non-persistent buffer corruption are different bugs." + }, + { + "left": "issue:44877", + "right": "issue:44991", + "accept": false, + "reason": "Both are loading regressions after v5, but one is a config-loading strictness issue and the other is tokenizer loading for a specific model." + }, + { + "left": "issue:42617", + "right": "issue:42915", + "accept": false, + "reason": "Different failures: running 3d_parallel.py versus Qwen3MoE with FineGrainedFP8Config; no shared code-path bug." + }, + { + "left": "issue:43450", + "right": "issue:45701", + "accept": false, + "reason": "Video batching shape bug versus version-dependent tokenization changes; unrelated behavior." + }, + { + "left": "issue:43550", + "right": "issue:43720", + "accept": false, + "reason": "Both are model/runtime failures, but torch.compile with SDPA and packed-weight unpacking during loading are distinct problems." + }, + { + "left": "issue:36331", + "right": "issue:42915", + "accept": false, + "reason": "Trainer API signature mismatch versus Qwen3MoE FP8 loading failure; not the same bug." + }, + { + "left": "issue:44610", + "right": "issue:44898", + "accept": false, + "reason": "Both are image-size/preprocessing issues, but they affect different models and different failure mechanisms." + }, + { + "left": "issue:44186", + "right": "issue:44610", + "accept": false, + "reason": "Tokenizer crash on NER padding/truncation is unrelated to OmDet-Turbo image-size mismatch." + }, + { + "left": "issue:36331", + "right": "issue:42617", + "accept": false, + "reason": "Unexpected trainer keyword argument and 3d_parallel execution failure are unrelated." + }, + { + "left": "issue:44387", + "right": "issue:45704", + "accept": false, + "reason": "CUDA memory growth under int4 quantization is not the same as the apex RMSNorm memory leak." + }, + { + "left": "issue:44479", + "right": "issue:44918", + "accept": false, + "reason": "Video-input regression in Qwen VL models and Qwen3.5 embedding unpacking with TRL are different bugs." + }, + { + "left": "issue:43316", + "right": "issue:44913", + "accept": false, + "reason": "Both are config-related, but one is Gemma3TextConfig API mismatch and the other is GPTNeoX rotary_pct reload loss." + }, + { + "left": "issue:42915", + "right": "issue:43317", + "accept": false, + "reason": "Qwen3MoE FP8 failure and dequantized model offload loading failure are different code paths." + }, + { + "left": "issue:43550", + "right": "issue:43606", + "accept": false, + "reason": "Both involve model execution/loading, but Bamba torch.compile SDPA and Bark CPU offload device mismatch are not the same bug." + }, + { + "left": "issue:40990", + "right": "issue:45701", + "accept": false, + "reason": "Perplexity on gpt-oss-20b is a model/eval issue, not the tokenizer-version regression described in 45701." + }, + { + "left": "issue:43295", + "right": "issue:45701", + "accept": false, + "reason": "Processor.tokenizer regression and tokenization changes across versions are related only superficially; different concrete failures." + }, + { + "left": "issue:41720", + "right": "issue:43322", + "accept": false, + "reason": "Qwen3 auto device mapping CUDA assert and Llava Next segfault are separate model-loading failures." + }, + { + "left": "issue:41720", + "right": "issue:45161", + "accept": false, + "reason": "Both mention Qwen/generation scaling issues, but one is CUDA assert under device mapping and the other is tensor-parallel not working." + }, + { + "left": "issue:44671", + "right": "issue:44743", + "accept": false, + "reason": "CamemBERT masked-LM output regression and Qwen3.5 recurrent-state reset are unrelated model bugs." + }, + { + "left": "issue:43531", + "right": "issue:43572", + "accept": false, + "reason": "Qwen3-MoE sliding-window behavior and StableLmConfig pad_token_idx loss are different issues." + }, + { + "left": "issue:44292", + "right": "issue:44479", + "accept": false, + "reason": "Qwen-3-8B-NVFP4 runtime error and Qwen VL video-input regression are different failures." + }, + { + "left": "issue:41720", + "right": "issue:43873", + "accept": false, + "reason": "Both involve Qwen/model loading, but the specific failures are device-mapping CUDA assert versus quantization/offload behavior." + }, + { + "left": "issue:43421", + "right": "issue:45701", + "accept": false, + "reason": "Runtime post-processor update bug is not the same as version-dependent tokenization changes." + }, + { + "left": "issue:30064", + "right": "issue:34634", + "accept": false, + "reason": "Image segmentation-map processing and Bark voice_preset are unrelated processor bugs." + }, + { + "left": "issue:44492", + "right": "issue:44509", + "accept": false, + "reason": "A typo in cache-strategy docs is unrelated to removed pipeline-task docs." + }, + { + "left": "issue:43606", + "right": "issue:43720", + "accept": false, + "reason": "Both are loading/offload related, but the failing mechanism differs: CPU offload device mismatch versus packed weights not unpacked." + }, + { + "left": "issue:43388", + "right": "issue:45059", + "accept": false, + "reason": "Metric-gathering label truncation and SAM3 text/bounding-box behavior are different code paths." + }, + { + "left": "issue:42915", + "right": "issue:43322", + "accept": false, + "reason": "Qwen3MoE FP8 failure and Llava Next segmentation fault are distinct runtime problems." + }, + { + "left": "issue:43550", + "right": "issue:43653", + "accept": false, + "reason": "Torch.compile/SDPA failure and BigBirdTokenizer special-token registration bug are unrelated." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 12, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-387", + "nodes": [ + "issue:33666", + "issue:34567", + "issue:34634", + "issue:35707", + "issue:39692", + "issue:41720", + "issue:42886", + "issue:42915", + "issue:43381", + "issue:43421", + "issue:43525", + "issue:43531", + "issue:43596", + "issue:43606", + "issue:43653", + "issue:43746", + "issue:43782", + "issue:43844", + "issue:43873", + "issue:43927", + "issue:43994", + "issue:44062", + "issue:44186", + "issue:44220", + "issue:44368", + "issue:44451", + "issue:44514", + "issue:44545", + "issue:44589", + "issue:44610", + "issue:44625", + "issue:44671", + "issue:44843", + "issue:44898", + "issue:44918", + "issue:44936", + "issue:44987", + "issue:44991", + "issue:45042", + "issue:45072", + "issue:45081", + "issue:45356", + "issue:45357", + "issue:45381", + "issue:45479", + "issue:45588", + "issue:45704", + "issue:45759" + ], + "soft_pairs": [ + "issue:44514|issue:44545", + "issue:43531|issue:45759", + "issue:43606|issue:43653", + "issue:43421|issue:43927", + "issue:34567|issue:35707", + "issue:44368|issue:45357", + "issue:41720|issue:43746", + "issue:43873|issue:45704", + "issue:44451|issue:44991", + "issue:44898|issue:45072", + "issue:44062|issue:44451", + "issue:43421|issue:43653", + "issue:45588|issue:45759", + "issue:33666|issue:35707", + "issue:44987|issue:45042", + "issue:42886|issue:44843", + "issue:43531|issue:44589", + "issue:39692|issue:44610", + "issue:43381|issue:43844", + "issue:43421|issue:44186", + "issue:44918|issue:44936", + "issue:45081|issue:45356", + "issue:44625|issue:45479", + "issue:43525|issue:43531", + "issue:43782|issue:45381", + "issue:43994|issue:44671", + "issue:34634|issue:41720", + "issue:43381|issue:45704", + "issue:39692|issue:43873", + "issue:42915|issue:45759", + "issue:43596|issue:44220" + ], + "prepared_review_unit_hash": "8ace39b3185234ee1c92db22afa53139bf2e054aa423396589cf3f43cbeb4b73", + "component_budget": { + "node_count": 387, + "item_count": 387, + "soft_pair_count": 1455, + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 37263, + "estimated_input_tokens": 9316, + "estimated_eval_tokens": 18888 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Mostly a heterogeneous cluster of unrelated bug reports. Two soft edges look like true duplicates: the Qwen2_5_VLProcessor batched `padding=False` crash, and the HF_HUB_OFFLINE tokenizer regression tied to `_patch_mistral_regex` / `model_info()`.", + "confidence": 0.84, + "canonical_issue_reason": "issue:44843 is the clearest root-cause report in the set: it pinpoints the offline failure to an unconditional `model_info()` call inside `_patch_mistral_regex`, which also explains the broader symptom in issue:42886.", + "canonical_pr_reason": null, + "best_issue_reason": "Best anchor for the cluster because it states the concrete failing code path and environment trigger, making it the most actionable representative of the offline tokenizer bug.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44514", + "right": "issue:44545", + "accept": true, + "reason": "Same Qwen2_5_VLProcessor `apply_chat_template` crash on batched input with `padding=False`; the titles describe the same failure." + }, + { + "left": "issue:43531", + "right": "issue:45759", + "accept": false, + "reason": "Both are Qwen3-family bugs, but one is `sliding_window` handling in Qwen3-MoE and the other is `from_config`/`text_config` unwrapping for composite multimodal configs." + }, + { + "left": "issue:43606", + "right": "issue:43653", + "accept": false, + "reason": "Different bugs: Bark CPU offload device mismatch vs BigBirdTokenizer special-token decode/registration issue." + }, + { + "left": "issue:43421", + "right": "issue:43927", + "accept": false, + "reason": "One is runtime post-processor updates for special tokens; the other is saving/loading custom token IDs in `DiaConfig`." + }, + { + "left": "issue:34567", + "right": "issue:35707", + "accept": false, + "reason": "Trainer state token counting and progressive generation with `inputs_embeds`/`past_key_values` are unrelated code paths." + }, + { + "left": "issue:44368", + "right": "issue:45357", + "accept": false, + "reason": "A warning about `tie_word_embeddings` in Qwen3.5 fine-tuning is not the same bug as incorrect visual encoder keys being saved in `save_pretrained`." + }, + { + "left": "issue:41720", + "right": "issue:43746", + "accept": false, + "reason": "Different failures: Qwen3 auto device mapping on A800 vs GraniteSpeech PEFT adapter loading from local checkpoints." + }, + { + "left": "issue:43873", + "right": "issue:45704", + "accept": false, + "reason": "Quantization/offloading behavior and a T5 RMSNorm memory leak are separate issues." + }, + { + "left": "issue:44451", + "right": "issue:44991", + "accept": false, + "reason": "Both are tokenizer-loading regressions, but for different models and likely different root causes." + }, + { + "left": "issue:44898", + "right": "issue:45072", + "accept": false, + "reason": "Perceiver resolution handling and bfloat16 dtype mismatches in SwitchTransformers/TimmWrapperModel are unrelated." + }, + { + "left": "issue:44062", + "right": "issue:44451", + "accept": false, + "reason": "`AddedToken(..., special=...)` argument duplication is a different tokenizer bug from loading `vesteinn/ScandiBERT`." + }, + { + "left": "issue:45588", + "right": "issue:45759", + "accept": false, + "reason": "Flash-attention `s_aux=None` handling and Qwen `text_config` unwrapping are different code-path problems." + }, + { + "left": "issue:33666", + "right": "issue:35707", + "accept": false, + "reason": "Multi-GPU training for Qwen2-VL is unrelated to progressive generation with embeddings and cache reuse." + }, + { + "left": "issue:44987", + "right": "issue:45042", + "accept": false, + "reason": "A model-loading failure for `physical-intelligence/fast` is not the same as PIL image processors incorrectly requiring torchvision." + }, + { + "left": "issue:42886", + "right": "issue:44843", + "accept": true, + "reason": "Same HF_HUB_OFFLINE tokenizer regression; 44843 identifies the concrete cause (`model_info()` called unconditionally in `_patch_mistral_regex`) behind 42886's cache/offline failure." + }, + { + "left": "issue:43531", + "right": "issue:44589", + "accept": false, + "reason": "`sliding_window` in Qwen3-MoE and missing `Float8_e4m3fnStorage` lookup are unrelated." + }, + { + "left": "issue:39692", + "right": "issue:44610", + "accept": false, + "reason": "SigLIP2 doc-example errors and OmDet-Turbo processor/model resolution mismatch are different model-specific issues." + }, + { + "left": "issue:43381", + "right": "issue:43844", + "accept": false, + "reason": "Gradient checkpointing being disallowed in eval mode is not the same as gradients exploding under ZeRO-3." + }, + { + "left": "issue:43421", + "right": "issue:44186", + "accept": false, + "reason": "Runtime special-token post-processor updates and LayoutLMv2Tokenizer NER/padding crashes are separate tokenizer problems." + }, + { + "left": "issue:44918", + "right": "issue:44936", + "accept": false, + "reason": "Qwen3.5 embedding unpacking with TRL SFT and `trainer.evaluate()` failing after `trainer.train()` are different trainer/model interactions." + }, + { + "left": "issue:45081", + "right": "issue:45356", + "accept": false, + "reason": "Both touch `fix_mistral_regex`, but one is a hard crash in `_patch_mistral_regex` while the other is a Kimi-K2.5 codec-handling regression with a misleading warning; not the same concrete bug." + }, + { + "left": "issue:44625", + "right": "issue:45479", + "accept": false, + "reason": "`num_labels` propagation in Qwen3.5 configs and degenerate zero loss for `num_labels=1` are distinct classification bugs." + }, + { + "left": "issue:43525", + "right": "issue:43531", + "accept": false, + "reason": "Missing `pad_token_id` on `Llama4Config` is unrelated to Qwen3-MoE `sliding_window` behavior." + }, + { + "left": "issue:43782", + "right": "issue:45381", + "accept": false, + "reason": "Qwen3VL `weight_only=True` loading failure and Qwen2.5-VL video `vision_position_ids` are different multimodal bugs." + }, + { + "left": "issue:43994", + "right": "issue:44671", + "accept": false, + "reason": "SigLIP2 nonsensical output with `AutoModel`/pipeline is not the same as CamemBERT masked-LM prediction regression." + }, + { + "left": "issue:34634", + "right": "issue:41720", + "accept": false, + "reason": "BarkProcessor `voice_preset` failure and Qwen3 auto device-mapping cuda assert are unrelated." + }, + { + "left": "issue:43381", + "right": "issue:45704", + "accept": false, + "reason": "Eval-mode gradient checkpointing and a T5 apex RMSNorm memory leak are separate issues." + }, + { + "left": "issue:42915", + "right": "issue:45759", + "accept": false, + "reason": "Qwen3Moe failing with `FineGrainedFP8Config` is unrelated to `AutoModelForCausalLM.from_config` not unwrapping `text_config`." + }, + { + "left": "issue:43596", + "right": "issue:44220", + "accept": false, + "reason": "A Zero-3/BertModel indexing crash and `_torch_extract_fbank_features()` are not the same bug." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet and stays conservative: it treats the cluster as mostly heterogeneous and only marks the two strongest title-aligned pairs as duplicates (`44514/44545` and `42886/44843`). The canonical anchor choice for `44843` is supported by the issue title, and the false soft-edge rejections do not overstate similarity." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 13, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-387", + "nodes": [ + "issue:30064", + "issue:36010", + "issue:42915", + "issue:43334", + "issue:43335", + "issue:43421", + "issue:43531", + "issue:43540", + "issue:43550", + "issue:43575", + "issue:43577", + "issue:43618", + "issue:43643", + "issue:43653", + "issue:43697", + "issue:43716", + "issue:43720", + "issue:43746", + "issue:43756", + "issue:43782", + "issue:43818", + "issue:43819", + "issue:43873", + "issue:44008", + "issue:44062", + "issue:44077", + "issue:44186", + "issue:44220", + "issue:44448", + "issue:44488", + "issue:44561", + "issue:44625", + "issue:44661", + "issue:44671", + "issue:44811", + "issue:44843", + "issue:44877", + "issue:44898", + "issue:44977", + "issue:45020", + "issue:45072", + "issue:45081", + "issue:45183", + "issue:45381", + "issue:45479", + "issue:45704", + "issue:45759" + ], + "soft_pairs": [ + "issue:45183|issue:45479", + "issue:43540|issue:44008", + "issue:44220|issue:44661", + "issue:43618|issue:43697", + "issue:43720|issue:43819", + "issue:43575|issue:45704", + "issue:44077|issue:44625", + "issue:30064|issue:36010", + "issue:43335|issue:43756", + "issue:43334|issue:45759", + "issue:43653|issue:43720", + "issue:43716|issue:44008", + "issue:42915|issue:43746", + "issue:43746|issue:45381", + "issue:43746|issue:43873", + "issue:43550|issue:45072", + "issue:43550|issue:43819", + "issue:44186|issue:44898", + "issue:43421|issue:44811", + "issue:44843|issue:45081", + "issue:44062|issue:44488", + "issue:43643|issue:44877", + "issue:43531|issue:43643", + "issue:43577|issue:43818", + "issue:43540|issue:45381", + "issue:43653|issue:43819", + "issue:44448|issue:44671", + "issue:43782|issue:44977", + "issue:44561|issue:45020" + ], + "prepared_review_unit_hash": "af742e957b7802b1094e04e82734b3a97daa4b98caccb73116303e28cb2e9970", + "component_budget": { + "node_count": 387, + "item_count": 387, + "soft_pair_count": 1455, + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 29, + "serialized_chars": 36139, + "estimated_input_tokens": 9035, + "estimated_eval_tokens": 18326 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a grab bag of unrelated Transformers bug reports. The closest thing to a center is the broad remote-code/config-loading regression theme, but most soft pairs are not true duplicates and should stay separate.", + "confidence": 0.36, + "canonical_issue_reason": "issue:45020 is the broadest and most central issue here: it captures the general class of recent-version regressions affecting remote_code models, which overlaps the strongest theme in the cluster.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45020 is the best representative issue because it is the most general umbrella report and can subsume adjacent config/remote-code loading regressions better than the narrower model-specific bugs.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45183", + "right": "issue:45479", + "accept": false, + "reason": "Different failures: input validation/XOR error messaging vs degenerate zero-loss classification behavior." + }, + { + "left": "issue:43540", + "right": "issue:44008", + "accept": false, + "reason": "Different models and code paths: Qwen3OmniMoe video processing vs Gemma 3n audio tensor name collision." + }, + { + "left": "issue:44220", + "right": "issue:44661", + "accept": false, + "reason": "Unrelated tokenizer/audio feature bug vs model-addition failure in TOKENIZER_MAPPING_NAMES." + }, + { + "left": "issue:43618", + "right": "issue:43697", + "accept": false, + "reason": "Different regressions: CLIPOutput attentions field omission vs RTDetrV2 output mismatch in v5." + }, + { + "left": "issue:43720", + "right": "issue:43819", + "accept": false, + "reason": "Different subsystems: BitNet packed-weight loading vs DAC latent decoding/STE mismatch." + }, + { + "left": "issue:43575", + "right": "issue:45704", + "accept": false, + "reason": "Qwen2 tensor-parallel OOM is unrelated to T5's apex RMSNorm memory leak." + }, + { + "left": "issue:44077", + "right": "issue:44625", + "accept": false, + "reason": "PatchTSMixer post_init API constraint and Qwen3.5 num_labels propagation are different bugs." + }, + { + "left": "issue:30064", + "right": "issue:36010", + "accept": false, + "reason": "Image processor void segmentation handling is unrelated to the GenerationMixin import error." + }, + { + "left": "issue:43335", + "right": "issue:43756", + "accept": false, + "reason": "SwitchTransformers sparse-layer creation and Smollm3 RoPE layer dropping are unrelated architecture/config bugs." + }, + { + "left": "issue:43334", + "right": "issue:45759", + "accept": false, + "reason": "Both involve Qwen composite configs, but one is a missing pad_token_id at load time while the other is from_config unwrapping behavior; too different to merge." + }, + { + "left": "issue:43653", + "right": "issue:43720", + "accept": false, + "reason": "BigBird special-token registration and BitNet accelerate loading are unrelated." + }, + { + "left": "issue:43716", + "right": "issue:44008", + "accept": false, + "reason": "Mistral-3 dtype mismatch and Gemma 3n tensor collision are different model bugs." + }, + { + "left": "issue:42915", + "right": "issue:43746", + "accept": false, + "reason": "Qwen3Moe FP8 config failure and GraniteSpeech PEFT checkpoint loading are unrelated." + }, + { + "left": "issue:43746", + "right": "issue:45381", + "accept": false, + "reason": "PEFT local checkpoint loading and qwen2.5-vl video position IDs are different code paths." + }, + { + "left": "issue:43746", + "right": "issue:43873", + "accept": false, + "reason": "Checkpoint loading with PEFT is unrelated to quantization/offloading behavior." + }, + { + "left": "issue:43550", + "right": "issue:45072", + "accept": false, + "reason": "torch.compile/SDPA failure in Bamba and dtype mismatch in SwitchTransformers/TimmWrapperModel are different issues." + }, + { + "left": "issue:43550", + "right": "issue:43819", + "accept": false, + "reason": "Different models and failures: compile-time SDPA bug vs DAC latent conversion mismatch." + }, + { + "left": "issue:44186", + "right": "issue:44898", + "accept": false, + "reason": "Tokenizer crash on NER padding/truncation is unrelated to Perceiver image-size handling." + }, + { + "left": "issue:43421", + "right": "issue:44811", + "accept": false, + "reason": "Runtime special-token post-processor updates and Whisper batch_decode skip_special_tokens behavior are not the same bug." + }, + { + "left": "issue:44843", + "right": "issue:45081", + "accept": false, + "reason": "Both touch _patch_mistral_regex, but one is an offline model_info call and the other is an attribute crash; related area, not the same concrete bug." + }, + { + "left": "issue:44062", + "right": "issue:44488", + "accept": false, + "reason": "AddedToken argument duplication is unrelated to a model-not-loading report." + }, + { + "left": "issue:43643", + "right": "issue:44877", + "accept": false, + "reason": "Both are config-related, but one is missing fields from trust_remote_code and the other is strict GraniteSpeech config loading; not the same defect." + }, + { + "left": "issue:43531", + "right": "issue:43643", + "accept": false, + "reason": "Qwen3-MoE sliding_window behavior and remote_code config field loss are unrelated." + }, + { + "left": "issue:43577", + "right": "issue:43818", + "accept": false, + "reason": "BLIP2 dtype propagation and Video-LLaVA missing temporal attention/shared weights are different model issues." + }, + { + "left": "issue:43540", + "right": "issue:45381", + "accept": false, + "reason": "Both involve video inputs, but they affect different Qwen model families and different failures." + }, + { + "left": "issue:43653", + "right": "issue:43819", + "accept": false, + "reason": "Tokenizer special-token registration and DAC forward/from_latents mismatch are unrelated." + }, + { + "left": "issue:44448", + "right": "issue:44671", + "accept": false, + "reason": "Two separate v5 output-regression reports for different models; not one concrete bug." + }, + { + "left": "issue:43782", + "right": "issue:44977", + "accept": false, + "reason": "Qwen3VL weight_only loading error and Qwen3.5 flash-attention generation issue are different problems." + }, + { + "left": "issue:44561", + "right": "issue:45020", + "accept": false, + "reason": "44561 is one specific remote_code regression cause; 45020 is a broad umbrella report, so they are related but not the same concrete bug." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 14, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-387", + "nodes": [ + "issue:30064", + "issue:34634", + "issue:36010", + "issue:36331", + "issue:42175", + "issue:43262", + "issue:43322", + "issue:43329", + "issue:43404", + "issue:43452", + "issue:43526", + "issue:43577", + "issue:43582", + "issue:43606", + "issue:43644", + "issue:43650", + "issue:43697", + "issue:43720", + "issue:43749", + "issue:43782", + "issue:43819", + "issue:43825", + "issue:43827", + "issue:43906", + "issue:43994", + "issue:44060", + "issue:44186", + "issue:44368", + "issue:44442", + "issue:44485", + "issue:44509", + "issue:44610", + "issue:44671", + "issue:44805", + "issue:44821", + "issue:44857", + "issue:44936", + "issue:44991", + "issue:45072", + "issue:45081", + "issue:45084", + "issue:45406", + "issue:45459", + "issue:45538", + "issue:45600", + "issue:45685", + "issue:45701", + "issue:45803" + ], + "soft_pairs": [ + "issue:43322|issue:45685", + "issue:43825|issue:43994", + "issue:44821|issue:45084", + "issue:30064|issue:36331", + "issue:43825|issue:43827", + "issue:44485|issue:44509", + "issue:43577|issue:43582", + "issue:44186|issue:44442", + "issue:43606|issue:45072", + "issue:34634|issue:36010", + "issue:43452|issue:44991", + "issue:44805|issue:44936", + "issue:44610|issue:45072", + "issue:42175|issue:45081", + "issue:43650|issue:43906", + "issue:43720|issue:45072", + "issue:45081|issue:45406", + "issue:44671|issue:45701", + "issue:43404|issue:43782", + "issue:44060|issue:44368", + "issue:43582|issue:44857", + "issue:45459|issue:45803", + "issue:43262|issue:43819", + "issue:43526|issue:43697", + "issue:45084|issue:45538", + "issue:43329|issue:45600", + "issue:43644|issue:43749" + ], + "prepared_review_unit_hash": "12b4eac731f53f2a701f03b34234840bb152335a1eb8dc049b62ae6c1782bd59", + "component_budget": { + "node_count": 387, + "item_count": 387, + "soft_pair_count": 1455, + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 27, + "serialized_chars": 36025, + "estimated_input_tokens": 9007, + "estimated_eval_tokens": 18270 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is highly heterogeneous: most items are unrelated transformer/model regressions, tokenizer/load errors, or docs cleanup issues. None of the soft-edge pairs look like the same underlying bug/change, so all should be rejected.", + "confidence": 0.94, + "canonical_issue_reason": "issue:43749 is the strongest standalone representative in this set: it is a concrete, high-impact regression with substantial discussion and inbound references. It is not a duplicate of the other items, but it is the best anchor issue for the cluster.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43749 is the most suitable global issue representative because it has the most signal (discussion, refs, and a clearly defined failure mode) among otherwise mixed topics.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43322", + "right": "issue:45685", + "accept": false, + "reason": "Different failures: Llava Next model load segfault vs MPS histogram kernel not implemented for int." + }, + { + "left": "issue:43825", + "right": "issue:43994", + "accept": false, + "reason": "One is a v5 pipeline error-message regression; the other is nonsensical outputs for SigLIP2. Different bugs and code paths." + }, + { + "left": "issue:44821", + "right": "issue:45084", + "accept": false, + "reason": "AutoImageProcessor URL loading and template-node compilation are unrelated." + }, + { + "left": "issue:30064", + "right": "issue:36331", + "accept": false, + "reason": "Void segmentation map handling and CustomTrainer.compute_loss signature mismatch are separate issues." + }, + { + "left": "issue:43825", + "right": "issue:43827", + "accept": false, + "reason": "Related v5 pipeline-removal theme, but one is runtime messaging and the other is docs cleanup; not the same bug." + }, + { + "left": "issue:44485", + "right": "issue:44509", + "accept": false, + "reason": "GLM-5 RoPE implementation discussion vs docs mentioning removed pipeline tasks. Different changes." + }, + { + "left": "issue:43577", + "right": "issue:43582", + "accept": false, + "reason": "BLIP2 dtype propagation and Apple Silicon warmup TypeError are unrelated." + }, + { + "left": "issue:44186", + "right": "issue:44442", + "accept": false, + "reason": "LayoutLMv2 tokenizer crashes and FastSpeech2ConformerTokenizer loading failure are different tokenizer bugs." + }, + { + "left": "issue:43606", + "right": "issue:45072", + "accept": false, + "reason": "CPU offload device mismatch for Bark-small vs dtype mismatch in SwitchTransformers/TimmWrapperModel." + }, + { + "left": "issue:34634", + "right": "issue:36010", + "accept": false, + "reason": "BarkProcessor voice_preset behavior and GenerationMixin import failure are unrelated." + }, + { + "left": "issue:43452", + "right": "issue:44991", + "accept": false, + "reason": "GGUF loading breakage and tokenizer loading failure for est-roberta are different regressions." + }, + { + "left": "issue:44805", + "right": "issue:44936", + "accept": false, + "reason": "Mask-shape IndexError and trainer.evaluate() after train() failure are distinct problems." + }, + { + "left": "issue:44610", + "right": "issue:45072", + "accept": false, + "reason": "OmDet input-size mismatch and dtype mismatch in inference are not the same code-path bug." + }, + { + "left": "issue:42175", + "right": "issue:45081", + "accept": false, + "reason": "Missing TensorFlow in torch install and Mistral tokenizer backend attribute error are unrelated." + }, + { + "left": "issue:43650", + "right": "issue:43906", + "accept": false, + "reason": "The placeholder issue and the isolated reproduction report do not describe the same concrete failure." + }, + { + "left": "issue:43720", + "right": "issue:45072", + "accept": false, + "reason": "BitNet packed-weights loading and SwitchTransformers/TimmWrapper dtype mismatch are different loading issues." + }, + { + "left": "issue:45081", + "right": "issue:45406", + "accept": false, + "reason": "Mistral tokenizer backend_tokenizer attribute error and Gemma4Processor missing _tokenizer are different attribute/access bugs." + }, + { + "left": "issue:44671", + "right": "issue:45701", + "accept": false, + "reason": "CamemBERT masked-LM regression and a broad tokenization-version report are not the same underlying bug." + }, + { + "left": "issue:43404", + "right": "issue:43782", + "accept": false, + "reason": "Mistral3 weight tying and Qwen3VL weight_only load error are unrelated model-specific issues." + }, + { + "left": "issue:44060", + "right": "issue:44368", + "accept": false, + "reason": "Qwen3-Next tied-weights warning bug and ms-swift LoRA warning about tie_word_embeddings are different problems." + }, + { + "left": "issue:43582", + "right": "issue:44857", + "accept": false, + "reason": "Apple Silicon warmup TypeError and LwDetrImageLoss AMP crash are unrelated runtime failures." + }, + { + "left": "issue:45459", + "right": "issue:45803", + "accept": false, + "reason": "Both involve bad exception handling, but they affect different functions and errors; not the same bug." + }, + { + "left": "issue:43262", + "right": "issue:43819", + "accept": false, + "reason": "Audio chat-template sampling-rate defaulting and DAC.from_latents STE mismatch are unrelated." + }, + { + "left": "issue:43526", + "right": "issue:43697", + "accept": false, + "reason": "BeitImageProcessorFast label reduction and RTDetrV2 v5 output differences are different model/processor regressions." + }, + { + "left": "issue:45084", + "right": "issue:45538", + "accept": false, + "reason": "Template compilation TypeError and CLIPTokenizer max-length sentinel are unrelated." + }, + { + "left": "issue:43329", + "right": "issue:45600", + "accept": false, + "reason": "Undefined variables in multimodal token counting and stale auto_mappings references are separate issues." + }, + { + "left": "issue:43644", + "right": "issue:43749", + "accept": false, + "reason": "Both touch loading, but one reports junk-filled non-persistent buffers and the other FSDP CPU RAM efficient loading; not the same concrete bug." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 15, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-387", + "nodes": [ + "issue:34567", + "issue:34634", + "issue:36010", + "issue:36331", + "issue:38175", + "issue:39692", + "issue:40990", + "issue:43257", + "issue:43262", + "issue:43329", + "issue:43450", + "issue:43452", + "issue:43454", + "issue:43526", + "issue:43531", + "issue:43550", + "issue:43577", + "issue:43582", + "issue:43638", + "issue:43643", + "issue:43645", + "issue:43653", + "issue:43697", + "issue:43720", + "issue:43746", + "issue:43749", + "issue:43819", + "issue:43854", + "issue:43927", + "issue:43931", + "issue:44291", + "issue:44479", + "issue:44554", + "issue:44568", + "issue:44589", + "issue:44610", + "issue:44811", + "issue:44849", + "issue:44912", + "issue:44936", + "issue:44977", + "issue:45042", + "issue:45356", + "issue:45357", + "issue:45488", + "issue:45701", + "issue:45759" + ], + "soft_pairs": [ + "issue:43653|issue:44912", + "issue:43638|issue:44936", + "issue:43746|issue:44811", + "issue:44589|issue:45042", + "issue:38175|issue:44610", + "issue:43454|issue:43577", + "issue:43854|issue:44977", + "issue:36010|issue:43645", + "issue:43526|issue:44554", + "issue:34634|issue:36331", + "issue:39692|issue:44912", + "issue:44568|issue:45488", + "issue:43819|issue:43927", + "issue:43257|issue:43329", + "issue:43720|issue:44610", + "issue:43749|issue:44291", + "issue:43550|issue:44912", + "issue:43262|issue:44811", + "issue:44568|issue:45356", + "issue:43531|issue:45042", + "issue:43643|issue:45759", + "issue:36010|issue:36331", + "issue:43329|issue:43450", + "issue:34567|issue:40990", + "issue:43329|issue:44479", + "issue:44849|issue:45701", + "issue:43257|issue:43931", + "issue:43697|issue:45357", + "issue:43452|issue:43582" + ], + "prepared_review_unit_hash": "fdb33c136ddae4466fa1dd1f1a07adfa05963ab1dbbd453cd90b226d48c70f0f", + "component_budget": { + "node_count": 387, + "item_count": 387, + "soft_pair_count": 1455, + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 29, + "serialized_chars": 36224, + "estimated_input_tokens": 9056, + "estimated_eval_tokens": 18368 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is mostly heterogeneous; the soft edges share broad subsystems but not the same concrete bug/change, so none should be merged. If forced to pick an issue representative, the tokenizer regression in issue 45488 is the clearest actionable exemplar, but it does not unify the whole set.", + "confidence": 0.68, + "canonical_issue_reason": "Issue 45488 is the strongest single representative because it describes a concrete, user-visible tokenizer regression with broad impact; the other items are too mixed to serve as one duplicate anchor.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 45488 is the best overall issue candidate: specific enough to be actionable, broad enough to represent a real regression class, and open rather than already closed.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43653", + "right": "issue:44912", + "accept": false, + "reason": "Different bugs: tokenizer special-token registration vs quantization/load fallback." + }, + { + "left": "issue:43638", + "right": "issue:44936", + "accept": false, + "reason": "Different training failures: ZeRO-3 empty-index crash vs post-train evaluate failure." + }, + { + "left": "issue:43746", + "right": "issue:44811", + "accept": false, + "reason": "Unrelated code paths: PEFT checkpoint loading vs Whisper batch decode flag handling." + }, + { + "left": "issue:44589", + "right": "issue:45042", + "accept": false, + "reason": "Different runtime problems: missing Float8 storage vs PIL image processors requiring torchvision." + }, + { + "left": "issue:38175", + "right": "issue:44610", + "accept": false, + "reason": "Different models and failures: zero probabilities vs processor/model input-size mismatch." + }, + { + "left": "issue:43454", + "right": "issue:43577", + "accept": false, + "reason": "Different multimodal load bugs: lm_head tying vs dtype propagation to Blip2/QFormer." + }, + { + "left": "issue:43854", + "right": "issue:44977", + "accept": false, + "reason": "Different model-specific regressions and code paths; not the same generation bug." + }, + { + "left": "issue:36010", + "right": "issue:43645", + "accept": false, + "reason": "Import-path regression vs Jupyter custom-model initialization issue." + }, + { + "left": "issue:43526", + "right": "issue:44554", + "accept": false, + "reason": "Image label reduction bug is unrelated to attention correctness on MPS." + }, + { + "left": "issue:34634", + "right": "issue:36331", + "accept": false, + "reason": "Different APIs entirely: BarkProcessor voice preset vs CustomTrainer compute_loss signature." + }, + { + "left": "issue:39692", + "right": "issue:44912", + "accept": false, + "reason": "Docs/example errors and quantization failure are not the same underlying bug." + }, + { + "left": "issue:44568", + "right": "issue:45488", + "accept": false, + "reason": "Both are tokenizer regressions, but they affect different tokenizers and different code paths." + }, + { + "left": "issue:43819", + "right": "issue:43927", + "accept": false, + "reason": "Different bugs: DAC forward/from_latents mismatch vs DiaConfig save/load token-ID loss." + }, + { + "left": "issue:43257", + "right": "issue:43329", + "accept": false, + "reason": "Qwen3 MoE weight conversion during loading is unrelated to multimodal video token counting." + }, + { + "left": "issue:43720", + "right": "issue:44610", + "accept": false, + "reason": "Packed-weight unpacking during accelerate loading is unrelated to OmDet processor image sizing." + }, + { + "left": "issue:43749", + "right": "issue:44291", + "accept": false, + "reason": "Different model-loading failures: FSDP CPU RAM efficient loading vs unexpected init arg in init_empty_weights." + }, + { + "left": "issue:43550", + "right": "issue:44912", + "accept": false, + "reason": "torch.compile/SDPA bug is unrelated to MXFP4 quantization fallback." + }, + { + "left": "issue:43262", + "right": "issue:44811", + "accept": false, + "reason": "Audio chat-template sampling-rate default is a different processor bug than Whisper decode flags." + }, + { + "left": "issue:44568", + "right": "issue:45356", + "accept": false, + "reason": "Different tokenizer regressions affecting different models and symptoms." + }, + { + "left": "issue:43531", + "right": "issue:45042", + "accept": false, + "reason": "Sliding-window model behavior is unrelated to the PIL backend torchvision dependency issue." + }, + { + "left": "issue:43643", + "right": "issue:45759", + "accept": false, + "reason": "Missing fields from trust_remote_code config loading is not the same as unwrapping text_config in from_config." + }, + { + "left": "issue:36010", + "right": "issue:36331", + "accept": false, + "reason": "Import failure from generation module is unrelated to a custom trainer loss signature change." + }, + { + "left": "issue:43329", + "right": "issue:43450", + "accept": false, + "reason": "Both concern video handling, but one is undefined helpers in token counting while the other is batched output shape; not the same bug." + }, + { + "left": "issue:34567", + "right": "issue:40990", + "accept": false, + "reason": "TrainerState token accounting bug and high perplexity on gpt-oss are different problems." + }, + { + "left": "issue:43329", + "right": "issue:44479", + "accept": false, + "reason": "Related video area, but one is an undefined-helper token-count bug and the other is a broader v5.3 video-input regression." + }, + { + "left": "issue:44849", + "right": "issue:45701", + "accept": false, + "reason": "Qwen3.5 hidden-state behavior and generic tokenization-version changes are not the same concrete issue." + }, + { + "left": "issue:43257", + "right": "issue:43931", + "accept": false, + "reason": "Different load failures: MoE weight conversion with accelerate/deepspeed vs Qwen3-VL shape mismatch." + }, + { + "left": "issue:43697", + "right": "issue:45357", + "accept": false, + "reason": "Different regressions: RTDetrV2 output divergence vs Qwen3.5 save_pretrained visual-key bug." + }, + { + "left": "issue:43452", + "right": "issue:43582", + "accept": false, + "reason": "Tokenizer/model loading with gguf_file is unrelated to Apple Silicon warmup TypeError." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 16, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-387", + "nodes": [ + "issue:33357", + "issue:34567", + "issue:36010", + "issue:36331", + "issue:38175", + "issue:41720", + "issue:42175", + "issue:42915", + "issue:43257", + "issue:43317", + "issue:43452", + "issue:43493", + "issue:43526", + "issue:43540", + "issue:43577", + "issue:43643", + "issue:43646", + "issue:43696", + "issue:43742", + "issue:43749", + "issue:43756", + "issue:43824", + "issue:43906", + "issue:43927", + "issue:43994", + "issue:44062", + "issue:44291", + "issue:44479", + "issue:44488", + "issue:44560", + "issue:44737", + "issue:44756", + "issue:44821", + "issue:44898", + "issue:44912", + "issue:44987", + "issue:44991", + "issue:45042", + "issue:45072", + "issue:45081", + "issue:45137", + "issue:45375", + "issue:45381", + "issue:45405", + "issue:45479", + "issue:45538", + "issue:45588" + ], + "soft_pairs": [ + "issue:41720|issue:43540", + "issue:44291|issue:45042", + "issue:42915|issue:43742", + "issue:33357|issue:34567", + "issue:43317|issue:45538", + "issue:44479|issue:45381", + "issue:44987|issue:45538", + "issue:43540|issue:44560", + "issue:42915|issue:44821", + "issue:43540|issue:44479", + "issue:43927|issue:44912", + "issue:43577|issue:45072", + "issue:43257|issue:43756", + "issue:42915|issue:45538", + "issue:43257|issue:43696", + "issue:45137|issue:45405", + "issue:38175|issue:44898", + "issue:44737|issue:45588", + "issue:44062|issue:44991", + "issue:43742|issue:44291", + "issue:43906|issue:45381", + "issue:43646|issue:43749", + "issue:43749|issue:44756", + "issue:34567|issue:36331", + "issue:44062|issue:45081", + "issue:43643|issue:45375", + "issue:36010|issue:42175", + "issue:43493|issue:43994", + "issue:43452|issue:44488", + "issue:42175|issue:43824", + "issue:43526|issue:45479", + "issue:44291|issue:44821" + ], + "prepared_review_unit_hash": "3386c60f1e048385fabff1ba9500f086983d44931a2399d5e755fc2cbb189e47", + "component_budget": { + "node_count": 387, + "item_count": 387, + "soft_pair_count": 1455, + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 32, + "serialized_chars": 36869, + "estimated_input_tokens": 9218, + "estimated_eval_tokens": 18692 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This is a noisy mixed cluster, but two duplicate-like pockets stand out: Qwen video-input regressions and a SigLIP2 output mismatch. Most other pairs are false positives across unrelated bugs.", + "confidence": 0.74, + "canonical_issue_reason": "issue:44479 is the broadest concrete regression in the cluster: it describes the shared Qwen video-input failure across multiple related model families, and it has the clearest duplicate-style overlap with issue:45381.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:44479 is the best representative issue because it is a broad, user-facing regression with multiple related variants and strong duplicate potential.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:41720", + "right": "issue:43540", + "accept": false, + "reason": "Different Qwen failures: auto device-mapping CUDA assert vs. video-input ValueError in a different model path." + }, + { + "left": "issue:44291", + "right": "issue:45042", + "accept": false, + "reason": "Unrelated problems: init_empty_weights argument handling vs. PIL backend image-processor dependency behavior." + }, + { + "left": "issue:42915", + "right": "issue:43742", + "accept": false, + "reason": "Different model/loading failures: FineGrainedFP8Config on Qwen3Moe vs. key error loading MobileLLM-125M." + }, + { + "left": "issue:33357", + "right": "issue:34567", + "accept": false, + "reason": "Completely different subsystems: MacOS CLIP bus error vs. TrainerState token accounting." + }, + { + "left": "issue:43317", + "right": "issue:45538", + "accept": false, + "reason": "Different code paths: dequantized device-map/offload loading bug vs. CLIPTokenizer max-length constant." + }, + { + "left": "issue:44479", + "right": "issue:45381", + "accept": true, + "reason": "Same underlying Qwen video-input regression: both point to the wrong video position-id handling in 5.3.0, with 44479 being the broader report." + }, + { + "left": "issue:44987", + "right": "issue:45538", + "accept": false, + "reason": "Different failures: loading a specific model on transformers>=5.1.0 vs. tokenizer max-length behavior." + }, + { + "left": "issue:43540", + "right": "issue:44560", + "accept": false, + "reason": "Both are video-related, but the model families and failure modes differ; not clearly the same bug." + }, + { + "left": "issue:42915", + "right": "issue:44821", + "accept": false, + "reason": "Unrelated model quantization/loading issue vs. AutoImageProcessor URL loading problem." + }, + { + "left": "issue:43540", + "right": "issue:44479", + "accept": false, + "reason": "Both involve video inputs, but one is a Qwen3OmniMoe processing ValueError and the other is a broader Qwen video regression; not the same concrete bug." + }, + { + "left": "issue:43927", + "right": "issue:44912", + "accept": false, + "reason": "Different features: DiaConfig save/load token IDs vs. MXFP4 quantization loading fallback." + }, + { + "left": "issue:43577", + "right": "issue:45072", + "accept": false, + "reason": "Both mention dtype, but the affected models and failure mechanisms are different." + }, + { + "left": "issue:43257", + "right": "issue:43756", + "accept": false, + "reason": "Different model-specific issues: Qwen3 MoE weight conversion vs. Smollm3 RoPE-layer dropping." + }, + { + "left": "issue:42915", + "right": "issue:45538", + "accept": false, + "reason": "Different areas entirely: Qwen3Moe FP8 loading vs. CLIPTokenizer max-length default." + }, + { + "left": "issue:43257", + "right": "issue:43696", + "accept": false, + "reason": "Different failure modes and models: MoE weight conversion vs. GPT-oss CUDA OOM." + }, + { + "left": "issue:45137", + "right": "issue:45405", + "accept": false, + "reason": "No relation: DeepSpeed ZeRO3 deque error vs. PEFT version pinning." + }, + { + "left": "issue:38175", + "right": "issue:44898", + "accept": false, + "reason": "Different models and symptoms: SigLIP2 zero probabilities vs. Perceiver resolution/interpolation failure." + }, + { + "left": "issue:44737", + "right": "issue:45588", + "accept": false, + "reason": "Unrelated implementation bugs: XLNet CPU arange placement vs. flash_attention auxiliary-state crash." + }, + { + "left": "issue:44062", + "right": "issue:44991", + "accept": false, + "reason": "Both tokenizer-related, but one is AddedToken argument duplication and the other is a model-specific tokenizer loading regression." + }, + { + "left": "issue:43742", + "right": "issue:44291", + "accept": false, + "reason": "Different layers: MobileLLM key lookup vs. init_empty_weights/_is_hf_initialized argument handling." + }, + { + "left": "issue:43906", + "right": "issue:45381", + "accept": false, + "reason": "The reproduction issue is too generic and does not match the specific Qwen2.5-VL video-position-id regression tightly enough." + }, + { + "left": "issue:43646", + "right": "issue:43749", + "accept": false, + "reason": "Different core problems: custom model init breakage in v5.0.0 vs. FSDP CPU RAM-efficient loading regression." + }, + { + "left": "issue:43749", + "right": "issue:44756", + "accept": false, + "reason": "Both involve loading/memory, but one is FSDP RAM-efficient loading and the other is mmap on Strix Halo; not the same bug." + }, + { + "left": "issue:34567", + "right": "issue:36331", + "accept": false, + "reason": "Training-state token counting vs. custom trainer compute_loss signature mismatch are unrelated." + }, + { + "left": "issue:44062", + "right": "issue:45081", + "accept": false, + "reason": "Both touch tokenizer internals, but one is AddedToken construction and the other is Mistral regex patching; different code paths." + }, + { + "left": "issue:43643", + "right": "issue:45375", + "accept": false, + "reason": "Similar theme of config-field handling, but they are different bugs and different model/config types." + }, + { + "left": "issue:36010", + "right": "issue:42175", + "accept": false, + "reason": "Completely unrelated: GenerationMixin import path vs. TensorFlow not included in the torch extra." + }, + { + "left": "issue:43493", + "right": "issue:43994", + "accept": true, + "reason": "Same SigLIP2 checkpoint family and same symptom class: incorrect model behavior/output compared to the expected implementation, plausibly fixed by one SigLIP2 code-path correction." + }, + { + "left": "issue:43452", + "right": "issue:44488", + "accept": false, + "reason": "Different loading problems: GGUF tokenizer/model loading vs. inability to load a specific bert model." + }, + { + "left": "issue:42175", + "right": "issue:43824", + "accept": false, + "reason": "Unrelated packaging/back-end inclusion issue vs. a specific model import error." + }, + { + "left": "issue:43526", + "right": "issue:45479", + "accept": false, + "reason": "Different image-processing and classification bugs: BEiT label reduction vs. sequence-classification zero-loss behavior." + }, + { + "left": "issue:44291", + "right": "issue:44821", + "accept": false, + "reason": "No overlap: init_empty_weights argument handling vs. AutoImageProcessor URL loading." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet: the Qwen video-input issues and the SigLIP2 model-output issues are the only plausible duplicate pockets, and the other pairwise rejections are conservative. I don\u2019t see any overstatement that would warrant rejection." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 17, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-387", + "nodes": [ + "issue:24643", + "issue:28282", + "issue:29127", + "issue:34689", + "issue:38175", + "issue:41762", + "issue:42175", + "issue:43295", + "issue:43452", + "issue:43531", + "issue:43577", + "issue:43582", + "issue:43643", + "issue:43653", + "issue:43697", + "issue:43720", + "issue:43742", + "issue:43756", + "issue:43819", + "issue:44466", + "issue:44484", + "issue:44488", + "issue:44514", + "issue:44561", + "issue:44568", + "issue:44589", + "issue:44625", + "issue:44704", + "issue:44792", + "issue:44811", + "issue:44821", + "issue:44843", + "issue:44877", + "issue:44898", + "issue:44936", + "issue:44964", + "issue:44977", + "issue:45005", + "issue:45042", + "issue:45081", + "issue:45216", + "issue:45290", + "issue:45325", + "issue:45356", + "issue:45381", + "issue:45405", + "issue:45600" + ], + "soft_pairs": [ + "issue:44792|issue:44964", + "issue:28282|issue:34689", + "issue:45042|issue:45081", + "issue:45405|issue:45600", + "issue:43756|issue:45325", + "issue:43742|issue:44589", + "issue:44484|issue:44704", + "issue:43531|issue:44821", + "issue:28282|issue:29127", + "issue:43697|issue:45216", + "issue:44514|issue:45381", + "issue:43643|issue:44561", + "issue:44625|issue:44877", + "issue:43582|issue:44977", + "issue:43452|issue:43577", + "issue:43720|issue:44898", + "issue:43452|issue:44843", + "issue:44514|issue:45290", + "issue:44466|issue:45005", + "issue:43819|issue:44811", + "issue:42175|issue:44488", + "issue:41762|issue:44936", + "issue:43295|issue:45356", + "issue:43653|issue:44568", + "issue:24643|issue:29127", + "issue:38175|issue:43295" + ], + "prepared_review_unit_hash": "060ab9fc870c1c9512a455a216de5f197431448956c674e4b101bbaa5b1ba504", + "component_budget": { + "node_count": 387, + "item_count": 387, + "soft_pair_count": 1455, + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 26, + "serialized_chars": 35269, + "estimated_input_tokens": 8818, + "estimated_eval_tokens": 17892 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These are mostly superficially similar bug reports across loader, tokenizer, processor, and config paths, but each pair has a distinct trigger or code path. No soft edges should be merged as duplicates.", + "confidence": 0.78, + "canonical_issue_reason": "issue:44704 is the best representative of the batch because it is a clear, cross-cutting `from_pretrained`/loader-plumbing regression with broader API impact than the model-specific reports.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:44704 is the strongest issue candidate overall: it has a precise failure mode, clear repro scope, and broader relevance than the narrowly model-specific symptom reports.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44792", + "right": "issue:44964", + "accept": false, + "reason": "Both are model-loading failures, but they affect different models and likely different root causes." + }, + { + "left": "issue:28282", + "right": "issue:34689", + "accept": false, + "reason": "One is a missing-PyTorch import error; the other is a Llama vision model-loading regression." + }, + { + "left": "issue:45042", + "right": "issue:45081", + "accept": false, + "reason": "These involve tokenizer/image-processor loading, but the reported failures are different: torchvision dependency vs Mistral regex crash." + }, + { + "left": "issue:45405", + "right": "issue:45600", + "accept": false, + "reason": "These are unrelated maintenance/config issues: PEFT version pinning versus stale auto-mapping references." + }, + { + "left": "issue:43756", + "right": "issue:45325", + "accept": false, + "reason": "Both mention RoPE/positioning in Qwen-family models, but one is about missing RoPE layers and the other about video position-id scaling." + }, + { + "left": "issue:43742", + "right": "issue:44589", + "accept": false, + "reason": "These are different loading/storage failures on different components, with no shared concrete code path." + }, + { + "left": "issue:44484", + "right": "issue:44704", + "accept": false, + "reason": "`max_shard_size` behavior and `cached_file` kwargs propagation are separate `save_pretrained`/load-pretrained concerns." + }, + { + "left": "issue:43531", + "right": "issue:44821", + "accept": false, + "reason": "One is a Qwen3-MoE sliding-window bug; the other is inability to load `AutoImageProcessor` from a URL." + }, + { + "left": "issue:28282", + "right": "issue:29127", + "accept": false, + "reason": "A missing PyTorch dependency error is unrelated to a LayoutLMv3 box-validation error message issue." + }, + { + "left": "issue:43697", + "right": "issue:45216", + "accept": false, + "reason": "One reports different runtime outputs in v5; the other is a regression in saved checkpoint serialization for Qwen3.5." + }, + { + "left": "issue:44514", + "right": "issue:45381", + "accept": false, + "reason": "Both are Qwen2.5-VL-related, but one crashes in chat-template batching while the other has incorrect video vision position IDs." + }, + { + "left": "issue:43643", + "right": "issue:44561", + "accept": false, + "reason": "These are distinct config/utility regressions: missing fields in `AutoConfig` versus removal of `is_torch_fx_available` breaking remote-code models." + }, + { + "left": "issue:44625", + "right": "issue:44877", + "accept": false, + "reason": "Qwen3.5 label propagation and strict granite_speech config loading are different configuration bugs." + }, + { + "left": "issue:43582", + "right": "issue:44977", + "accept": false, + "reason": "Apple Silicon allocator warmup and Qwen3.5 flash-attention generation are unrelated failures." + }, + { + "left": "issue:43452", + "right": "issue:43577", + "accept": false, + "reason": "Both concern model/tokenizer loading, but one is broken by `gguf_file` handling and the other by dtype propagation in Blip2." + }, + { + "left": "issue:43720", + "right": "issue:44898", + "accept": false, + "reason": "These are different model-path bugs: BitNet packed-weight unpacking versus Perceiver interpolation at non-default resolution." + }, + { + "left": "issue:43452", + "right": "issue:44843", + "accept": false, + "reason": "`gguf_file` loading and offline `model_info()` calls in Mistral regex patching are separate failure modes." + }, + { + "left": "issue:44514", + "right": "issue:45290", + "accept": false, + "reason": "Both touch chat templates, but the failures are different: batched `padding=False` versus tool-call assistant messages without content." + }, + { + "left": "issue:44466", + "right": "issue:45005", + "accept": false, + "reason": "Both involve tied weights, but one is about serialization depending on device and the other is a translation-model issue in v5." + }, + { + "left": "issue:43819", + "right": "issue:44811", + "accept": false, + "reason": "DAC latent decoding mismatch and Whisper `batch_decode` ignoring `skip_special_tokens` are unrelated decoding bugs." + }, + { + "left": "issue:42175", + "right": "issue:44488", + "accept": false, + "reason": "A packaging/backend dependency omission is not the same as a model-specific load failure for `cjvt/sleng-bert`." + }, + { + "left": "issue:41762", + "right": "issue:44936", + "accept": false, + "reason": "Gemma3 ZeRO-3 loading and `trainer.evaluate()` after `trainer.train()` are different runtime paths." + }, + { + "left": "issue:43295", + "right": "issue:45356", + "accept": false, + "reason": "These are separate regressions: processor/tokenizer access in one case and Kimi-K2.5 codec/fix-mistral-regex behavior in the other." + }, + { + "left": "issue:43653", + "right": "issue:44568", + "accept": false, + "reason": "BigBird special-token registration and mDeBERTa BOS/EOS insertion are different tokenizer behaviors." + }, + { + "left": "issue:24643", + "right": "issue:29127", + "accept": false, + "reason": "DeepSpeed training shape error and LayoutLMv3 box-validation messaging are not the same bug." + }, + { + "left": "issue:38175", + "right": "issue:43295", + "accept": false, + "reason": "SigLIP2 zero-probability output and a processor/tokenizer regression are unrelated issues." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 18, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-387", + "nodes": [ + "issue:28282", + "issue:30064", + "issue:32090", + "issue:33666", + "issue:34567", + "issue:35141", + "issue:36010", + "issue:41720", + "issue:43122", + "issue:43262", + "issue:43295", + "issue:43329", + "issue:43452", + "issue:43493", + "issue:43504", + "issue:43525", + "issue:43531", + "issue:43550", + "issue:43572", + "issue:43577", + "issue:43653", + "issue:43688", + "issue:43697", + "issue:43720", + "issue:43742", + "issue:43749", + "issue:43825", + "issue:43906", + "issue:43927", + "issue:44361", + "issue:44442", + "issue:44451", + "issue:44464", + "issue:44479", + "issue:44610", + "issue:44743", + "issue:44821", + "issue:44843", + "issue:44849", + "issue:44936", + "issue:44977", + "issue:44991", + "issue:45325", + "issue:45356", + "issue:45357", + "issue:45440", + "issue:45588" + ], + "soft_pairs": [ + "issue:44361|issue:44442", + "issue:43653|issue:44361", + "issue:30064|issue:33666", + "issue:43577|issue:43697", + "issue:44464|issue:44743", + "issue:44843|issue:44991", + "issue:44479|issue:44936", + "issue:41720|issue:43906", + "issue:44977|issue:45356", + "issue:43688|issue:45440", + "issue:43531|issue:43742", + "issue:43577|issue:45588", + "issue:43653|issue:43927", + "issue:44464|issue:44977", + "issue:43550|issue:43927", + "issue:43550|issue:44361", + "issue:32090|issue:34567", + "issue:43122|issue:43906", + "issue:43504|issue:43720", + "issue:43504|issue:43550", + "issue:32090|issue:33666", + "issue:28282|issue:35141", + "issue:43504|issue:44610", + "issue:43452|issue:44451", + "issue:43742|issue:44821", + "issue:41720|issue:44464", + "issue:43329|issue:45325", + "issue:33666|issue:36010", + "issue:43688|issue:43749", + "issue:43525|issue:43572", + "issue:44849|issue:45357", + "issue:43262|issue:43295", + "issue:43493|issue:43825" + ], + "prepared_review_unit_hash": "31333bb96adb2b6e5edc13f3651c1b8c326606a9cef8289d18b5ec3ec0b34e64", + "component_budget": { + "node_count": 387, + "item_count": 387, + "soft_pair_count": 1455, + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 33, + "serialized_chars": 37174, + "estimated_input_tokens": 9294, + "estimated_eval_tokens": 18844 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is mostly a collection of unrelated Transformers bug reports rather than a single duplicate set. I did not find any soft pair that clearly shares the same underlying bug or fix path. The most central issue in the set is the FSDP CPU RAM efficient loading regression.", + "confidence": 0.71, + "canonical_issue_reason": "Issue 43749 has the strongest cluster-centrality signal (highest discussion/inbound refs) and reads like a broader regression report, while the other items are largely distinct model/tokenizer/training bugs.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 43749 is the best anchor for this cluster because it is well-discussed, widely referenced, and represents a substantive regression rather than a narrowly model-specific edge case.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44361", + "right": "issue:44442", + "accept": false, + "reason": "Both are tokenizer-loading failures, but they affect different tokenizer classes and different failure modes; not the same underlying bug." + }, + { + "left": "issue:43653", + "right": "issue:44361", + "accept": false, + "reason": "Different tokenizer/model paths and different symptoms; no evidence they share a common code-path bug." + }, + { + "left": "issue:30064", + "right": "issue:33666", + "accept": false, + "reason": "Image processor void-segmentation handling and Qwen2-VL multi-GPU training are unrelated issues." + }, + { + "left": "issue:43577", + "right": "issue:43697", + "accept": false, + "reason": "One is a dtype propagation problem in Blip2 loading, the other is an output regression in RTDetrV2; not the same fix." + }, + { + "left": "issue:44464", + "right": "issue:44743", + "accept": false, + "reason": "Both involve generation/caching behavior, but they describe different mechanisms and code paths." + }, + { + "left": "issue:44843", + "right": "issue:44991", + "accept": false, + "reason": "Both are tokenizer-loading regressions, but one is specifically offline model_info patching and the other is a different tokenizer failure." + }, + { + "left": "issue:44479", + "right": "issue:44936", + "accept": false, + "reason": "Video-input regression in Qwen VL models and trainer evaluate-after-train failure are unrelated." + }, + { + "left": "issue:41720", + "right": "issue:43906", + "accept": false, + "reason": "The isolated reproduction does not clearly target the same Qwen3 auto-device-map CUDA assert; insufficient evidence of a shared bug." + }, + { + "left": "issue:44977", + "right": "issue:45356", + "accept": false, + "reason": "Qwen3.5 flash-attention generation and Kimi-K2.5 tokenizer codec regression are different problems." + }, + { + "left": "issue:43688", + "right": "issue:45440", + "accept": false, + "reason": "Both concern MoE models, but one is auxiliary-loss normalization and the other is implementation divergence; not the same underlying defect." + }, + { + "left": "issue:43531", + "right": "issue:43742", + "accept": false, + "reason": "Sliding-window behavior in Qwen3-MoE and a key error loading MobileLLM are unrelated." + }, + { + "left": "issue:43577", + "right": "issue:45588", + "accept": false, + "reason": "Dtype propagation in Blip2 loading is unrelated to flash-attention crashing on s_aux=None." + }, + { + "left": "issue:43653", + "right": "issue:43927", + "accept": false, + "reason": "BigBird tokenizer special-token registration and DiaConfig custom token IDs are different loading/tokenization issues." + }, + { + "left": "issue:44464", + "right": "issue:44977", + "accept": false, + "reason": "Both are generation regressions, but one is compiled-forward inconsistency and the other is flash-attention-specific generation behavior." + }, + { + "left": "issue:43550", + "right": "issue:43927", + "accept": false, + "reason": "torch.compile + SDPA failure for Bamba and DiaConfig token-ID persistence are unrelated." + }, + { + "left": "issue:43550", + "right": "issue:44361", + "accept": false, + "reason": "Different models and failure modes; no shared concrete code-path problem." + }, + { + "left": "issue:32090", + "right": "issue:34567", + "accept": false, + "reason": "Trainer broadcast NoneType error and num_input_tokens_seen not updating are unrelated trainer bugs." + }, + { + "left": "issue:43122", + "right": "issue:43906", + "accept": false, + "reason": "Tokenization differences across versions and an isolated reproduction of another issue are not enough to conclude the same bug." + }, + { + "left": "issue:43504", + "right": "issue:43720", + "accept": false, + "reason": "Legacy-field model loading and BitNet packed-weight unpacking are different loading paths and fixes." + }, + { + "left": "issue:43504", + "right": "issue:43550", + "accept": false, + "reason": "Beit preset loading and Bamba torch.compile failure are unrelated." + }, + { + "left": "issue:32090", + "right": "issue:33666", + "accept": false, + "reason": "Trainer NoneType broadcast and Qwen2-VL multi-GPU training are not the same issue." + }, + { + "left": "issue:28282", + "right": "issue:35141", + "accept": false, + "reason": "PyTorch-missing import error and embedding reinitialization after resize are unrelated." + }, + { + "left": "issue:43504", + "right": "issue:44610", + "accept": false, + "reason": "Model preset loading and processor/model input-size mismatch are different bugs." + }, + { + "left": "issue:43452", + "right": "issue:44451", + "accept": false, + "reason": "gguf_file loading breakage and inability to load a specific model are not clearly the same underlying failure." + }, + { + "left": "issue:43742", + "right": "issue:44821", + "accept": false, + "reason": "MobileLLM key error and AutoImageProcessor-from-URL loading are separate code paths." + }, + { + "left": "issue:41720", + "right": "issue:44464", + "accept": false, + "reason": "CUDA assert in auto device mapping and compiled-forward generation inconsistency are unrelated." + }, + { + "left": "issue:43329", + "right": "issue:45325", + "accept": false, + "reason": "Video token counting bug and Qwen2.5-VL rope-index scaling bug are different multimodal issues." + }, + { + "left": "issue:33666", + "right": "issue:36010", + "accept": false, + "reason": "Qwen2-VL multi-GPU training and GenerationMixin import failure are unrelated." + }, + { + "left": "issue:43688", + "right": "issue:43749", + "accept": false, + "reason": "Auxiliary-loss normalization and FSDP CPU RAM efficient loading are different bug classes with no shared fix path." + }, + { + "left": "issue:43525", + "right": "issue:43572", + "accept": false, + "reason": "Both mention missing padding token config fields, but they affect different model configs and are not clearly one bug." + }, + { + "left": "issue:44849", + "right": "issue:45357", + "accept": false, + "reason": "Qwen3.5 hidden-states generation bug and incorrect visual encoder keys on save_pretrained are different regressions." + }, + { + "left": "issue:43262", + "right": "issue:43295", + "accept": false, + "reason": "Audio chat-template sampling-rate default and processor.tokenizer/image passing regression are unrelated processor issues." + }, + { + "left": "issue:43493", + "right": "issue:43825", + "accept": false, + "reason": "SigLIP2 implementation discrepancy and pipeline translation-task messaging are entirely separate concerns." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 19, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-387", + "nodes": [ + "issue:28282", + "issue:32090", + "issue:33357", + "issue:33666", + "issue:36331", + "issue:39692", + "issue:43257", + "issue:43278", + "issue:43295", + "issue:43299", + "issue:43334", + "issue:43493", + "issue:43504", + "issue:43540", + "issue:43572", + "issue:43575", + "issue:43577", + "issue:43606", + "issue:43646", + "issue:43653", + "issue:43688", + "issue:43716", + "issue:43720", + "issue:43756", + "issue:43824", + "issue:43844", + "issue:43927", + "issue:44291", + "issue:44361", + "issue:44423", + "issue:44466", + "issue:44479", + "issue:44488", + "issue:44554", + "issue:44589", + "issue:44625", + "issue:44734", + "issue:44843", + "issue:44913", + "issue:44964", + "issue:44977", + "issue:45083", + "issue:45200", + "issue:45229", + "issue:45405", + "issue:45520", + "issue:45701", + "issue:45704" + ], + "soft_pairs": [ + "issue:43756|issue:44913", + "issue:43720|issue:43927", + "issue:43540|issue:43716", + "issue:43606|issue:43927", + "issue:28282|issue:32090", + "issue:28282|issue:33357", + "issue:43334|issue:43572", + "issue:43577|issue:44554", + "issue:33666|issue:36331", + "issue:43688|issue:45704", + "issue:39692|issue:43493", + "issue:43720|issue:44361", + "issue:44964|issue:45405", + "issue:44479|issue:44625", + "issue:43540|issue:44977", + "issue:43493|issue:43844", + "issue:44488|issue:44843", + "issue:43756|issue:45083", + "issue:45200|issue:45229", + "issue:43278|issue:43844", + "issue:43257|issue:43299", + "issue:43646|issue:44466", + "issue:43824|issue:44589", + "issue:39692|issue:43295", + "issue:43688|issue:45701", + "issue:44589|issue:45520", + "issue:33357|issue:33666", + "issue:43504|issue:43606", + "issue:44423|issue:44734", + "issue:43572|issue:44291", + "issue:43927|issue:44361", + "issue:43575|issue:45405", + "issue:43504|issue:43653" + ], + "prepared_review_unit_hash": "9edd495e375239f7a96a0c944c4e0e9ac29507c9e42d24b4ddbcb1b8e785f857", + "component_budget": { + "node_count": 387, + "item_count": 387, + "soft_pair_count": 1455, + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 33, + "serialized_chars": 37881, + "estimated_input_tokens": 9471, + "estimated_eval_tokens": 19198 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This is a heterogeneous cluster of mostly unrelated Transformers bug reports. The soft pairs share broad vocabulary like loading, configs, dtype, or Qwen, but they describe different concrete failures and code paths, so they should not be merged as duplicates.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43756", + "right": "issue:44913", + "accept": false, + "reason": "Different bugs: RoPE layer selection in Smollm3 vs GPTNeoX rotary_pct not persisting on reload." + }, + { + "left": "issue:43720", + "right": "issue:43927", + "accept": false, + "reason": "Different failures: packed BitNet weights during accelerate loading vs DiaConfig losing custom token IDs on save/load." + }, + { + "left": "issue:43540", + "right": "issue:43716", + "accept": false, + "reason": "Different multimodal issues: Qwen3OmniMoe video processing error vs Mistral-3 image preprocessor/model dtype mismatch." + }, + { + "left": "issue:43606", + "right": "issue:43927", + "accept": false, + "reason": "Different code paths: CPU offload device mismatch in bark-small vs config token ID persistence bug in DiaConfig." + }, + { + "left": "issue:28282", + "right": "issue:32090", + "accept": false, + "reason": "Both are runtime errors, but one is missing PyTorch import dependency and the other is Trainer GPU broadcast handling of NoneType." + }, + { + "left": "issue:28282", + "right": "issue:33357", + "accept": false, + "reason": "Unrelated problems: missing PyTorch dependency vs MacOS bus error loading a CLIP model." + }, + { + "left": "issue:43334", + "right": "issue:43572", + "accept": false, + "reason": "Both mention missing config fields, but Qwen3-VL pad_token_id and StableLm pad_token_idx are different model/config regressions." + }, + { + "left": "issue:43577", + "right": "issue:44554", + "accept": false, + "reason": "Different dtype issues: BLIP2 components staying float32 vs an MPS attention correctness bug with mismatched head dimensions." + }, + { + "left": "issue:33666", + "right": "issue:36331", + "accept": false, + "reason": "Qwen2-VL multi-GPU training support question/problem is not the same as CustomTrainer compute_loss signature breakage." + }, + { + "left": "issue:43688", + "right": "issue:45704", + "accept": false, + "reason": "Both involve training behavior, but OLMoE/GPT OSS auxiliary loss normalization is unrelated to T5 using apex FusedRMSNorm memory leak." + }, + { + "left": "issue:39692", + "right": "issue:43493", + "accept": false, + "reason": "SigLIP2 docs/example errors are not the same as a deeper HF vs JAX implementation discrepancy." + }, + { + "left": "issue:43720", + "right": "issue:44361", + "accept": false, + "reason": "Different loading issues: BitNet packed weight unpacking vs MLukeTokenizer task-time AttributeError." + }, + { + "left": "issue:44964", + "right": "issue:45405", + "accept": false, + "reason": "Cannot load Phi-4 multimodal model is unrelated to an external minimum PEFT version bump that is unreleased on PyPI." + }, + { + "left": "issue:44479", + "right": "issue:44625", + "accept": false, + "reason": "Qwen3.5 video input regression is unrelated to num_labels propagation between configs." + }, + { + "left": "issue:43540", + "right": "issue:44977", + "accept": false, + "reason": "Both are Qwen3-family issues, but one is video input handling and the other is flash-attention generation correctness." + }, + { + "left": "issue:43493", + "right": "issue:43844", + "accept": false, + "reason": "SigLIP2 implementation discrepancy is not the same as ZeRO-3 gradient inflation with HfDeepSpeedConfig." + }, + { + "left": "issue:44488", + "right": "issue:44843", + "accept": false, + "reason": "Loading a specific model failing is unrelated to AutoTokenizer making an unconditional hub model_info call in offline mode." + }, + { + "left": "issue:43756", + "right": "issue:45083", + "accept": false, + "reason": "Smollm3 RoPE layer count issue does not match a helper-length calculation bug in qwen3_omni_moe." + }, + { + "left": "issue:45200", + "right": "issue:45229", + "accept": false, + "reason": "Gemma 4 text-only mm_token_type_ids defaulting is a configuration bug, while 31B multi-GPU inference OOM is a capacity/runtime issue." + }, + { + "left": "issue:43278", + "right": "issue:43844", + "accept": false, + "reason": "Embedding dtype changing between train and eval is unrelated to DeepSpeed ZeRO-3 gradient scaling behavior." + }, + { + "left": "issue:43257", + "right": "issue:43299", + "accept": false, + "reason": "Qwen3 MOE weight conversion with accelerate+deepspeed is not the same as Qwen3VL loading breakage from a dev-version config/API change." + }, + { + "left": "issue:43646", + "right": "issue:44466", + "accept": false, + "reason": "Transformers 5 custom model initialization breakage is broader and different from inconsistent lm_head.weight serialization depending on device." + }, + { + "left": "issue:43824", + "right": "issue:44589", + "accept": false, + "reason": "Missing Qwen2_5_VL class import is not the same underlying issue as missing Float8 storage support." + }, + { + "left": "issue:39692", + "right": "issue:43295", + "accept": false, + "reason": "Doc/example errors in SigLIP2 are unrelated to a v4.57.5 regression in processor.tokenizer and image tokenization behavior." + }, + { + "left": "issue:43688", + "right": "issue:45701", + "accept": false, + "reason": "Aux loss normalization and version-dependent tokenization changes are different bugs with different failure modes." + }, + { + "left": "issue:44589", + "right": "issue:45520", + "accept": false, + "reason": "Both are import/storage related, but Float8 storage lookup and flash_attn KeyError on Python 3.13 are distinct issues." + }, + { + "left": "issue:33357", + "right": "issue:33666", + "accept": false, + "reason": "MacOS bus error loading a CLIP model is unrelated to Qwen2-VL multi-GPU training behavior." + }, + { + "left": "issue:43504", + "right": "issue:43606", + "accept": false, + "reason": "Legacy field preventing Beit semantic segmentation loading is unrelated to bark-small CPU offload device mismatch." + }, + { + "left": "issue:44423", + "right": "issue:44734", + "accept": false, + "reason": "Continuous-batching crash on multimodal models is a different code path from /v1/responses KV-cache continuation indexing." + }, + { + "left": "issue:43572", + "right": "issue:44291", + "accept": false, + "reason": "Missing pad_token_idx in StableLmConfig is not the same as init_empty_weights passing an unexpected _is_hf_initialized argument." + }, + { + "left": "issue:43927", + "right": "issue:44361", + "accept": false, + "reason": "DiaConfig token ID persistence and MLukeTokenizer task AttributeError are unrelated." + }, + { + "left": "issue:43575", + "right": "issue:45405", + "accept": false, + "reason": "Tensor-parallel OOM while loading Qwen2-57B is unrelated to a PEFT minimum-version packaging issue." + }, + { + "left": "issue:43504", + "right": "issue:43653", + "accept": false, + "reason": "Beit legacy field loading bug is not the same as BigBirdTokenizer mask token registration/empty decode output." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 20, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-387", + "nodes": [ + "issue:15354", + "issue:28218", + "issue:29942", + "issue:30333", + "issue:33357", + "issue:34567", + "issue:38175", + "issue:39401", + "issue:43232", + "issue:43278", + "issue:43295", + "issue:43334", + "issue:43525", + "issue:43540", + "issue:43606", + "issue:43688", + "issue:43701", + "issue:43844", + "issue:43873", + "issue:43994", + "issue:44112", + "issue:44315", + "issue:44464", + "issue:44466", + "issue:44493", + "issue:44521", + "issue:44554", + "issue:44568", + "issue:44610", + "issue:44625", + "issue:44792", + "issue:44849", + "issue:44964", + "issue:45072", + "issue:45084", + "issue:45127", + "issue:45198", + "issue:45200", + "issue:45290", + "issue:45325", + "issue:45356", + "issue:45405", + "issue:45446", + "issue:45520", + "issue:45584", + "issue:45701", + "issue:45759" + ], + "soft_pairs": [ + "issue:44625|issue:45759", + "issue:44792|issue:45405", + "issue:43606|issue:44112", + "issue:43994|issue:44493", + "issue:44521|issue:45290", + "issue:44464|issue:45584", + "issue:29942|issue:30333", + "issue:28218|issue:34567", + "issue:43844|issue:44964", + "issue:45084|issue:45520", + "issue:43688|issue:43873", + "issue:15354|issue:38175", + "issue:44554|issue:45446", + "issue:43540|issue:45200", + "issue:44568|issue:45701", + "issue:43232|issue:44849", + "issue:45072|issue:45198", + "issue:39401|issue:44568", + "issue:43278|issue:43701", + "issue:44315|issue:44849", + "issue:28218|issue:33357", + "issue:45200|issue:45325", + "issue:43295|issue:44610", + "issue:44554|issue:45356", + "issue:43334|issue:43525", + "issue:44466|issue:45127" + ], + "prepared_review_unit_hash": "e5145c0e163026d7898acccc7789cb8c4838488fc10669c1d1fd9c44a109cd1f", + "component_budget": { + "node_count": 387, + "item_count": 387, + "soft_pair_count": 1455, + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 26, + "serialized_chars": 34990, + "estimated_input_tokens": 8748, + "estimated_eval_tokens": 17752 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Most items are unrelated bug reports spanning tokenizers, multimodal processors, generation, and training infra. The only plausible duplicate pair is the missing `pad_token_id` config-load failure in Qwen3-VL and Llama4; the rest are distinct issues.", + "confidence": 0.61, + "canonical_issue_reason": "issue:43525 is the best root-style representative because it captures the shared composite-config loading failure around missing `pad_token_id`, which is the closest true duplicate theme in this set.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43525 is the most reusable global representative of the cluster: it is a concrete, actionable loading bug and generalizes better than the model-specific report.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44625", + "right": "issue:45759", + "accept": false, + "reason": "Both involve Qwen 3.5 config behavior, but one is `num_labels` propagation and the other is `from_config` unwrapping `text_config`; different code paths and symptoms." + }, + { + "left": "issue:44792", + "right": "issue:45405", + "accept": false, + "reason": "A failed janus image-generation test and an unmet PEFT version requirement are unrelated." + }, + { + "left": "issue:43606", + "right": "issue:44112", + "accept": false, + "reason": "Both mention device/CI problems, but one is a real CPU-offload mismatch and the other is a stale test; not the same bug." + }, + { + "left": "issue:43994", + "right": "issue:44493", + "accept": false, + "reason": "Different failures: one is bad model outputs for SigLIP2, the other is unexpected position-id keys across many models." + }, + { + "left": "issue:44521", + "right": "issue:45290", + "accept": false, + "reason": "Both touch chat templating, but one is all-zero assistant masks and the other is a crash on tool-call assistant messages." + }, + { + "left": "issue:44464", + "right": "issue:45584", + "accept": false, + "reason": "Chunked generation with compiled forward and Whisper empty-transcription handling are distinct generation bugs." + }, + { + "left": "issue:29942", + "right": "issue:30333", + "accept": false, + "reason": "A FlashAttention CI failure and an MLflow reporting issue are unrelated." + }, + { + "left": "issue:28218", + "right": "issue:34567", + "accept": false, + "reason": "Tokenizer spacing after added tokens and TrainerState token counters do not share the same underlying defect." + }, + { + "left": "issue:43844", + "right": "issue:44964", + "accept": false, + "reason": "ZeRO-3 gradient growth and loading Phi-4 multimodal are different training vs model-loading issues." + }, + { + "left": "issue:45084", + "right": "issue:45520", + "accept": false, + "reason": "Template-node compilation and a Python 3.13 import KeyError are separate failures." + }, + { + "left": "issue:43688", + "right": "issue:43873", + "accept": false, + "reason": "Aux-loss normalization and quantization offload behavior are unrelated." + }, + { + "left": "issue:15354", + "right": "issue:38175", + "accept": false, + "reason": "TorchScript generator-expression export failure and SigLIP2 zero probabilities are different problems." + }, + { + "left": "issue:44554", + "right": "issue:45446", + "accept": false, + "reason": "MPS attention correctness and a flex_attention PyTorch version check are unrelated code paths." + }, + { + "left": "issue:43540", + "right": "issue:45200", + "accept": false, + "reason": "Qwen3OmniMoe video-processing failure and Gemma 4 token-type defaults are different multimodal bugs." + }, + { + "left": "issue:44568", + "right": "issue:45701", + "accept": false, + "reason": "One is a specific `add_special_tokens` regression for mDeBERTa; the other is a broad tokenizer-version change report." + }, + { + "left": "issue:43232", + "right": "issue:44849", + "accept": false, + "reason": "Generation kwargs handling after `sync_gpus` and a Qwen3.5 hidden-states bug are not the same issue." + }, + { + "left": "issue:45072", + "right": "issue:45198", + "accept": false, + "reason": "bfloat16 dtype mismatches and Wav2Vec2 save/tokenization failure are unrelated." + }, + { + "left": "issue:39401", + "right": "issue:44568", + "accept": false, + "reason": "Both are tokenizer bugs, but offset mapping and special-token insertion are distinct behaviors." + }, + { + "left": "issue:43278", + "right": "issue:43701", + "accept": false, + "reason": "Embedding dtype drift between train/eval and checkpoint-key mismatch are unrelated." + }, + { + "left": "issue:44315", + "right": "issue:44849", + "accept": false, + "reason": "Liger Kernel application with `model_init` and Qwen3.5 hidden-states output are different bugs." + }, + { + "left": "issue:28218", + "right": "issue:33357", + "accept": false, + "reason": "Tokenizer spacing and a MacOS bus error in CLIP are unrelated." + }, + { + "left": "issue:45200", + "right": "issue:45325", + "accept": false, + "reason": "Both are multimodal config issues, but one is missing `mm_token_type_ids` defaults and the other is incorrect temporal position scaling." + }, + { + "left": "issue:43295", + "right": "issue:44610", + "accept": false, + "reason": "Both are processor/model regressions, but one is about `processor.tokenizer`/image passing and the other is an image-size mismatch." + }, + { + "left": "issue:44554", + "right": "issue:45356", + "accept": false, + "reason": "Attention correctness on MPS and Kimi tokenizer codec handling are unrelated." + }, + { + "left": "issue:43334", + "right": "issue:43525", + "accept": true, + "reason": "Both report loading composite multimodal configs failing because `pad_token_id` is missing on the text/config object; same underlying config-handling bug." + }, + { + "left": "issue:44466", + "right": "issue:45127", + "accept": false, + "reason": "Serialization of tied weights and LoRA merge collapse are different failures." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The cluster summary is grounded in the packet, and the only accepted soft edge\u2014issue:43334 with issue:43525\u2014matches a clearly shared pad_token_id composite-config loading bug and looks mergeable into one fix. The other pairwise rejections stay conservative and are supported by the titles." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 21, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-387", + "nodes": [ + "issue:37428", + "issue:38175", + "issue:42915", + "issue:43262", + "issue:43278", + "issue:43295", + "issue:43425", + "issue:43450", + "issue:43493", + "issue:43572", + "issue:43575", + "issue:43653", + "issue:43701", + "issue:43819", + "issue:43824", + "issue:43825", + "issue:43827", + "issue:43844", + "issue:43856", + "issue:43873", + "issue:43901", + "issue:43957", + "issue:43975", + "issue:43994", + "issue:44112", + "issue:44206", + "issue:44360", + "issue:44442", + "issue:44464", + "issue:44514", + "issue:44521", + "issue:44561", + "issue:44610", + "issue:44792", + "issue:44857", + "issue:44898", + "issue:44936", + "issue:44987", + "issue:45083", + "issue:45127", + "issue:45198", + "issue:45245", + "issue:45290", + "issue:45381", + "issue:45412", + "issue:45563", + "issue:45584", + "issue:45593" + ], + "soft_pairs": [ + "issue:44898|issue:45412", + "issue:43493|issue:43701", + "issue:37428|issue:43957", + "issue:43572|issue:44987", + "issue:43873|issue:44464", + "issue:43572|issue:43957", + "issue:43295|issue:43450", + "issue:43824|issue:43975", + "issue:43844|issue:44936", + "issue:43295|issue:43873", + "issue:44514|issue:44521", + "issue:38175|issue:45412", + "issue:43278|issue:43856", + "issue:43295|issue:43994", + "issue:44442|issue:45198", + "issue:37428|issue:43824", + "issue:38175|issue:43493", + "issue:42915|issue:43824", + "issue:43827|issue:43901", + "issue:43653|issue:44112", + "issue:43450|issue:43493", + "issue:43450|issue:43575", + "issue:44360|issue:44561", + "issue:44792|issue:45563", + "issue:45127|issue:45245", + "issue:43819|issue:45083", + "issue:44206|issue:45584", + "issue:43262|issue:43450", + "issue:43819|issue:44112", + "issue:44610|issue:44857", + "issue:37428|issue:43425", + "issue:44857|issue:45593", + "issue:45290|issue:45381", + "issue:43278|issue:43825" + ], + "prepared_review_unit_hash": "eff1c7af9ccdd26c0e9141dc09922c3267a15603e72c4ab606a10f2754e5c6d9", + "component_budget": { + "node_count": 387, + "item_count": 387, + "soft_pair_count": 1455, + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 34, + "serialized_chars": 37791, + "estimated_input_tokens": 9448, + "estimated_eval_tokens": 19152 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This set is mostly heterogeneous and does not form one duplicate cluster. The only plausible duplicate-looking pair is the two SigLIP2 issues; the rest are separate bugs across unrelated models, processors, and training/runtime paths.", + "confidence": 0.82, + "canonical_issue_reason": "No single canonical issue fits the whole set because the items span many unrelated failure modes. If forced to anchor the most coherent duplicate subcluster, issue 43493 is the best representative because it is the broad SigLIP2 implementation-parity report that plausibly covers the zero-probabilities symptom.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 43493 is the strongest issue candidate: it is broader than the specific zero-probabilities report, points at the underlying model-implementation discrepancy, and is the best consolidation target for that subcluster.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44898", + "right": "issue:45412", + "accept": false, + "reason": "Different models and symptoms: Perceiver interpolation failure vs RT-DETR memory not being released." + }, + { + "left": "issue:43493", + "right": "issue:43701", + "accept": false, + "reason": "SigLIP2 model-parity bug vs checkpoint resume key mismatch; unrelated code paths." + }, + { + "left": "issue:37428", + "right": "issue:43957", + "accept": false, + "reason": "ImportError for a flash-attention helper vs meta-device loading breakage; not the same defect." + }, + { + "left": "issue:43572", + "right": "issue:44987", + "accept": false, + "reason": "Missing config field in StableLm vs loading failure for a different model package." + }, + { + "left": "issue:43873", + "right": "issue:44464", + "accept": false, + "reason": "Quantization/offloading behavior vs compiled-forward chunked generation outputs; separate problems." + }, + { + "left": "issue:43572", + "right": "issue:43957", + "accept": false, + "reason": "Config regression vs meta-device model loading failure; only broad version overlap." + }, + { + "left": "issue:43295", + "right": "issue:43450", + "accept": false, + "reason": "Processor/tokenizer regression vs batched video shape bug; different APIs and failure modes." + }, + { + "left": "issue:43824", + "right": "issue:43975", + "accept": false, + "reason": "Missing exported class ImportError vs incorrect detokenization in v5; not the same issue." + }, + { + "left": "issue:43844", + "right": "issue:44936", + "accept": false, + "reason": "ZeRO-3 gradient explosion vs trainer.evaluate() state issue after training; unrelated training bugs." + }, + { + "left": "issue:43295", + "right": "issue:43873", + "accept": false, + "reason": "Processor regression vs quantization/offloading behavior; no shared root cause." + }, + { + "left": "issue:44514", + "right": "issue:44521", + "accept": false, + "reason": "Both involve apply_chat_template, but one is a batched crash with tool calls and the other is all-zero assistant masks; too different to merge confidently." + }, + { + "left": "issue:38175", + "right": "issue:45412", + "accept": false, + "reason": "SigLIP2 output bug vs RT-DETR memory leak; completely different subsystems." + }, + { + "left": "issue:43278", + "right": "issue:43856", + "accept": false, + "reason": "Dtype drift in evaluation vs Qwen3 MoE memory inefficiency; no common bug." + }, + { + "left": "issue:43295", + "right": "issue:43994", + "accept": false, + "reason": "Custom processor/tokenizer regression vs SigLIP2 inference quality issue; different models and code paths." + }, + { + "left": "issue:44442", + "right": "issue:45198", + "accept": false, + "reason": "Both are tokenizer-related, but for different model families and distinct save/load failures." + }, + { + "left": "issue:37428", + "right": "issue:43824", + "accept": false, + "reason": "Both are ImportErrors, but for different missing symbols in unrelated model modules." + }, + { + "left": "issue:38175", + "right": "issue:43493", + "accept": true, + "reason": "Both are SigLIP2 correctness reports; the broader parity/discrepancy issue plausibly subsumes the zero-probabilities symptom." + }, + { + "left": "issue:42915", + "right": "issue:43824", + "accept": false, + "reason": "Qwen3Moe FP8 failure vs missing Qwen2.5-VL export; different models and problems." + }, + { + "left": "issue:43827", + "right": "issue:43901", + "accept": false, + "reason": "Both are docs-ish around pipeline behavior, but they cover different tasks and different regressions." + }, + { + "left": "issue:43653", + "right": "issue:44112", + "accept": false, + "reason": "Tokenizer special-token registration bug vs GraniteSpeech CI device override test failure." + }, + { + "left": "issue:43450", + "right": "issue:43493", + "accept": false, + "reason": "Batched video processor shape issue vs SigLIP2 implementation discrepancy; unrelated areas." + }, + { + "left": "issue:43450", + "right": "issue:43575", + "accept": false, + "reason": "Video batch-shape bug vs tensor-parallel OOM on loading a large model; no shared root cause." + }, + { + "left": "issue:44360", + "right": "issue:44561", + "accept": false, + "reason": "Discussion about a DSA indexer activation vs removal of is_torch_fx_available breaking trust_remote_code models." + }, + { + "left": "issue:44792", + "right": "issue:45563", + "accept": false, + "reason": "Janus image-generation test failure vs a stale generate() warning; different severity and code paths." + }, + { + "left": "issue:45127", + "right": "issue:45245", + "accept": false, + "reason": "LoRA merge collapse with extended vocab vs category-count runtime limit; unrelated." + }, + { + "left": "issue:43819", + "right": "issue:45083", + "accept": false, + "reason": "DAC latent/STE mismatch vs Qwen3 Omni feature-length helper behavior; different model families." + }, + { + "left": "issue:44206", + "right": "issue:45584", + "accept": false, + "reason": "Feature extractor center-arg regression vs Whisper empty-transcription failure after token alignment." + }, + { + "left": "issue:43262", + "right": "issue:43450", + "accept": false, + "reason": "Audio apply_chat_template sampling-rate bug vs video processor batched-shape bug; not the same issue." + }, + { + "left": "issue:43819", + "right": "issue:44112", + "accept": false, + "reason": "DAC forward/from_latents mismatch vs GraniteSpeech CI device override failure; unrelated." + }, + { + "left": "issue:44610", + "right": "issue:44857", + "accept": false, + "reason": "OmDet processor size mismatch vs LwDetrImageLoss AMP/CUDA crash; separate defects." + }, + { + "left": "issue:37428", + "right": "issue:43425", + "accept": false, + "reason": "Specific flash-attention import failure vs a broad torch compatibility complaint; not the same bug." + }, + { + "left": "issue:44857", + "right": "issue:45593", + "accept": false, + "reason": "AMP/CUDA loss crash vs missing auxiliary losses when denoising is off." + }, + { + "left": "issue:45290", + "right": "issue:45381", + "accept": false, + "reason": "Assistant-message tool-call crash in chat templating vs Qwen2.5-VL video vision_position_ids bug." + }, + { + "left": "issue:43278", + "right": "issue:43825", + "accept": false, + "reason": "Embedding dtype regression vs pipeline docs/error-message mismatch; different layers of the stack." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet and stays conservative: it correctly treats the set as mostly heterogeneous, and the single accepted soft edge (38175 \u2194 43493) is reasonably supported by the two SigLIP2 titles. The rejected pairs are also framed cautiously based on distinct model families, APIs, or failure modes." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 22, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-387", + "nodes": [ + "issue:28218", + "issue:29127", + "issue:29942", + "issue:32090", + "issue:35141", + "issue:37428", + "issue:38175", + "issue:39692", + "issue:40444", + "issue:43335", + "issue:43388", + "issue:43404", + "issue:43504", + "issue:43526", + "issue:43550", + "issue:43606", + "issue:43611", + "issue:43653", + "issue:43720", + "issue:43819", + "issue:43931", + "issue:44112", + "issue:44292", + "issue:44351", + "issue:44448", + "issue:44462", + "issue:44493", + "issue:44561", + "issue:44568", + "issue:44743", + "issue:44779", + "issue:44855", + "issue:44857", + "issue:44977", + "issue:45072", + "issue:45083", + "issue:45198", + "issue:45200", + "issue:45229", + "issue:45310", + "issue:45459", + "issue:45478", + "issue:45491", + "issue:45520", + "issue:45588", + "issue:45593", + "issue:45712", + "issue:45803" + ], + "soft_pairs": [ + "issue:45200|issue:45588", + "issue:43606|issue:45083", + "issue:44448|issue:44743", + "issue:43335|issue:43550", + "issue:37428|issue:44351", + "issue:44561|issue:44779", + "issue:44462|issue:44561", + "issue:43611|issue:43931", + "issue:39692|issue:43526", + "issue:43388|issue:43526", + "issue:43819|issue:44493", + "issue:43335|issue:43653", + "issue:29942|issue:32090", + "issue:44855|issue:45083", + "issue:43404|issue:45200", + "issue:45310|issue:45478", + "issue:39692|issue:44493", + "issue:43550|issue:44112", + "issue:43335|issue:45083", + "issue:43335|issue:43504", + "issue:44292|issue:45520", + "issue:44448|issue:44779", + "issue:38175|issue:44493", + "issue:28218|issue:35141", + "issue:43335|issue:45072", + "issue:44977|issue:45229", + "issue:43335|issue:45593", + "issue:43720|issue:44112", + "issue:44448|issue:44977", + "issue:45459|issue:45712", + "issue:44568|issue:45491", + "issue:45198|issue:45803", + "issue:43606|issue:44857", + "issue:29127|issue:32090", + "issue:40444|issue:43526", + "issue:44857|issue:45072" + ], + "prepared_review_unit_hash": "d115817b0b9cd2460692970ac1afe0012d58e31dfac6ec47e0d70ffcf643f5b7", + "component_budget": { + "node_count": 387, + "item_count": 387, + "soft_pair_count": 1455, + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 36, + "serialized_chars": 38848, + "estimated_input_tokens": 9712, + "estimated_eval_tokens": 19680 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The set is mostly a broad false-positive similarity cluster: most issues share subsystem keywords but describe different bugs. The only clear duplicate in the soft edges is the pair of identical Qwen3.5 MoE from_pretrained reports.", + "confidence": 0.96, + "canonical_issue_reason": "issue:45310 is the earlier, fuller report of the same Qwen3.5 MoE from_pretrained failure; issue:45478 is a near-identical duplicate.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45310 is the strongest representative: it is specific, reproducible, and clearly states the concrete Qwen3.5 MoE loading failure that the duplicate mirrors.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45200", + "right": "issue:45588", + "accept": false, + "reason": "Different failures: Gemma 4 finetuning needs mm_token_type_ids; flash_attention.py crashes on s_aux=None." + }, + { + "left": "issue:43606", + "right": "issue:45083", + "accept": false, + "reason": "Different code paths and symptoms: CPU offload device mismatch vs a feature-extractor helper behavior issue." + }, + { + "left": "issue:44448", + "right": "issue:44743", + "accept": false, + "reason": "Both mention generation behavior, but one is Pegasus output drift and the other is Qwen3.5 recurrent state resetting." + }, + { + "left": "issue:43335", + "right": "issue:43550", + "accept": false, + "reason": "Unrelated bugs: SwitchTransformers config creates an unwanted sparse layer; Bamba-9B-v2 fails under torch.compile with SDPA." + }, + { + "left": "issue:37428", + "right": "issue:44351", + "accept": false, + "reason": "Both are import errors, but for different missing symbols and different root causes." + }, + { + "left": "issue:44561", + "right": "issue:44779", + "accept": false, + "reason": "Different regressions: removal of is_torch_fx_available vs DeepSeek tokenizer output changes in v5." + }, + { + "left": "issue:44462", + "right": "issue:44561", + "accept": false, + "reason": "AutoTokenizer ignoring tokenizer.json is unrelated to the v5 trust_remote_code import breakage." + }, + { + "left": "issue:43611", + "right": "issue:43931", + "accept": false, + "reason": "Different loading problems: base_model_prefix regression vs Qwen3-VL weight-shape mismatch." + }, + { + "left": "issue:39692", + "right": "issue:43526", + "accept": false, + "reason": "Documentation/example mistakes are not the same underlying bug as BeitImageProcessorFast reduce_labels returning the wrong labels." + }, + { + "left": "issue:43388", + "right": "issue:43526", + "accept": false, + "reason": "Different data-handling bugs: gather_for_metrics dropping tuple labels vs reduce_labels truncating label arrays." + }, + { + "left": "issue:43819", + "right": "issue:44493", + "accept": false, + "reason": "DAC from_latents/STE mismatch is unrelated to position-id key warnings across many models." + }, + { + "left": "issue:43335", + "right": "issue:43653", + "accept": false, + "reason": "Different tokenizer/config areas and different concrete failures." + }, + { + "left": "issue:29942", + "right": "issue:32090", + "accept": false, + "reason": "Flash Attention 2 test failures and _gpu_broadcast_one NoneType errors are separate issues." + }, + { + "left": "issue:44855", + "right": "issue:45083", + "accept": false, + "reason": "Python 3.13 import parsing failure in DebertaV2Model is unrelated to the qwen3_omni_moe helper bug." + }, + { + "left": "issue:43404", + "right": "issue:45200", + "accept": false, + "reason": "lm_head tying bug in Mistral3ForConditionalGeneration is distinct from Gemma 4 token-type-id handling." + }, + { + "left": "issue:45310", + "right": "issue:45478", + "accept": true, + "reason": "Near-identical title and failure mode: same Qwen3.5 MoE from_pretrained error, with 45478 looking like a duplicate follow-up." + }, + { + "left": "issue:39692", + "right": "issue:44493", + "accept": false, + "reason": "Doc/example errors are not the same as the broad unexpected-key-position-id regression." + }, + { + "left": "issue:43550", + "right": "issue:44112", + "accept": false, + "reason": "Bamba compile/SDPA failure is unrelated to a stale GraniteSpeech CI test." + }, + { + "left": "issue:43335", + "right": "issue:45083", + "accept": false, + "reason": "SwitchTransformers sparse-layer construction bug does not match the qwen3_omni_moe output-length helper bug." + }, + { + "left": "issue:43335", + "right": "issue:43504", + "accept": false, + "reason": "Sparse-layer creation and BEiT legacy-field loading are different model-specific bugs." + }, + { + "left": "issue:44292", + "right": "issue:45520", + "accept": false, + "reason": "Qwen-3-8B-NVFP4 runtime error and flash_attn import KeyError are different failure classes." + }, + { + "left": "issue:44448", + "right": "issue:44779", + "accept": false, + "reason": "Pegasus output differences and DeepSeek tokenizer regressions do not share the same code path." + }, + { + "left": "issue:38175", + "right": "issue:44493", + "accept": false, + "reason": "Zero probabilities in SigLIP2 are unrelated to unexpected position-id keys." + }, + { + "left": "issue:28218", + "right": "issue:35141", + "accept": false, + "reason": "Tokenizer spacing after added tokens is unrelated to embedding reinitialization after resize." + }, + { + "left": "issue:43335", + "right": "issue:45072", + "accept": false, + "reason": "SwitchTransformers sparse-layer config bug is unrelated to dtype mismatches in bfloat16 inference." + }, + { + "left": "issue:44977", + "right": "issue:45229", + "accept": false, + "reason": "Qwen3.5 flash-attention generation bug is unrelated to Gemma4 multi-GPU OOM." + }, + { + "left": "issue:43335", + "right": "issue:45593", + "accept": false, + "reason": "Sparse-layer creation is unrelated to D-FINE auxiliary-loss behavior when denoising is off." + }, + { + "left": "issue:43720", + "right": "issue:44112", + "accept": false, + "reason": "Packed-weight loading in BitNet is unrelated to the GraniteSpeech device-override CI failure." + }, + { + "left": "issue:44448", + "right": "issue:44977", + "accept": false, + "reason": "Different models and symptoms: Pegasus output drift vs Qwen3.5 flash-attention generation failure." + }, + { + "left": "issue:45459", + "right": "issue:45712", + "accept": false, + "reason": "Tokenizer exception masking is unrelated to leftover dummy classes in dummy_pt_objects.py." + }, + { + "left": "issue:44568", + "right": "issue:45491", + "accept": false, + "reason": "Missing BOS/EOS with add_special_tokens and NaN embeddings from sliding-window attention are distinct bugs." + }, + { + "left": "issue:45198", + "right": "issue:45803", + "accept": false, + "reason": "Wav2Vec2 save/tokenization failure is unrelated to the bare-except FuyuBatchFeature issue." + }, + { + "left": "issue:43606", + "right": "issue:44857", + "accept": false, + "reason": "CPU offload device mismatch and float16 AMP crashes in LwDetrImageLoss are different problems." + }, + { + "left": "issue:29127", + "right": "issue:32090", + "accept": false, + "reason": "A clearer LayoutLMV3 error message request is not the same as the Trainer _gpu_broadcast_one NoneType crash." + }, + { + "left": "issue:40444", + "right": "issue:43526", + "accept": false, + "reason": "Multi-image IterableDataset finetuning failure is unrelated to BeitImageProcessorFast reduce_labels returning only one label." + }, + { + "left": "issue:44857", + "right": "issue:45072", + "accept": false, + "reason": "AMP/CUDA crash in LwDetrImageLoss is unrelated to SwitchTransformers/TimmWrapperModel bfloat16 dtype mismatches." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The clustering is conservative and grounded in the packet. The only accepted soft-edge pair, issue:45310 and issue:45478, is strongly supported by identical titles and highly similar metadata. The remaining false verdicts are reasonably cautious and do not overstate sameness." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 23, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-387", + "nodes": [ + "issue:25251", + "issue:29942", + "issue:32090", + "issue:34689", + "issue:35141", + "issue:36010", + "issue:37428", + "issue:42915", + "issue:42994", + "issue:43065", + "issue:43066", + "issue:43278", + "issue:43381", + "issue:43412", + "issue:43421", + "issue:43550", + "issue:43653", + "issue:43701", + "issue:43873", + "issue:43927", + "issue:43976", + "issue:43994", + "issue:44186", + "issue:44206", + "issue:44265", + "issue:44361", + "issue:44448", + "issue:44451", + "issue:44462", + "issue:44466", + "issue:44479", + "issue:44493", + "issue:44743", + "issue:44857", + "issue:44871", + "issue:44977", + "issue:45083", + "issue:45198", + "issue:45245", + "issue:45372", + "issue:45412", + "issue:45440", + "issue:45459", + "issue:45491", + "issue:45520", + "issue:45563", + "issue:45588", + "issue:45698" + ], + "soft_pairs": [ + "issue:34689|issue:35141", + "issue:43701|issue:45083", + "issue:45372|issue:45588", + "issue:43927|issue:45698", + "issue:44977|issue:45563", + "issue:43550|issue:45083", + "issue:43412|issue:45588", + "issue:43066|issue:44462", + "issue:45245|issue:45563", + "issue:43873|issue:44493", + "issue:29942|issue:35141", + "issue:36010|issue:37428", + "issue:45372|issue:45459", + "issue:43278|issue:43381", + "issue:25251|issue:32090", + "issue:44448|issue:44466", + "issue:44857|issue:45588", + "issue:43065|issue:45412", + "issue:43653|issue:44743", + "issue:44451|issue:45520", + "issue:44186|issue:45198", + "issue:43976|issue:45440", + "issue:43994|issue:45412", + "issue:44361|issue:45198", + "issue:42915|issue:42994", + "issue:44871|issue:44977", + "issue:44265|issue:45198", + "issue:43421|issue:45491", + "issue:44206|issue:44479" + ], + "prepared_review_unit_hash": "245300eca58a1f229b1d94efd95a9bb039d295d99cf8147c4d44154bc8fe0db5", + "component_budget": { + "node_count": 387, + "item_count": 387, + "soft_pair_count": 1455, + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36372, + "estimated_input_tokens": 9093, + "estimated_eval_tokens": 18442 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is noisy: the soft edges connect clearly unrelated issues across tokenization, generation, export, quantization, memory, and model-loading behavior. No soft pair looks like the same underlying bug; issue 45588 is only the most central representative, not a true duplicate target.", + "confidence": 0.81, + "canonical_issue_reason": "Issue 45588 is the most central and activity-rich node in this otherwise heterogeneous cluster, with multiple nearby soft links around the same flash-attention crash theme.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 45588 is the best representative issue here because it is the densest hub in the cluster and describes a concrete runtime failure, even though the cluster does not form a real duplicate set.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:34689", + "right": "issue:35141", + "accept": false, + "reason": "Model-loading regression for Llama 3.2 Vision vs embedding reinit on resize; unrelated failures." + }, + { + "left": "issue:43701", + "right": "issue:45083", + "accept": false, + "reason": "Checkpoint key mismatch vs helper output-length behavior in qwen3_omni_moe; different bug surfaces." + }, + { + "left": "issue:45372", + "right": "issue:45588", + "accept": false, + "reason": "Gemma 4 processor import failure due to mistral_common vs flash_attention s_aux=None crash; different code paths." + }, + { + "left": "issue:43927", + "right": "issue:45698", + "accept": false, + "reason": "Saved custom token IDs lost on load vs wrong custom module loaded after save_pretrained; related area but distinct bugs." + }, + { + "left": "issue:44977", + "right": "issue:45563", + "accept": false, + "reason": "Qwen3.5 flash-attention generation regression vs stale num_return_sequences warning in paged generate(); not the same issue." + }, + { + "left": "issue:43550", + "right": "issue:45083", + "accept": false, + "reason": "Bamba torch.compile + SDPA failure vs qwen3_omni_moe feature-length helper behavior; unrelated subsystems." + }, + { + "left": "issue:43412", + "right": "issue:45588", + "accept": false, + "reason": "Gemma3 executorch export guard/erfinv support issue vs sink-less flash_attention AttributeError; different failures." + }, + { + "left": "issue:43066", + "right": "issue:44462", + "accept": false, + "reason": "Tokenizer decoder type mismatch in v5 vs AutoTokenizer ignoring tokenizer.json; both tokenizer-related but different bugs." + }, + { + "left": "issue:45245", + "right": "issue:45563", + "accept": false, + "reason": "Category-count runtime error vs paged generate warning; not the same underlying problem." + }, + { + "left": "issue:43873", + "right": "issue:44493", + "accept": false, + "reason": "Quantization offloading issue vs position-id key warning across many models; different behavior." + }, + { + "left": "issue:29942", + "right": "issue:35141", + "accept": false, + "reason": "Flash Attention 2 test failures vs embedding reinitialization after resize; different issues." + }, + { + "left": "issue:36010", + "right": "issue:37428", + "accept": false, + "reason": "GenerationMixin import error vs _flash_supports_window_size import error; both imports but unrelated symbols and modules." + }, + { + "left": "issue:45372", + "right": "issue:45459", + "accept": false, + "reason": "Gemma 4 processor import dependency issue vs protobuf exception masking tokenizer errors; distinct import/error handling bugs." + }, + { + "left": "issue:43278", + "right": "issue:43381", + "accept": false, + "reason": "Embedding dtype drift in eval vs gradient checkpointing disallowed in eval mode; separate training/eval regressions." + }, + { + "left": "issue:25251", + "right": "issue:32090", + "accept": false, + "reason": "top_k pipeline output nesting change vs Trainer _gpu_broadcast_one NoneType crash; unrelated." + }, + { + "left": "issue:44448", + "right": "issue:44466", + "accept": false, + "reason": "Pegasus output mismatch across v4/v5 vs lm_head.weight serialization inconsistency; different model behavior." + }, + { + "left": "issue:44857", + "right": "issue:45588", + "accept": false, + "reason": "LwDetrImageLoss AMP/CUDA crash vs flash_attention s_aux=None AttributeError; different components." + }, + { + "left": "issue:43065", + "right": "issue:45412", + "accept": false, + "reason": "Dummy Conv2d in Sam3PixelDecoder vs RT-DETR memory not released; unrelated model internals." + }, + { + "left": "issue:43653", + "right": "issue:44743", + "accept": false, + "reason": "BigBirdTokenizer special-token registration bug vs Qwen3.5 recurrent state reset with cache; different tokenizer/model logic." + }, + { + "left": "issue:44451", + "right": "issue:45520", + "accept": false, + "reason": "ScandiBERT loading failure vs Python 3.13 flash_attn import KeyError; unrelated import/load paths." + }, + { + "left": "issue:44186", + "right": "issue:45198", + "accept": false, + "reason": "LayoutLMv2Tokenizer NER padding/truncation crash vs Wav2Vec2 save_pretrained/tokenization failure; different tokenizers and symptoms." + }, + { + "left": "issue:43976", + "right": "issue:45440", + "accept": false, + "reason": "Python 3.9/3.10 compatibility issue vs DeepseekV3MoE divergence from remote implementation; unrelated." + }, + { + "left": "issue:43994", + "right": "issue:45412", + "accept": false, + "reason": "SigLIP2 nonsensical outputs with AutoModel/pipeline vs RT-DETR memory leak; different model families and bugs." + }, + { + "left": "issue:44361", + "right": "issue:45198", + "accept": false, + "reason": "MLukeTokenizer AttributeError vs Wav2Vec2 save/tokenization failure; both tokenizer-related but not the same defect." + }, + { + "left": "issue:42915", + "right": "issue:42994", + "accept": false, + "reason": "Qwen3Moe FineGrainedFP8Config failure vs quantized model saving failure; quantization-adjacent but separate problems." + }, + { + "left": "issue:44871", + "right": "issue:44977", + "accept": false, + "reason": "Gemma-3 eos_token_id mismatch vs Qwen3.5 flash-attention generation regression; different model/config issues." + }, + { + "left": "issue:44265", + "right": "issue:45198", + "accept": false, + "reason": "torch.export.export failure with torch_compilable_check vs Wav2Vec2 save/tokenization failure; unrelated export vs tokenizer bug." + }, + { + "left": "issue:43421", + "right": "issue:45491", + "accept": false, + "reason": "Runtime special-token post-processor update feature request vs Gemma3 NaN embeddings with sliding-window attention; different concerns." + }, + { + "left": "issue:44206", + "right": "issue:44479", + "accept": false, + "reason": "LasrFeatureExtractor unsupported center arg vs Qwen v5.3.0 video input regression; different input-processing bugs." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 24, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-387", + "nodes": [ + "issue:29127", + "issue:29942", + "issue:30064", + "issue:32090", + "issue:33357", + "issue:34567", + "issue:36010", + "issue:39692", + "issue:41720", + "issue:42915", + "issue:43065", + "issue:43262", + "issue:43278", + "issue:43329", + "issue:43334", + "issue:43381", + "issue:43441", + "issue:43452", + "issue:43526", + "issue:43531", + "issue:43575", + "issue:43653", + "issue:43701", + "issue:43723", + "issue:43756", + "issue:43824", + "issue:43873", + "issue:43931", + "issue:43994", + "issue:44186", + "issue:44206", + "issue:44279", + "issue:44442", + "issue:44554", + "issue:44610", + "issue:44734", + "issue:44857", + "issue:44977", + "issue:45072", + "issue:45103", + "issue:45229", + "issue:45245", + "issue:45372", + "issue:45406", + "issue:45412", + "issue:45440", + "issue:45588", + "issue:45593" + ], + "soft_pairs": [ + "issue:29942|issue:30064", + "issue:43065|issue:43994", + "issue:44734|issue:45406", + "issue:33357|issue:39692", + "issue:29127|issue:34567", + "issue:44610|issue:45412", + "issue:45103|issue:45372", + "issue:43873|issue:45229", + "issue:43756|issue:45440", + "issue:45372|issue:45406", + "issue:43723|issue:44554", + "issue:43278|issue:43575", + "issue:43381|issue:43701", + "issue:43278|issue:43329", + "issue:43653|issue:44442", + "issue:41720|issue:45229", + "issue:45440|issue:45593", + "issue:29942|issue:36010", + "issue:32090|issue:33357", + "issue:43334|issue:43531", + "issue:43526|issue:44186", + "issue:44977|issue:45245", + "issue:43824|issue:44279", + "issue:43526|issue:44857", + "issue:43526|issue:44610", + "issue:42915|issue:43931", + "issue:43262|issue:44206", + "issue:43452|issue:45588", + "issue:43441|issue:45588", + "issue:44442|issue:45072", + "issue:43575|issue:43701" + ], + "prepared_review_unit_hash": "743184a7cab918dd86b88f093dbc0b190c2a59726c01fa8742b30715f83056cd", + "component_budget": { + "node_count": 387, + "item_count": 387, + "soft_pair_count": 1455, + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 36979, + "estimated_input_tokens": 9245, + "estimated_eval_tokens": 18746 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These items are a loose similarity cluster, not a true duplicate set: they span many unrelated bugs across tokenizers, processors, training, serving, quantization, and model-specific inference issues. I don\u2019t see a single canonical issue or PR for the whole cluster.", + "confidence": 0.96, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:29942", + "right": "issue:30064", + "accept": false, + "reason": "Unrelated failures: FlashAttention 2 tests vs void segmentation-map processing." + }, + { + "left": "issue:43065", + "right": "issue:43994", + "accept": false, + "reason": "Different vision-model bugs: dummy Conv2d in Sam3PixelDecoder vs bad SigLIP2 outputs." + }, + { + "left": "issue:44734", + "right": "issue:45406", + "accept": false, + "reason": "Both touch serve/processor code, but one is KV-cache tensor indexing and the other is missing `_tokenizer` on Gemma4Processor." + }, + { + "left": "issue:33357", + "right": "issue:39692", + "accept": false, + "reason": "Different problems: MacOS bus error for community CLIP vs documentation example errors for SigLIP2." + }, + { + "left": "issue:29127", + "right": "issue:34567", + "accept": false, + "reason": "LayoutLMv3 error-message clarity is unrelated to TrainerState token-count updating." + }, + { + "left": "issue:44610", + "right": "issue:45412", + "accept": false, + "reason": "Different object-detection issues: wrong processor image size vs RT-DETR memory not being released." + }, + { + "left": "issue:45103", + "right": "issue:45372", + "accept": false, + "reason": "auto_docstring future-annotations crash is unrelated to Gemma4 loading failure from missing ReasoningEffort." + }, + { + "left": "issue:43873", + "right": "issue:45229", + "accept": false, + "reason": "Both involve memory/quantization broadly, but offloading behavior and Gemma4 multi-GPU OOM are not the same bug." + }, + { + "left": "issue:43756", + "right": "issue:45440", + "accept": false, + "reason": "Different model-implementation divergences in unrelated architectures." + }, + { + "left": "issue:45372", + "right": "issue:45406", + "accept": false, + "reason": "Both mention Gemma4, but one is an import/dependency issue and the other is a serve-time `_tokenizer` attribute crash." + }, + { + "left": "issue:43723", + "right": "issue:44554", + "accept": false, + "reason": "Tokenizer loading in v5 is unrelated to the MPS attention correctness issue." + }, + { + "left": "issue:43278", + "right": "issue:43575", + "accept": false, + "reason": "BF16/FP32 eval dtype change is unrelated to Qwen2 TP loading OOM." + }, + { + "left": "issue:43381", + "right": "issue:43701", + "accept": false, + "reason": "Gradient checkpointing in eval mode is unrelated to checkpoint resume key mismatches." + }, + { + "left": "issue:43278", + "right": "issue:43329", + "accept": false, + "reason": "Different bugs: embedding dtype regression vs multimodal token counting video-branch NameError." + }, + { + "left": "issue:43653", + "right": "issue:44442", + "accept": false, + "reason": "Tokenizer special-token registration bug is unrelated to FastSpeech2Conformer AutoTokenizer loading failure." + }, + { + "left": "issue:41720", + "right": "issue:45229", + "accept": false, + "reason": "Different model/loading failures: Qwen3 auto device mapping assert vs Gemma4 OOM." + }, + { + "left": "issue:45440", + "right": "issue:45593", + "accept": false, + "reason": "DeepSeekV3 implementation divergence is unrelated to D-FINE auxiliary-loss handling." + }, + { + "left": "issue:29942", + "right": "issue:36010", + "accept": false, + "reason": "FlashAttention test failures are unrelated to GenerationMixin import errors." + }, + { + "left": "issue:32090", + "right": "issue:33357", + "accept": false, + "reason": "Trainer GPU broadcast NoneType crash is unrelated to MacOS CLIP bus error." + }, + { + "left": "issue:43334", + "right": "issue:43531", + "accept": false, + "reason": "Different Qwen3-family bugs: missing pad_token_id in VL config vs sliding-window MoE behavior." + }, + { + "left": "issue:43526", + "right": "issue:44186", + "accept": false, + "reason": "BeitImageProcessorFast reduce_labels bug is unrelated to LayoutLMv2Tokenizer NER padding/truncation crashes." + }, + { + "left": "issue:44977", + "right": "issue:45245", + "accept": false, + "reason": "Flash-attention generation failure is unrelated to category-count limits." + }, + { + "left": "issue:43824", + "right": "issue:44279", + "accept": false, + "reason": "Specific missing Qwen2.5-VL export is not the same as a generic dependency issue." + }, + { + "left": "issue:43526", + "right": "issue:44857", + "accept": false, + "reason": "Different failures: Beit label reduction vs LwDetrImageLoss AMP CUDA crash." + }, + { + "left": "issue:43526", + "right": "issue:44610", + "accept": false, + "reason": "Both are image-processor related, but they concern different models and different bugs." + }, + { + "left": "issue:42915", + "right": "issue:43931", + "accept": false, + "reason": "Qwen3MoE FP8 config failure is different from Qwen3-VL weight-shape mismatch." + }, + { + "left": "issue:43262", + "right": "issue:44206", + "accept": false, + "reason": "Audio chat-template sampling-rate default is unrelated to unsupported `center` argument in feature extraction." + }, + { + "left": "issue:43452", + "right": "issue:45588", + "accept": false, + "reason": "Tokenizer/model loading via gguf_file is unrelated to flash_attention sink-less `s_aux=None` crash." + }, + { + "left": "issue:43441", + "right": "issue:45588", + "accept": false, + "reason": "Both mention flash attention, but the concrete failing paths and fixes are different." + }, + { + "left": "issue:44442", + "right": "issue:45072", + "accept": false, + "reason": "Different issues: tokenizer loading vs bfloat16 dtype mismatches in inference." + }, + { + "left": "issue:43575", + "right": "issue:43701", + "accept": false, + "reason": "TP OOM when loading Qwen2 is unrelated to resume-from-checkpoint key mismatch." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 25, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-387", + "nodes": [ + "issue:28218", + "issue:29127", + "issue:32090", + "issue:33290", + "issue:33357", + "issue:34567", + "issue:34689", + "issue:35141", + "issue:38175", + "issue:39692", + "issue:42907", + "issue:43316", + "issue:43381", + "issue:43493", + "issue:43525", + "issue:43550", + "issue:43575", + "issue:43611", + "issue:43646", + "issue:43761", + "issue:43827", + "issue:43873", + "issue:43931", + "issue:43994", + "issue:44077", + "issue:44079", + "issue:44188", + "issue:44206", + "issue:44387", + "issue:44442", + "issue:44464", + "issue:44479", + "issue:44743", + "issue:44811", + "issue:44898", + "issue:44964", + "issue:44977", + "issue:45072", + "issue:45216", + "issue:45229", + "issue:45290", + "issue:45440", + "issue:45491", + "issue:45584", + "issue:45588", + "issue:45600", + "issue:45698", + "issue:45803" + ], + "soft_pairs": [ + "issue:44743|issue:44811", + "issue:44464|issue:45491", + "issue:44442|issue:45803", + "issue:29127|issue:35141", + "issue:43646|issue:44077", + "issue:43575|issue:44387", + "issue:45072|issue:45229", + "issue:44977|issue:45584", + "issue:33357|issue:38175", + "issue:45588|issue:45600", + "issue:44479|issue:45290", + "issue:43316|issue:43525", + "issue:44964|issue:45440", + "issue:43761|issue:44079", + "issue:28218|issue:29127", + "issue:43931|issue:45216", + "issue:28218|issue:32090", + "issue:42907|issue:45698", + "issue:33290|issue:34567", + "issue:34689|issue:43611", + "issue:43761|issue:44188", + "issue:44442|issue:44898", + "issue:39692|issue:45491", + "issue:43381|issue:43873", + "issue:43493|issue:43827", + "issue:43550|issue:44743", + "issue:43994|issue:44206" + ], + "prepared_review_unit_hash": "b85d752dbb3e516cfe35822bfb0fadff3400ec348d67e28ec4ae1e2bef30c3b9", + "component_budget": { + "node_count": 387, + "item_count": 387, + "soft_pair_count": 1455, + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 27, + "serialized_chars": 35848, + "estimated_input_tokens": 8962, + "estimated_eval_tokens": 18180 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The set is mostly heterogeneous; nearly all candidate pairs are separate bugs. The only strong duplicate-like pair is the generic `ModelOutput` None-key assignment issue and the CLIPVision hidden_states regression it can explain.", + "confidence": 0.84, + "canonical_issue_reason": "issue:44079 is the best canonical issue because it describes the broader root-cause bug in `ModelOutput` key assignment; `issue:43761` looks like a model-specific symptom of that same problem.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:44079 is the strongest representative of the only plausible duplicate subcluster and has the most general, reusable fix scope.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44743", + "right": "issue:44811", + "accept": false, + "reason": "Different subsystems and failure modes: cache/recurrent-state reset vs Whisper decoding flags." + }, + { + "left": "issue:44464", + "right": "issue:45491", + "accept": false, + "reason": "Both involve generation quality, but the reported bugs are distinct: compiled chunked generation vs Gemma3 NaN embeddings with mixed-length windows." + }, + { + "left": "issue:44442", + "right": "issue:45803", + "accept": false, + "reason": "Unrelated tokenizer-loading failure vs a tensor conversion exception swallowing interrupts." + }, + { + "left": "issue:29127", + "right": "issue:35141", + "accept": false, + "reason": "LayoutLMv3 box validation/error messaging is unrelated to embedding reinitialization after resize." + }, + { + "left": "issue:43646", + "right": "issue:44077", + "accept": false, + "reason": "Related theme, but not the same concrete bug: generic custom-model init breakage vs a specific `patchtsmixer` `post_init` compatibility issue." + }, + { + "left": "issue:43575", + "right": "issue:44387", + "accept": false, + "reason": "Both are OOM reports, but with different triggers and code paths: tensor parallel loading vs int4 quantization memory growth." + }, + { + "left": "issue:45072", + "right": "issue:45229", + "accept": false, + "reason": "Different problems: dtype mismatch in inference vs multi-GPU CUDA OOM." + }, + { + "left": "issue:44977", + "right": "issue:45584", + "accept": false, + "reason": "Unrelated model-generation regressions in different models and code paths." + }, + { + "left": "issue:33357", + "right": "issue:38175", + "accept": false, + "reason": "Different models and symptoms: MacOS bus error on CLIP vs zero-probability outputs on SigLIP2." + }, + { + "left": "issue:45588", + "right": "issue:45600", + "accept": false, + "reason": "Different code paths: flash-attention crash on `s_aux=None` vs stale auto-mapping references breaking CI." + }, + { + "left": "issue:44479", + "right": "issue:45290", + "accept": false, + "reason": "Video-input regression and chat-template/tool-call crash are separate bugs." + }, + { + "left": "issue:43316", + "right": "issue:43525", + "accept": false, + "reason": "Both are config/API issues, but they concern different missing/uneven attributes in different model configs." + }, + { + "left": "issue:44964", + "right": "issue:45440", + "accept": false, + "reason": "Different model families and failure modes: loading Phi-4 multimodal vs DeepSeekV3 implementation divergence." + }, + { + "left": "issue:43761", + "right": "issue:44079", + "accept": true, + "reason": "`issue:44079` describes the generic `ModelOutput` None-key assignment bug that plausibly causes `issue:43761`'s missing `hidden_states` symptom." + }, + { + "left": "issue:28218", + "right": "issue:32090", + "accept": false, + "reason": "Tokenizer spacing bug and Trainer/NCCL broadcast TypeError are unrelated." + }, + { + "left": "issue:43931", + "right": "issue:45216", + "accept": false, + "reason": "Both mention Qwen and loading/saving, but one is a shape-mismatch load error and the other is an incorrect checkpoint save regression." + }, + { + "left": "issue:42907", + "right": "issue:45698", + "accept": false, + "reason": "Dequantized-model saving failure is not the same as loading the wrong custom module after `save_pretrained`." + }, + { + "left": "issue:33290", + "right": "issue:34567", + "accept": false, + "reason": "OOM with Adafactor/DeepSpeed is unrelated to the trainer state token counter not updating." + }, + { + "left": "issue:34689", + "right": "issue:43611", + "accept": false, + "reason": "Both are loading regressions, but they appear to stem from different version changes and different affected mechanics." + }, + { + "left": "issue:43761", + "right": "issue:44188", + "accept": false, + "reason": "`hidden_states` output handling is unrelated to attention-kernel divergence under `torch.compile`." + }, + { + "left": "issue:44442", + "right": "issue:44898", + "accept": false, + "reason": "Different vision-tokenizer/model failures with no shared concrete bug." + }, + { + "left": "issue:39692", + "right": "issue:45491", + "accept": false, + "reason": "Documentation example errors are unrelated to the Gemma3 mixed-length NaN embedding regression." + }, + { + "left": "issue:43381", + "right": "issue:43873", + "accept": false, + "reason": "Gradient checkpointing in eval mode and quantization offloading are separate features/issues." + }, + { + "left": "issue:43493", + "right": "issue:43827", + "accept": false, + "reason": "SigLIP2 implementation drift is unrelated to docs still referencing removed `pipeline()` APIs." + }, + { + "left": "issue:43550", + "right": "issue:44743", + "accept": false, + "reason": "torch.compile+SDPA failure and cache/recurrent-state reset are distinct bugs." + }, + { + "left": "issue:43994", + "right": "issue:44206", + "accept": false, + "reason": "SigLIP2 wrong outputs and a LASR feature-extractor crash are unrelated." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet and stays mostly conservative. The single accepted pair (`issue:43761` / `issue:44079`) is defensible from the titles alone: a generic `ModelOutput` key-assignment bug plausibly explains a model-specific `hidden_states=None` regression. The other soft-edge rejections are also cautious and do not overclaim shared root causes." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 26, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-387", + "nodes": [ + "issue:25251", + "issue:28282", + "issue:29127", + "issue:32090", + "issue:33290", + "issue:33357", + "issue:34689", + "issue:36010", + "issue:41628", + "issue:41950", + "issue:42175", + "issue:43232", + "issue:43295", + "issue:43316", + "issue:43381", + "issue:43388", + "issue:43404", + "issue:43441", + "issue:43475", + "issue:43540", + "issue:43550", + "issue:43606", + "issue:43643", + "issue:43701", + "issue:43720", + "issue:43742", + "issue:43825", + "issue:43856", + "issue:43873", + "issue:44077", + "issue:44162", + "issue:44190", + "issue:44442", + "issue:44479", + "issue:44496", + "issue:44514", + "issue:44568", + "issue:44792", + "issue:44855", + "issue:44871", + "issue:44877", + "issue:44913", + "issue:45137", + "issue:45372", + "issue:45479", + "issue:45584", + "issue:45656", + "issue:45759" + ], + "soft_pairs": [ + "issue:43295|issue:45584", + "issue:43316|issue:45759", + "issue:44568|issue:45479", + "issue:41628|issue:45372", + "issue:43720|issue:44442", + "issue:43643|issue:44913", + "issue:32090|issue:33290", + "issue:44190|issue:44496", + "issue:43873|issue:44871", + "issue:34689|issue:36010", + "issue:43381|issue:43856", + "issue:44077|issue:44479", + "issue:44855|issue:45137", + "issue:33290|issue:45656", + "issue:29127|issue:33357", + "issue:43381|issue:43742", + "issue:43540|issue:44514", + "issue:43316|issue:44877", + "issue:33290|issue:34689", + "issue:42175|issue:43316", + "issue:25251|issue:28282", + "issue:44162|issue:44792", + "issue:43404|issue:43441", + "issue:43232|issue:43388", + "issue:43295|issue:43825", + "issue:41950|issue:43475", + "issue:44871|issue:45584", + "issue:43550|issue:43701", + "issue:43606|issue:43701" + ], + "prepared_review_unit_hash": "126822630cb276da560052c51516b2b065a9f843e3bba95bd01ce8e006c826eb", + "component_budget": { + "node_count": 387, + "item_count": 387, + "soft_pair_count": 1455, + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36610, + "estimated_input_tokens": 9153, + "estimated_eval_tokens": 18562 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "No convincing duplicates here: the soft pairs span unrelated issues across config loading, generation, multimodal processors, DeepSpeed, tokenizers, and model-specific regressions. I would not merge any of the suggested pairs, and there is no safe canonical representative for the set.", + "confidence": 0.93, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": "No issue in this set is a safe canonical representative; the items are largely unrelated one-off bugs rather than variants of the same underlying problem.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43295", + "right": "issue:45584", + "accept": false, + "reason": "Different bugs: processor.tokenizer/images-to-tokenizer regression vs Whisper empty-transcription handling after align_special_tokens." + }, + { + "left": "issue:43316", + "right": "issue:45759", + "accept": false, + "reason": "Both involve config/model loading, but the concrete bugs differ: Gemma3TextConfig API mismatch vs Qwen composite config unwrap logic." + }, + { + "left": "issue:44568", + "right": "issue:45479", + "accept": false, + "reason": "Unrelated: tokenizer special-token insertion regression vs zero-loss classification bug for num_labels=1." + }, + { + "left": "issue:41628", + "right": "issue:45372", + "accept": false, + "reason": "Both are import errors, but they stem from different missing symbols and different load paths." + }, + { + "left": "issue:43720", + "right": "issue:44442", + "accept": false, + "reason": "Packed-weight loading bug vs AutoTokenizer failing for a specific tokenizer class; no shared code path." + }, + { + "left": "issue:43643", + "right": "issue:44913", + "accept": false, + "reason": "Different config persistence issues: trust_remote_code field loss vs rotary_pct resetting on reload." + }, + { + "left": "issue:32090", + "right": "issue:33290", + "accept": false, + "reason": "Trainer broadcast NoneType error and DeepSpeed+Adafactor OOM are distinct failures." + }, + { + "left": "issue:44190", + "right": "issue:44496", + "accept": false, + "reason": "Local dataset loading in an example script vs unrecognized model/config metadata loading." + }, + { + "left": "issue:43873", + "right": "issue:44871", + "accept": false, + "reason": "Quantization/offloading behavior vs Gemma eos_token_id config mismatch are unrelated." + }, + { + "left": "issue:34689", + "right": "issue:36010", + "accept": false, + "reason": "Model loading regression for Llama 3.2 Vision vs GenerationMixin import failure are different issues." + }, + { + "left": "issue:43381", + "right": "issue:43856", + "accept": false, + "reason": "Gradient checkpointing in eval mode and Qwen3 MoE memory inefficiency are separate training concerns." + }, + { + "left": "issue:44077", + "right": "issue:44479", + "accept": false, + "reason": "PatchTSMixer post_init validation vs Qwen video-input regression are unrelated." + }, + { + "left": "issue:44855", + "right": "issue:45137", + "accept": false, + "reason": "Python 3.13 torch.jit parsing error vs DeepSpeed ZeRO3 deque underflow are different failures." + }, + { + "left": "issue:33290", + "right": "issue:45656", + "accept": false, + "reason": "Both mention DeepSpeed/optimizer, but one is OOM during Adafactor use and the other is double optimizer stepping." + }, + { + "left": "issue:29127", + "right": "issue:33357", + "accept": false, + "reason": "LayoutLMv3 error-message clarity vs MacOS bus error on CLIP are not the same bug." + }, + { + "left": "issue:43381", + "right": "issue:43742", + "accept": false, + "reason": "Gradient checkpointing eval-mode issue and MobileLLM key error are unrelated." + }, + { + "left": "issue:43540", + "right": "issue:44514", + "accept": false, + "reason": "Different Qwen multimodal failures: video processing ValueError vs batched chat-template crash with padding=False." + }, + { + "left": "issue:43316", + "right": "issue:44877", + "accept": false, + "reason": "Gemma3TextConfig API discrepancy is not the same as granite_speech being rejected by strict config loading." + }, + { + "left": "issue:33290", + "right": "issue:34689", + "accept": false, + "reason": "Adafactor/DeepSpeed OOM and Llama 3.2 Vision loading breakage are unrelated." + }, + { + "left": "issue:42175", + "right": "issue:43316", + "accept": false, + "reason": "Package/backend dependency issue vs Gemma3 config API mismatch; no shared underlying defect." + }, + { + "left": "issue:25251", + "right": "issue:28282", + "accept": false, + "reason": "Pipeline nesting behavior and missing PyTorch ImportError are completely different." + }, + { + "left": "issue:44162", + "right": "issue:44792", + "accept": false, + "reason": "ESM2 breakage and janus image-generation test failure are different model-specific issues." + }, + { + "left": "issue:43404", + "right": "issue:43441", + "accept": false, + "reason": "Mistral3 weight tying bug and Ministral FlashAttention failure are distinct code paths." + }, + { + "left": "issue:43232", + "right": "issue:43388", + "accept": false, + "reason": "Generation kwargs update bug vs metric-gathering label truncation are unrelated." + }, + { + "left": "issue:43295", + "right": "issue:43825", + "accept": false, + "reason": "Processor/tokenizer regression vs pipeline error-message wording are different problems." + }, + { + "left": "issue:41950", + "right": "issue:43475", + "accept": false, + "reason": "Video-classification pipeline choosing image processors is unrelated to SAM3 output attribute missing." + }, + { + "left": "issue:44871", + "right": "issue:45584", + "accept": false, + "reason": "Gemma eos_token_id mismatch and Whisper empty transcription failure are different bugs." + }, + { + "left": "issue:43550", + "right": "issue:43701", + "accept": false, + "reason": "torch.compile/SDPA issue in Bamba is unrelated to resume_from_checkpoint key mismatch." + }, + { + "left": "issue:43606", + "right": "issue:43701", + "accept": false, + "reason": "CPU offload device mismatch and checkpoint-resume key mismatch are separate issues." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 27, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-387", + "nodes": [ + "issue:33290", + "issue:35141", + "issue:36331", + "issue:41720", + "issue:42617", + "issue:43065", + "issue:43295", + "issue:43452", + "issue:43575", + "issue:43646", + "issue:43673", + "issue:43742", + "issue:43761", + "issue:43827", + "issue:43906", + "issue:43931", + "issue:43994", + "issue:44060", + "issue:44077", + "issue:44190", + "issue:44242", + "issue:44464", + "issue:44479", + "issue:44496", + "issue:44521", + "issue:44534", + "issue:44610", + "issue:44617", + "issue:44734", + "issue:44737", + "issue:44743", + "issue:44792", + "issue:44855", + "issue:44871", + "issue:44913", + "issue:44936", + "issue:44964", + "issue:44977", + "issue:45127", + "issue:45200", + "issue:45245", + "issue:45290", + "issue:45381", + "issue:45584", + "issue:45588", + "issue:45600", + "issue:45698", + "issue:45803" + ], + "soft_pairs": [ + "issue:45127|issue:45200", + "issue:43295|issue:44871", + "issue:33290|issue:35141", + "issue:43761|issue:44479", + "issue:43065|issue:44617", + "issue:33290|issue:36331", + "issue:43575|issue:43742", + "issue:43452|issue:44496", + "issue:42617|issue:43575", + "issue:43646|issue:44913", + "issue:43673|issue:44743", + "issue:44936|issue:45698", + "issue:43761|issue:44242", + "issue:41720|issue:43575", + "issue:44077|issue:44534", + "issue:44479|issue:44734", + "issue:44464|issue:44871", + "issue:43931|issue:44060", + "issue:44617|issue:45584", + "issue:43906|issue:45290", + "issue:43065|issue:44610", + "issue:43827|issue:43994", + "issue:44792|issue:44977", + "issue:44534|issue:44913", + "issue:41720|issue:45245", + "issue:44190|issue:44964", + "issue:44792|issue:45600", + "issue:44737|issue:45803", + "issue:44855|issue:45588", + "issue:44521|issue:45381" + ], + "prepared_review_unit_hash": "33acde81c7497f34ccecb505c4e460001b31d1cfa31fcdb580d22b85fd578674", + "component_budget": { + "node_count": 387, + "item_count": 387, + "soft_pair_count": 1455, + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 36824, + "estimated_input_tokens": 9206, + "estimated_eval_tokens": 18668 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "No soft-edge pair looks like a true duplicate; the issues span unrelated regressions, model-specific bugs, and API changes. The most generally representative issue is the open, highly-referenced multimodal chat-template bug, but it is not a good duplicate target for the rest.", + "confidence": 0.95, + "canonical_issue_reason": "issue:44521 is the clearest representative entry: it is open, has the most inbound references, and describes a concrete, user-facing bug in a widely used API. That said, the cluster does not contain a single underlying duplicate problem.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:44521 is the best issue to anchor this set because it is active, broadly applicable, and has clear reproduction/context; the others are too diverse to serve as one duplicate target.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45127", + "right": "issue:45200", + "accept": false, + "reason": "Different bugs: LoRA/extended-vocab collapse on tied embeddings vs Gemma 4 multimodal token-type IDs defaulting to zeros." + }, + { + "left": "issue:43295", + "right": "issue:44871", + "accept": false, + "reason": "Both are regression-style config issues, but one is about processor/tokenizer API breakage and the other is an eos_token_id mismatch in Gemma-3; not the same failure path." + }, + { + "left": "issue:33290", + "right": "issue:35141", + "accept": false, + "reason": "OOM in Adafactor/DeepSpeed training is unrelated to token-embedding resizing and post_init reinitialization." + }, + { + "left": "issue:43761", + "right": "issue:44479", + "accept": false, + "reason": "Different model families and symptoms: CLIPVisionModel missing hidden_states vs Qwen multimodal video-input regression." + }, + { + "left": "issue:43065", + "right": "issue:44617", + "accept": false, + "reason": "Sam3PixelDecoder dummy Conv2d is a modeling/implementation issue, while Sam3Video OOM is a memory/performance problem." + }, + { + "left": "issue:33290", + "right": "issue:36331", + "accept": false, + "reason": "Training OOM with Adafactor/DeepSpeed is unrelated to a Trainer API signature mismatch in compute_loss()." + }, + { + "left": "issue:43575", + "right": "issue:43742", + "accept": false, + "reason": "Both concern loading large models, but one is a TP/OOM issue and the other is a key/config loading error for a specific model." + }, + { + "left": "issue:43452", + "right": "issue:44496", + "accept": false, + "reason": "AutoTokenizer/AutoModel gguf_file breakage is a loading-path issue; the other is an unrecognized model/config.json issue for a different checkpoint." + }, + { + "left": "issue:42617", + "right": "issue:43575", + "accept": false, + "reason": "Failure running a 3d_parallel example is not the same as OOM when loading Qwen2-57B with tensor parallelism." + }, + { + "left": "issue:43646", + "right": "issue:44913", + "accept": false, + "reason": "Transformers v5 custom model initialization breakage is broader API behavior, while GPTNeoX rotary_pct reset is a specific config reload bug." + }, + { + "left": "issue:43673", + "right": "issue:44743", + "accept": false, + "reason": "GenerationMixin cache missing during chunked_prefill is a v5 cache API issue; recurrent state reset in modular_qwen3_5.py is a separate cache/state bug." + }, + { + "left": "issue:44936", + "right": "issue:45698", + "accept": false, + "reason": "trainer.evaluate() after train() is a Trainer lifecycle bug; wrong custom module after save_pretrained concerns model class resolution on load." + }, + { + "left": "issue:43761", + "right": "issue:44242", + "accept": false, + "reason": "CLIPVision hidden_states regression and missing load-balancing loss when output_router_logits=False are unrelated model/subsystem bugs." + }, + { + "left": "issue:41720", + "right": "issue:43575", + "accept": false, + "reason": "Qwen3 auto device mapping cudaErrorAssert is a runtime placement/assertion bug, not the same as TP OOM when loading Qwen2-57B." + }, + { + "left": "issue:44077", + "right": "issue:44534", + "accept": false, + "reason": "PatchTST post_init allowance and non-persistent buffers filled with junk are different initialization-policy regressions." + }, + { + "left": "issue:44479", + "right": "issue:44734", + "accept": false, + "reason": "Qwen video-input vision_position_ids bug is unrelated to transformers serve KV-cache continuation tensor indexing." + }, + { + "left": "issue:44464", + "right": "issue:44871", + "accept": false, + "reason": "Chunked generation inconsistency under compiled forward is a generation/runtime bug; Gemma-3 eos_token_id mismatch is a config mismatch." + }, + { + "left": "issue:43931", + "right": "issue:44060", + "accept": false, + "reason": "Qwen3-VL shape mismatch on load and incorrect tied-weights warning in Qwen3-Next are separate model-specific loading/config issues." + }, + { + "left": "issue:44617", + "right": "issue:45584", + "accept": false, + "reason": "Sam3Video CUDA OOM and Whisper empty-transcription failure after align_special_tokens are unrelated bugs." + }, + { + "left": "issue:43906", + "right": "issue:45290", + "accept": false, + "reason": "An isolated reproduction of a prior issue is not the same bug as apply_chat_template crashing on assistant tool-call messages with no content." + }, + { + "left": "issue:43065", + "right": "issue:44610", + "accept": false, + "reason": "Sam3PixelDecoder dummy layer issue is unrelated to OmDet-Turbo processor/model input-size mismatch." + }, + { + "left": "issue:43827", + "right": "issue:43994", + "accept": false, + "reason": "Docs still referencing pipeline() after removals is documentation cleanup; siglip2 producing nonsensical outputs is a model/pipeline behavior bug." + }, + { + "left": "issue:44792", + "right": "issue:44977", + "accept": false, + "reason": "janus image-generation test failure and Qwen3.5 flash-attention generation bug are different models and failure modes." + }, + { + "left": "issue:44534", + "right": "issue:44913", + "accept": false, + "reason": "Non-persistent buffer junk in v5 is an initialization serialization issue, not the same as GPTNeoX rotary_pct reset on reload." + }, + { + "left": "issue:41720", + "right": "issue:45245", + "accept": false, + "reason": "Both mention runtime failures, but one is a Qwen3 device-mapping assert and the other is a categorical-cardinality limit error; not the same bug." + }, + { + "left": "issue:44190", + "right": "issue:44964", + "accept": false, + "reason": "Local dataset loading failure in an example script is unrelated to loading microsoft/Phi-4-multimodal-instruct with latest transformers." + }, + { + "left": "issue:44792", + "right": "issue:45600", + "accept": false, + "reason": "A failed janus image-generation test is not the same as auto_mappings.py still referencing removed Sam3LiteText configs." + }, + { + "left": "issue:44737", + "right": "issue:45803", + "accept": false, + "reason": "XLNet CPU allocation bug from missing device= in torch.arange is unrelated to a bare-except swallowing interrupts/errors in FuyuBatchFeature.convert_to_tensors()." + }, + { + "left": "issue:44855", + "right": "issue:45588", + "accept": false, + "reason": "Python 3.13 IndentationError in DebertaV2Model import is unrelated to flash_attention.py crashing on s_aux=None for sink-less models." + }, + { + "left": "issue:44521", + "right": "issue:45381", + "accept": false, + "reason": "Both concern multimodal chat/template behavior, but one is assistant_masks being all zero and the other is video input vision_position_ids being wrong in Qwen2.5-VL; different code paths." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 28, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-387", + "nodes": [ + "issue:36246", + "issue:38175", + "issue:41628", + "issue:43065", + "issue:43116", + "issue:43295", + "issue:43335", + "issue:43404", + "issue:43479", + "issue:43540", + "issue:43550", + "issue:43644", + "issue:43650", + "issue:43761", + "issue:43976", + "issue:44077", + "issue:44112", + "issue:44186", + "issue:44190", + "issue:44220", + "issue:44246", + "issue:44464", + "issue:44466", + "issue:44496", + "issue:44561", + "issue:44683", + "issue:44734", + "issue:44743", + "issue:44855", + "issue:44877", + "issue:44913", + "issue:44933", + "issue:44938", + "issue:45003", + "issue:45127", + "issue:45216", + "issue:45245", + "issue:45276", + "issue:45357", + "issue:45362", + "issue:45375", + "issue:45446", + "issue:45542", + "issue:45563", + "issue:45584", + "issue:45600", + "issue:45632", + "issue:45803" + ], + "soft_pairs": [ + "issue:45127|issue:45357", + "issue:43335|issue:44855", + "issue:44077|issue:44877", + "issue:43644|issue:44913", + "issue:44938|issue:45542", + "issue:43976|issue:44855", + "issue:43550|issue:44190", + "issue:43540|issue:44734", + "issue:43761|issue:44743", + "issue:44496|issue:45803", + "issue:45276|issue:45584", + "issue:44112|issue:45600", + "issue:43650|issue:44220", + "issue:43116|issue:44190", + "issue:43295|issue:45276", + "issue:43335|issue:44190", + "issue:45245|issue:45584", + "issue:43404|issue:45216", + "issue:43479|issue:45375", + "issue:36246|issue:41628", + "issue:44933|issue:45362", + "issue:44186|issue:44734", + "issue:44464|issue:45563", + "issue:38175|issue:43065", + "issue:44683|issue:45446", + "issue:44466|issue:44913", + "issue:44561|issue:45632", + "issue:44734|issue:45632", + "issue:44246|issue:45003" + ], + "prepared_review_unit_hash": "5c64d0d7b2c7a9e2bfe187a9cdcff32575e313c8fd746b4e9ec4729efaa0d2ac", + "component_budget": { + "node_count": 387, + "item_count": 387, + "soft_pair_count": 1455, + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36481, + "estimated_input_tokens": 9121, + "estimated_eval_tokens": 18498 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is highly heterogeneous: these are mostly unrelated single-bug issues, with no soft-edge pair close enough to merge as duplicates. I found no same-root-cause issue pairs under a strict duplicate standard.", + "confidence": 0.95, + "canonical_issue_reason": "issue:41628 is the cleanest, most concrete, and actionable report in the set: a specific import regression with a clear failing symbol and module path.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:41628 is the best representative issue because it is narrowly scoped, externally visible, and likely to map to a precise code change or export fix.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45127", + "right": "issue:45357", + "accept": false, + "reason": "Different bugs: LoRA merge/tied embeddings vs. incorrect visual encoder keys in save_pretrained for Qwen3.5." + }, + { + "left": "issue:43335", + "right": "issue:44855", + "accept": false, + "reason": "Unrelated failures: SwitchTransformers sparse-layer config creation vs. Python 3.13 import/torch.jit parsing error." + }, + { + "left": "issue:44077", + "right": "issue:44877", + "accept": false, + "reason": "Different config-validation problems affecting different models and fields; not the same bug." + }, + { + "left": "issue:43644", + "right": "issue:44913", + "accept": false, + "reason": "Non-persistent buffer junk initialization is unrelated to rotary_pct not persisting on config reload." + }, + { + "left": "issue:44938", + "right": "issue:45542", + "accept": false, + "reason": "Python 3.14 import/load failure is a different issue from tensorboard-only backend resolution without TensorFlow." + }, + { + "left": "issue:43976", + "right": "issue:44855", + "accept": false, + "reason": "Different Python-version compatibility problems with different failure modes." + }, + { + "left": "issue:43550", + "right": "issue:44190", + "accept": false, + "reason": "Torch.compile/SDPA model failure is unrelated to the image-classification script dataset loading bug." + }, + { + "left": "issue:43540", + "right": "issue:44734", + "accept": false, + "reason": "Different Qwen serving/multimodal code paths: video-input validation vs. KV-cache continuation indexing." + }, + { + "left": "issue:43761", + "right": "issue:44743", + "accept": false, + "reason": "CLIPVision hidden-states regression and Qwen recurrent-state reset are separate model-specific bugs." + }, + { + "left": "issue:44496", + "right": "issue:45803", + "accept": false, + "reason": "Model auto-loading/config resolution failure is unrelated to FuyuBatchFeature swallowing exceptions." + }, + { + "left": "issue:45276", + "right": "issue:45584", + "accept": false, + "reason": "Token-embedding resize semantics are unrelated to Whisper empty-transcription generation failure." + }, + { + "left": "issue:44112", + "right": "issue:45600", + "accept": false, + "reason": "Both touch CI around GraniteSpeech/Sam3LiteText, but the causes are different: stale device override test vs removed auto-mapping config references." + }, + { + "left": "issue:43650", + "right": "issue:44220", + "accept": false, + "reason": "Placeholder issue text and fbank feature extraction error are not the same underlying bug." + }, + { + "left": "issue:43116", + "right": "issue:44190", + "accept": false, + "reason": "Multi-label classification script behavior is unrelated to local dataset loading in a different training example." + }, + { + "left": "issue:43295", + "right": "issue:45276", + "accept": false, + "reason": "Processor/tokenizer regression is unrelated to resize_token_embeddings not updating tied embeddings." + }, + { + "left": "issue:43335", + "right": "issue:44190", + "accept": false, + "reason": "SwitchTransformers config construction and example-script dataset loading are unrelated." + }, + { + "left": "issue:45245", + "right": "issue:45584", + "accept": false, + "reason": "Category-count runtime limit is unrelated to Whisper empty-transcription handling." + }, + { + "left": "issue:43404", + "right": "issue:45216", + "accept": false, + "reason": "Different tied-weight/save bugs in different multimodal model families." + }, + { + "left": "issue:43479", + "right": "issue:45375", + "accept": false, + "reason": "Both are multimodal config issues, but one is default-init behavior and the other is a missing field dropped by strict config parsing." + }, + { + "left": "issue:36246", + "right": "issue:41628", + "accept": false, + "reason": "Both are import errors, but they affect different symbols and paths; not the same export bug." + }, + { + "left": "issue:44933", + "right": "issue:45362", + "accept": false, + "reason": "Missing image_utils import is unrelated to Qwen3.5 chat crash behavior." + }, + { + "left": "issue:44186", + "right": "issue:44734", + "accept": false, + "reason": "Tokenizer NER/padding bug is unrelated to serve-time KV-cache continuation indexing." + }, + { + "left": "issue:44464", + "right": "issue:45563", + "accept": false, + "reason": "Compiled-forward chunked generation inconsistency is unrelated to a stale warning in paged generate." + }, + { + "left": "issue:38175", + "right": "issue:43065", + "accept": false, + "reason": "Siglip2 zero-probabilities and Sam3PixelDecoder dummy Conv2d are different model-specific issues." + }, + { + "left": "issue:44683", + "right": "issue:45446", + "accept": false, + "reason": "Torch>=2.9 flex_attention failure and a PyTorch version-check import bug are related only by subsystem, not the same defect." + }, + { + "left": "issue:44466", + "right": "issue:44913", + "accept": false, + "reason": "Serialization of tied lm_head weights is unrelated to rotary_pct not round-tripping." + }, + { + "left": "issue:44561", + "right": "issue:45632", + "accept": false, + "reason": "Removal of is_torch_fx_available affecting trust_remote_code is unrelated to cache-path collisions for local models." + }, + { + "left": "issue:44734", + "right": "issue:45632", + "accept": false, + "reason": "Serve API KV-cache indexing crash is unrelated to trust_remote_code cache path collisions." + }, + { + "left": "issue:44246", + "right": "issue:45003", + "accept": false, + "reason": "Import latency and unsafe sys.modules access are not the same concrete bug." + }, + { + "left": "issue:44220", + "right": "issue:45003", + "accept": false, + "reason": "Audio fbank feature extraction failure is unrelated to sys.modules access in modeling_utils." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 29, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-387", + "nodes": [ + "issue:36246", + "issue:38617", + "issue:41950", + "issue:42175", + "issue:42907", + "issue:43295", + "issue:43299", + "issue:43381", + "issue:43388", + "issue:43441", + "issue:43526", + "issue:43531", + "issue:43577", + "issue:43643", + "issue:43673", + "issue:43688", + "issue:43827", + "issue:43854", + "issue:43901", + "issue:43976", + "issue:44016", + "issue:44062", + "issue:44075", + "issue:44112", + "issue:44220", + "issue:44242", + "issue:44279", + "issue:44336", + "issue:44351", + "issue:44368", + "issue:44496", + "issue:44964", + "issue:45020", + "issue:45042", + "issue:45200", + "issue:45216", + "issue:45278", + "issue:45335", + "issue:45357", + "issue:45405", + "issue:45406", + "issue:45563", + "issue:45584", + "issue:45593", + "issue:45600", + "issue:45632", + "issue:45759" + ], + "soft_pairs": [ + "issue:36246|issue:38617", + "issue:43643|issue:44496", + "issue:45406|issue:45632", + "issue:45563|issue:45584", + "issue:44964|issue:45600", + "issue:41950|issue:44016", + "issue:44112|issue:44964", + "issue:43388|issue:43688", + "issue:42907|issue:43441", + "issue:43295|issue:45335", + "issue:43854|issue:44964", + "issue:44496|issue:45042", + "issue:43673|issue:44336", + "issue:44016|issue:44220", + "issue:43299|issue:45216", + "issue:43531|issue:45600", + "issue:43526|issue:43901", + "issue:42175|issue:44279", + "issue:45335|issue:45584", + "issue:44075|issue:44368", + "issue:43381|issue:45593", + "issue:43299|issue:45357", + "issue:44496|issue:45759", + "issue:44242|issue:45200", + "issue:43901|issue:44496", + "issue:43577|issue:45216", + "issue:41950|issue:45020", + "issue:44016|issue:44062", + "issue:44351|issue:45278", + "issue:43976|issue:45405", + "issue:43388|issue:43827" + ], + "prepared_review_unit_hash": "535e704e6f0e5035a72733ac00d63afc2fab04f43281938267198e55eaa145aa", + "component_budget": { + "node_count": 387, + "item_count": 387, + "soft_pair_count": 1455, + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 31, + "serialized_chars": 36582, + "estimated_input_tokens": 9146, + "estimated_eval_tokens": 18548 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These are mostly distinct bug reports across model loading, generation, processor behavior, docs, and backend/dependency regressions. The only broad umbrella is the open import-error regression report, but the soft pairs are not close enough to merge as duplicates.", + "confidence": 0.73, + "canonical_issue_reason": "issue:45278 is the broadest representative of the cluster: a post-upgrade import-regression umbrella. The other issues are narrower, model- or subsystem-specific failures rather than the same bug.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45278 is the best single issue to anchor triage because it captures the widest cross-version import breakage theme and is already phrased as a general regression report.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:36246", + "right": "issue:38617", + "accept": false, + "reason": "Both are ImportErrors, but they concern different missing symbols in different modules; not the same underlying regression." + }, + { + "left": "issue:43643", + "right": "issue:44496", + "accept": false, + "reason": "One is missing fields from trust_remote_code config loading, the other is an unrecognized model/config problem; different failure modes." + }, + { + "left": "issue:45406", + "right": "issue:45632", + "accept": false, + "reason": "Gemma4Processor tokenizer attribute loss and trust_remote_code cache-path collisions are unrelated bugs." + }, + { + "left": "issue:45563", + "right": "issue:45584", + "accept": false, + "reason": "A stale warning in paged generate() and Whisper failing on empty transcription are separate generation-path issues." + }, + { + "left": "issue:44964", + "right": "issue:45600", + "accept": false, + "reason": "Phi-4 multimodal loading and auto_mappings referencing removed Sam3LiteText configs are different code paths." + }, + { + "left": "issue:41950", + "right": "issue:44016", + "accept": false, + "reason": "Pipeline image-processor lookup bug versus a notebook syntax error; not the same issue." + }, + { + "left": "issue:44112", + "right": "issue:44964", + "accept": false, + "reason": "CI stale device-override test failure is unrelated to Phi-4 model loading." + }, + { + "left": "issue:43388", + "right": "issue:43688", + "accept": false, + "reason": "Metric-gathering label truncation and auxiliary-loss normalization are distinct training/distributed bugs." + }, + { + "left": "issue:42907", + "right": "issue:43441", + "accept": false, + "reason": "Different model-specific failures: saving dequantized models versus FlashAttention loading for Ministral-3." + }, + { + "left": "issue:43295", + "right": "issue:45335", + "accept": false, + "reason": "Processor/tokenizer regression and t5gemma embedding resize behavior do not share the same root cause." + }, + { + "left": "issue:43854", + "right": "issue:44964", + "accept": false, + "reason": "Both are model-loading reports, but for different models and different breakages." + }, + { + "left": "issue:44496", + "right": "issue:45042", + "accept": false, + "reason": "Unrecognized config/model-type loading error is unrelated to PIL backend requiring torchvision." + }, + { + "left": "issue:43673", + "right": "issue:44336", + "accept": false, + "reason": "Generation cache regression and ANSI output formatting in loading_report are unrelated." + }, + { + "left": "issue:44016", + "right": "issue:44220", + "accept": false, + "reason": "Notebook syntax error and audio feature extraction failure are unrelated bugs." + }, + { + "left": "issue:43299", + "right": "issue:45216", + "accept": false, + "reason": "Qwen3VL MoE loading failure and Qwen3.5 save_pretrained checkpoint corruption are different regression classes." + }, + { + "left": "issue:43531", + "right": "issue:45600", + "accept": false, + "reason": "Sliding-window behavior in Qwen3-MoE and auto_mappings references to removed configs are unrelated." + }, + { + "left": "issue:43526", + "right": "issue:43901", + "accept": false, + "reason": "BeitImageProcessorFast label reduction bug versus TextClassificationPipeline docs mismatch; not the same issue." + }, + { + "left": "issue:42175", + "right": "issue:44279", + "accept": false, + "reason": "A missing optional backend in an extra install and a generic dependency issue are too vague and not clearly the same bug." + }, + { + "left": "issue:45335", + "right": "issue:45584", + "accept": false, + "reason": "Embedding resize behavior in t5gemma and Whisper empty-transcription generation failure are unrelated." + }, + { + "left": "issue:44075", + "right": "issue:44368", + "accept": false, + "reason": "SGD optimizer args being ignored and a tie_word_embeddings warning are different configuration issues." + }, + { + "left": "issue:43381", + "right": "issue:45593", + "accept": false, + "reason": "Gradient checkpointing in eval mode and D-FINE auxiliary-loss handling are separate training logic bugs." + }, + { + "left": "issue:43299", + "right": "issue:45357", + "accept": false, + "reason": "Both involve Qwen3.5/Qwen3VL regressions, but one is load-time and the other is save_pretrained key serialization; not the same bug." + }, + { + "left": "issue:44496", + "right": "issue:45759", + "accept": false, + "reason": "Unrecognized model/config loading and from_config text_config unwrapping are different AutoModel problems." + }, + { + "left": "issue:44242", + "right": "issue:45200", + "accept": false, + "reason": "MoE load-balancing loss omission and Gemma4 multimodal token-type defaults are unrelated." + }, + { + "left": "issue:43901", + "right": "issue:44496", + "accept": false, + "reason": "Documentation mismatch and model-config loading failure do not describe the same underlying problem." + }, + { + "left": "issue:43577", + "right": "issue:45216", + "accept": false, + "reason": "BLIP2 dtype propagation and Qwen3.5 checkpoint saving are different regressions." + }, + { + "left": "issue:41950", + "right": "issue:45020", + "accept": false, + "reason": "Video-classification pipeline bug and broad remote_code breakage are related only at a high level, not the same concrete bug." + }, + { + "left": "issue:44016", + "right": "issue:44062", + "accept": false, + "reason": "Notebook syntax error and AddedToken keyword conflict are unrelated." + }, + { + "left": "issue:44351", + "right": "issue:45278", + "accept": false, + "reason": "HybridCache import failure may be one symptom of the broader import-regression report, but these are not clearly the same underlying bug." + }, + { + "left": "issue:43976", + "right": "issue:45405", + "accept": false, + "reason": "Python version support breakage and a PEFT minimum-version bump are dependency/versioning issues, but not the same specific problem." + }, + { + "left": "issue:43388", + "right": "issue:43827", + "accept": false, + "reason": "Distributed label gathering bug and docs still referencing pipeline() are unrelated." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 30, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-387", + "nodes": [ + "issue:32090", + "issue:35141", + "issue:38175", + "issue:39692", + "issue:41950", + "issue:42994", + "issue:43065", + "issue:43295", + "issue:43299", + "issue:43317", + "issue:43441", + "issue:43450", + "issue:43479", + "issue:43525", + "issue:43650", + "issue:43653", + "issue:43723", + "issue:43824", + "issue:43854", + "issue:43901", + "issue:43976", + "issue:44162", + "issue:44261", + "issue:44263", + "issue:44279", + "issue:44297", + "issue:44462", + "issue:44466", + "issue:44521", + "issue:44617", + "issue:44623", + "issue:44671", + "issue:44683", + "issue:44749", + "issue:44779", + "issue:44898", + "issue:44933", + "issue:44987", + "issue:45081", + "issue:45276", + "issue:45290", + "issue:45325", + "issue:45397", + "issue:45405", + "issue:45542", + "issue:45593", + "issue:45600", + "issue:45712" + ], + "soft_pairs": [ + "issue:43479|issue:43525", + "issue:43824|issue:44933", + "issue:44462|issue:45081", + "issue:44683|issue:45542", + "issue:32090|issue:35141", + "issue:44279|issue:45405", + "issue:43299|issue:43854", + "issue:39692|issue:43901", + "issue:38175|issue:44617", + "issue:43295|issue:44779", + "issue:44297|issue:44462", + "issue:41950|issue:43650", + "issue:44462|issue:44779", + "issue:44521|issue:45325", + "issue:43976|issue:44279", + "issue:43976|issue:44987", + "issue:45600|issue:45712", + "issue:43450|issue:44279", + "issue:44466|issue:45276", + "issue:42994|issue:43317", + "issue:43065|issue:44898", + "issue:44261|issue:44263", + "issue:43650|issue:45290", + "issue:43441|issue:44683", + "issue:44623|issue:45397", + "issue:43723|issue:44162", + "issue:44623|issue:44749", + "issue:44671|issue:44779", + "issue:43653|issue:45593" + ], + "prepared_review_unit_hash": "429213a116068e49d2f66d55f2e8c6f27fd098a6edf497b3db6a54c241d03c92", + "component_budget": { + "node_count": 387, + "item_count": 387, + "soft_pair_count": 1455, + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36345, + "estimated_input_tokens": 9087, + "estimated_eval_tokens": 18430 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is highly heterogeneous: most items are unrelated model-specific bugs, docs issues, or dependency/regression reports. The soft-edge pairs mostly share only broad subsystem vocabulary (tokenizers, multimodal, quantization, attention) and should not be merged as duplicates.", + "confidence": 0.89, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": "Issue 44462 is the broadest representative of the tokenizer/loading regression subset, but even that theme is only a loose neighborhood here; the cluster is too mixed to support a single true canonical duplicate.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43479", + "right": "issue:43525", + "accept": false, + "reason": "Both are config/initialization bugs, but for different model families and different failures (`None` defaults vs missing `pad_token_id`)." + }, + { + "left": "issue:43824", + "right": "issue:44933", + "accept": false, + "reason": "Both are import-related, but one is a missing model export and the other is a non-existent utility import; different APIs and root causes." + }, + { + "left": "issue:44462", + "right": "issue:45081", + "accept": false, + "reason": "Both touch tokenizer loading, but one ignores `tokenizer.json` while the other crashes in a Mistral regex patch path; not the same bug." + }, + { + "left": "issue:44683", + "right": "issue:45542", + "accept": false, + "reason": "Both involve dependencies, but one is a torch/attention backend failure and the other is an optional TensorFlow backend issue." + }, + { + "left": "issue:32090", + "right": "issue:35141", + "accept": false, + "reason": "Trainer GPU broadcast `NoneType` and embedding reinit after resize are unrelated code paths." + }, + { + "left": "issue:44279", + "right": "issue:45405", + "accept": false, + "reason": "A generic dependency problem and an invalid PEFT version bump are different issues." + }, + { + "left": "issue:43299", + "right": "issue:43854", + "accept": false, + "reason": "Different model families and different failure modes; both are loading bugs only at a very high level." + }, + { + "left": "issue:39692", + "right": "issue:43901", + "accept": false, + "reason": "Both are docs-related, but one is a SigLIP2 example/model mismatch and the other is a pipeline parameter docs mismatch." + }, + { + "left": "issue:38175", + "right": "issue:44617", + "accept": false, + "reason": "Zero probabilities in SigLIP2 and Sam3Video OOM are unrelated model/runtime failures." + }, + { + "left": "issue:43295", + "right": "issue:44779", + "accept": false, + "reason": "Both are tokenizer regressions, but one is processor/tokenizer access in multimodal code and the other is DeepSeek tokenization correctness; different paths." + }, + { + "left": "issue:44297", + "right": "issue:44462", + "accept": false, + "reason": "Both concern tokenizer serialization/loading, but one is wrong `tokenizer_class` on save and the other is repository `tokenizer.json` being ignored." + }, + { + "left": "issue:41950", + "right": "issue:43650", + "accept": false, + "reason": "Completely unrelated: pipeline processor selection vs an unclear data-related issue." + }, + { + "left": "issue:44462", + "right": "issue:44779", + "accept": false, + "reason": "Repository tokenizer loading and DeepSeek tokenizer output correctness are different bugs." + }, + { + "left": "issue:44521", + "right": "issue:45325", + "accept": false, + "reason": "Both are multimodal, but assistant mask generation and Qwen2.5-VL temporal position scaling are separate code paths." + }, + { + "left": "issue:43976", + "right": "issue:44279", + "accept": false, + "reason": "Python version compatibility and a dependency issue are not the same underlying problem." + }, + { + "left": "issue:43976", + "right": "issue:44987", + "accept": false, + "reason": "Different regression domains: interpreter-version support vs a specific model load failure." + }, + { + "left": "issue:45600", + "right": "issue:45712", + "accept": false, + "reason": "Both are repository-maintenance style issues, but one is stale auto-mappings and the other is leftover dummy classes." + }, + { + "left": "issue:43450", + "right": "issue:44279", + "accept": false, + "reason": "Video processor batching shape bug and a dependency issue are unrelated." + }, + { + "left": "issue:44466", + "right": "issue:45276", + "accept": false, + "reason": "Both involve weight/embedding handling, but one is serialization of tied weights and the other is resize_token_embeddings not updating Gemma-4 outputs." + }, + { + "left": "issue:42994", + "right": "issue:43317", + "accept": false, + "reason": "Quantized-model saving and dequantized model offload loading are different phases of the pipeline." + }, + { + "left": "issue:43065", + "right": "issue:44898", + "accept": false, + "reason": "A dummy Conv2d implementation and a Perceiver non-default resolution failure are unrelated." + }, + { + "left": "issue:44261", + "right": "issue:44263", + "accept": false, + "reason": "Both mention GLM internals, but missing `rms_norm_eps` and `torch.split()` return handling are different bugs." + }, + { + "left": "issue:43650", + "right": "issue:45290", + "accept": false, + "reason": "No meaningful overlap; one is an undefined data issue, the other is a chat-template/tool-call crash." + }, + { + "left": "issue:43441", + "right": "issue:44683", + "accept": false, + "reason": "Both touch attention backends, but they are different models and different backend failures." + }, + { + "left": "issue:44623", + "right": "issue:45397", + "accept": false, + "reason": "Processor save-file omissions and Gemma-4 Zero3 loading are unrelated." + }, + { + "left": "issue:43723", + "right": "issue:44162", + "accept": false, + "reason": "AutoTokenizer loading in v5 and ESM2 breakage are different model/loading issues." + }, + { + "left": "issue:44623", + "right": "issue:44749", + "accept": false, + "reason": "Missing files on save and a performance regression after upgrade are not the same bug." + }, + { + "left": "issue:44671", + "right": "issue:44779", + "accept": false, + "reason": "Different models and different symptoms: masked LM prediction error vs tokenizer output regression." + }, + { + "left": "issue:43653", + "right": "issue:45593", + "accept": false, + "reason": "Tokenizer special-token registration and D-FINE auxiliary-loss behavior are unrelated." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 31, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-387", + "nodes": [ + "issue:16998", + "issue:22355", + "issue:36246", + "issue:41628", + "issue:41950", + "issue:42491", + "issue:42757", + "issue:43122", + "issue:43262", + "issue:43299", + "issue:43352", + "issue:43450", + "issue:43475", + "issue:43525", + "issue:43650", + "issue:43673", + "issue:43761", + "issue:43784", + "issue:43881", + "issue:43901", + "issue:43976", + "issue:43994", + "issue:44016", + "issue:44188", + "issue:44220", + "issue:44242", + "issue:44246", + "issue:44291", + "issue:44295", + "issue:44297", + "issue:44336", + "issue:44351", + "issue:44361", + "issue:44373", + "issue:44462", + "issue:44485", + "issue:44496", + "issue:44521", + "issue:44589", + "issue:44617", + "issue:44987", + "issue:45250", + "issue:45356", + "issue:45362", + "issue:45397", + "issue:45542", + "issue:45593", + "issue:45759" + ], + "soft_pairs": [ + "issue:43784|issue:45542", + "issue:43901|issue:43994", + "issue:43122|issue:43650", + "issue:44361|issue:45593", + "issue:44987|issue:45542", + "issue:44016|issue:44291", + "issue:43262|issue:44521", + "issue:44297|issue:45397", + "issue:44188|issue:44336", + "issue:43673|issue:43761", + "issue:43475|issue:44016", + "issue:43525|issue:45759", + "issue:43525|issue:43881", + "issue:43450|issue:44617", + "issue:44589|issue:45362", + "issue:43901|issue:44246", + "issue:42491|issue:43299", + "issue:41628|issue:43976", + "issue:44336|issue:44373", + "issue:16998|issue:22355", + "issue:44242|issue:45593", + "issue:43352|issue:45250", + "issue:44295|issue:45397", + "issue:44246|issue:44496", + "issue:41950|issue:44220", + "issue:44462|issue:45356", + "issue:44336|issue:44485", + "issue:36246|issue:42757", + "issue:36246|issue:44351" + ], + "prepared_review_unit_hash": "6ae55f6d6a9dae10ab2afe5b6d212c1b1c6dccc7fa309d327ae4da72561f8d2c", + "component_budget": { + "node_count": 387, + "item_count": 387, + "soft_pair_count": 1455, + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36092, + "estimated_input_tokens": 9023, + "estimated_eval_tokens": 18302 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a set of unrelated bug reports that only share broad surface similarity (imports, loading, tokenizer, multimodal, or model-regression themes). None of the soft pairs look like the same underlying bug or change, so all soft edges should be rejected.", + "confidence": 0.94, + "canonical_issue_reason": "issue:41628 is the broadest and most central-looking core-library import regression in the set, and it is the closest thematic representative of the import-name failures here, even though the cluster is not truly duplicate-heavy.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:41628 is the best representative because it is a clear, user-facing core import failure with a well-scoped symptom; the others are either narrower model-specific regressions, docs issues, or unrelated bugs.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43784", + "right": "issue:45542", + "accept": false, + "reason": "Both involve import/backend problems, but one is a missing `nn` import on sentence-transformers integration and the other is a TensorBoard/TensorFlow backend detection issue; different code paths." + }, + { + "left": "issue:43901", + "right": "issue:43994", + "accept": false, + "reason": "Docs mismatch for `return_all_scores` and incorrect SigLIP2 outputs are unrelated problems in different areas of the library." + }, + { + "left": "issue:43122", + "right": "issue:43650", + "accept": false, + "reason": "Tokenization changes across versions and a nonspecific 'ADD THE DATA' issue do not describe the same bug." + }, + { + "left": "issue:44361", + "right": "issue:45593", + "accept": false, + "reason": "`MLukeTokenizer` task AttributeError and missing auxiliary losses in D-FINE are separate model-specific bugs." + }, + { + "left": "issue:44987", + "right": "issue:45542", + "accept": false, + "reason": "A model-loading failure for `physical-intelligence/fast` is not the same as a TensorBoard-without-TF backend error." + }, + { + "left": "issue:44016", + "right": "issue:44291", + "accept": false, + "reason": "Notebook syntax error and `init_empty_weights` TypeError are unrelated failures with different triggers and fix paths." + }, + { + "left": "issue:43262", + "right": "issue:44521", + "accept": false, + "reason": "These are both multimodal/chat-template related, but one is an audio sampling-rate default bug and the other is zero assistant masks for multimodal inputs." + }, + { + "left": "issue:44297", + "right": "issue:45397", + "accept": false, + "reason": "Tokenizer metadata mismatch on save and Gemma-4 zero3 `from_pretrained` failure are different behaviors and code paths." + }, + { + "left": "issue:44188", + "right": "issue:44336", + "accept": false, + "reason": "Attention-kernel divergence under `torch.compile` and ANSI codes in loading report are completely unrelated." + }, + { + "left": "issue:43673", + "right": "issue:43761", + "accept": false, + "reason": "Missing cache in chunked prefill and CLIPVisionModel hidden_states regression affect different components and outputs." + }, + { + "left": "issue:43475", + "right": "issue:44016", + "accept": false, + "reason": "SAM 3 video missing attribute and notebook syntax error are unrelated." + }, + { + "left": "issue:43525", + "right": "issue:45759", + "accept": false, + "reason": "Llama4 `pad_token_id` attribute error and Qwen multimodal `text_config` unwrapping are different model/config bugs." + }, + { + "left": "issue:43525", + "right": "issue:43881", + "accept": false, + "reason": "Llama4 `pad_token_id` and glm-4v-9b loading failure are separate model-loading issues with no shared code-path evidence." + }, + { + "left": "issue:43450", + "right": "issue:44617", + "accept": false, + "reason": "Batched video processor shape bug and SAM 3 CUDA OOM are not the same underlying issue." + }, + { + "left": "issue:44589", + "right": "issue:45362", + "accept": false, + "reason": "Float8 storage lookup error and Qwen3.5 chat crash are different runtime failures." + }, + { + "left": "issue:43901", + "right": "issue:44246", + "accept": false, + "reason": "A documentation inconsistency and a slow-import performance issue are not duplicates." + }, + { + "left": "issue:42491", + "right": "issue:43299", + "accept": false, + "reason": "Both mention Qwen3 MoE loading/version regressions, but they target different version transitions and symptoms; not enough to be the same concrete bug." + }, + { + "left": "issue:41628", + "right": "issue:43976", + "accept": false, + "reason": "Missing `AutoImageProcessor` import and a Python version compatibility problem are different issues." + }, + { + "left": "issue:44336", + "right": "issue:44373", + "accept": false, + "reason": "ANSI escape handling in loading report and a wrong docstring for `position_ids` are unrelated." + }, + { + "left": "issue:16998", + "right": "issue:22355", + "accept": false, + "reason": "Model max length behavior for DeBERTa-V3 and missing `transformers.onnx` are unrelated." + }, + { + "left": "issue:44242", + "right": "issue:45593", + "accept": false, + "reason": "Load-balancing loss routing bug and D-FINE auxiliary-loss behavior are separate model-training issues." + }, + { + "left": "issue:43352", + "right": "issue:45250", + "accept": false, + "reason": "A specific Nemotron Flash Attention 2 limitation and a generic Flash Attention 2 issue are too different to merge as one bug." + }, + { + "left": "issue:44295", + "right": "issue:45397", + "accept": false, + "reason": "Reading `position_ids` after buffer registration and Gemma-4 zero3 loading failure are unrelated." + }, + { + "left": "issue:44246", + "right": "issue:44496", + "accept": false, + "reason": "Slow import and unrecognized OLMo-Hybrid model config are not the same underlying problem." + }, + { + "left": "issue:41950", + "right": "issue:44220", + "accept": false, + "reason": "Video-classification pipeline selecting image processors and `_torch_extract_fbank_features()` are different multimodal/audio bugs." + }, + { + "left": "issue:44462", + "right": "issue:45356", + "accept": false, + "reason": "AutoTokenizer ignoring repository tokenizer.json and Kimi-K2.5 codec/regex regression are separate tokenizer problems." + }, + { + "left": "issue:44336", + "right": "issue:44485", + "accept": false, + "reason": "Loading-report ANSI codes and GLM-5 RoPE implementation are unrelated." + }, + { + "left": "issue:36246", + "right": "issue:42757", + "accept": false, + "reason": "Both are import errors, but they involve different symbols and different dependency layers (`transformers.models...` vs `huggingface_hub`)." + }, + { + "left": "issue:36246", + "right": "issue:44351", + "accept": false, + "reason": "Different missing imports (`Qwen2_5_VLImageProcessor` vs `HybridCache`) indicate distinct breakages and fixes." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 32, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-387", + "nodes": [ + "issue:36246", + "issue:41950", + "issue:42886", + "issue:43065", + "issue:43232", + "issue:43257", + "issue:43262", + "issue:43408", + "issue:43441", + "issue:43475", + "issue:43526", + "issue:43540", + "issue:43596", + "issue:43673", + "issue:43824", + "issue:43901", + "issue:43976", + "issue:44162", + "issue:44242", + "issue:44246", + "issue:44295", + "issue:44336", + "issue:44451", + "issue:44466", + "issue:44496", + "issue:44561", + "issue:44749", + "issue:44779", + "issue:44857", + "issue:44871", + "issue:44877", + "issue:44964", + "issue:44987", + "issue:44995", + "issue:45003", + "issue:45137", + "issue:45200", + "issue:45229", + "issue:45292", + "issue:45335", + "issue:45362", + "issue:45375", + "issue:45459", + "issue:45561", + "issue:45563", + "issue:45584", + "issue:45701", + "issue:45759" + ], + "soft_pairs": [ + "issue:43673|issue:44561", + "issue:43824|issue:43976", + "issue:43976|issue:44964", + "issue:44246|issue:44871", + "issue:44242|issue:44857", + "issue:44162|issue:45200", + "issue:45362|issue:45375", + "issue:44295|issue:44749", + "issue:44779|issue:45292", + "issue:43441|issue:43540", + "issue:45292|issue:45584", + "issue:44877|issue:45375", + "issue:45292|issue:45701", + "issue:44987|issue:45459", + "issue:43901|issue:45003", + "issue:44779|issue:45701", + "issue:36246|issue:43596", + "issue:41950|issue:44451", + "issue:45561|issue:45563", + "issue:43232|issue:44336", + "issue:45375|issue:45759", + "issue:43526|issue:45003", + "issue:43257|issue:45137", + "issue:43065|issue:43408", + "issue:44995|issue:45200", + "issue:44496|issue:45003", + "issue:43262|issue:44246", + "issue:43475|issue:43540", + "issue:42886|issue:45229", + "issue:44466|issue:45335" + ], + "prepared_review_unit_hash": "0a773b8fe9546dcbd45068479c7271f958cf87dd43996ec14f0575662a151d26", + "component_budget": { + "node_count": 387, + "item_count": 387, + "soft_pair_count": 1455, + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 36729, + "estimated_input_tokens": 9183, + "estimated_eval_tokens": 18622 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a loose similarity bucket of unrelated issues. The only plausible duplicate pair is the resize_token_embeddings embedding-propagation bug between the generic and T5Gemma-specific reports.", + "confidence": 0.68, + "canonical_issue_reason": "issue:45292 is the broadest, most general report of the shared resize_token_embeddings / output embedding bug; issue:45335 looks like a model-specific manifestation.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45292 is the best representative issue for tracking the underlying embedding-resize bug across models.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43673", + "right": "issue:44561", + "accept": false, + "reason": "Different failures: generation cache/chunked prefill vs trust_remote_code breakage from a removed helper." + }, + { + "left": "issue:43824", + "right": "issue:43976", + "accept": false, + "reason": "Both are import/load problems, but for different symbols and different root causes." + }, + { + "left": "issue:43976", + "right": "issue:44964", + "accept": false, + "reason": "One is a Python-version compatibility issue; the other is a model-loading/config regression." + }, + { + "left": "issue:44246", + "right": "issue:44871", + "accept": false, + "reason": "Completely unrelated: import-time slowness versus Gemma eos_token_id config mismatch." + }, + { + "left": "issue:44242", + "right": "issue:44857", + "accept": false, + "reason": "Different model paths and symptoms; load-balancing loss is not the same as an AMP crash." + }, + { + "left": "issue:44162", + "right": "issue:45200", + "accept": false, + "reason": "Distinct model families and bugs; ESM2 breakage is unrelated to Gemma 4 token type IDs." + }, + { + "left": "issue:45362", + "right": "issue:45375", + "accept": false, + "reason": "Both mention Qwen, but one is chat/runtime failure and the other is a config field being dropped." + }, + { + "left": "issue:44295", + "right": "issue:44749", + "accept": false, + "reason": "Unrelated issues: position_ids buffer handling vs a performance regression." + }, + { + "left": "issue:44779", + "right": "issue:45292", + "accept": false, + "reason": "Tokenizer correctness in Deepseek is not the same bug as resize_token_embeddings not updating embeddings." + }, + { + "left": "issue:43441", + "right": "issue:43540", + "accept": false, + "reason": "Different code paths and model families: FlashAttention failure vs video-input validation." + }, + { + "left": "issue:45292", + "right": "issue:45584", + "accept": false, + "reason": "Resize-token-embeddings bug is unrelated to Whisper empty-transcription generation." + }, + { + "left": "issue:44877", + "right": "issue:45375", + "accept": false, + "reason": "Different config-loading issues with different models; not the same defect." + }, + { + "left": "issue:45292", + "right": "issue:45701", + "accept": false, + "reason": "Token embedding resizing and tokenization changes are different problems." + }, + { + "left": "issue:44987", + "right": "issue:45459", + "accept": false, + "reason": "Loading a specific model and protobuf error masking are unrelated." + }, + { + "left": "issue:43901", + "right": "issue:45003", + "accept": false, + "reason": "Docs mismatch in pipelines vs unsafe sys.modules access are unrelated." + }, + { + "left": "issue:44779", + "right": "issue:45701", + "accept": false, + "reason": "Both mention tokenizer behavior, but one is a Deepseek regression and the other is a version-wide tokenization change claim." + }, + { + "left": "issue:36246", + "right": "issue:43596", + "accept": false, + "reason": "Different failures: Qwen2.5-VL import error vs DeepSpeed/BertModel index error." + }, + { + "left": "issue:41950", + "right": "issue:44451", + "accept": false, + "reason": "Video-classification pipeline processor lookup and ScandiBERT load failure are not the same bug." + }, + { + "left": "issue:45561", + "right": "issue:45563", + "accept": false, + "reason": "A pytest-xdist race and a stale warning in generate() are unrelated." + }, + { + "left": "issue:43232", + "right": "issue:44336", + "accept": false, + "reason": "Generation kwargs handling and ANSI output formatting are unrelated." + }, + { + "left": "issue:45375", + "right": "issue:45759", + "accept": false, + "reason": "Both involve Qwen multimodal configs, but the concrete bugs differ: missing field vs text_config unwrapping." + }, + { + "left": "issue:43526", + "right": "issue:45003", + "accept": false, + "reason": "Image processor label reduction bug is unrelated to sys.modules access." + }, + { + "left": "issue:43257", + "right": "issue:45137", + "accept": false, + "reason": "Different DeepSpeed-related failures with different symptoms and code paths." + }, + { + "left": "issue:43065", + "right": "issue:43408", + "accept": false, + "reason": "Both are SAM3-related, but one is a dummy layer concern and the other is a model-type warning." + }, + { + "left": "issue:44995", + "right": "issue:45200", + "accept": false, + "reason": "Stale indexer cache in GlmMoeDsa is unrelated to Gemma 4 token type IDs." + }, + { + "left": "issue:44496", + "right": "issue:45003", + "accept": false, + "reason": "Unrecognized model/config loading versus unsafe sys.modules access are distinct bugs." + }, + { + "left": "issue:43262", + "right": "issue:44246", + "accept": false, + "reason": "Audio chat-template sampling-rate bug is unrelated to import-time slowness." + }, + { + "left": "issue:43475", + "right": "issue:43540", + "accept": false, + "reason": "Different models and failures: missing vision-encoder attribute vs video input validation error." + }, + { + "left": "issue:42886", + "right": "issue:45229", + "accept": false, + "reason": "Tokenizer cache/offline loading and Gemma4 multi-GPU OOM are unrelated." + }, + { + "left": "issue:44466", + "right": "issue:45335", + "accept": false, + "reason": "Both mention embeddings, but one is tied-weight serialization and the other is resize_token_embeddings not propagating." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 33, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-387", + "nodes": [ + "issue:28282", + "issue:30990", + "issue:41628", + "issue:42371", + "issue:42673", + "issue:42994", + "issue:43232", + "issue:43452", + "issue:43475", + "issue:43482", + "issue:43525", + "issue:43643", + "issue:43646", + "issue:43653", + "issue:43673", + "issue:43704", + "issue:43827", + "issue:43931", + "issue:44038", + "issue:44060", + "issue:44077", + "issue:44162", + "issue:44297", + "issue:44361", + "issue:44368", + "issue:44623", + "issue:44734", + "issue:44749", + "issue:44869", + "issue:44933", + "issue:44977", + "issue:44995", + "issue:45003", + "issue:45081", + "issue:45092", + "issue:45103", + "issue:45137", + "issue:45230", + "issue:45310", + "issue:45356", + "issue:45381", + "issue:45397", + "issue:45459", + "issue:45561", + "issue:45563", + "issue:45588", + "issue:45600", + "issue:45656" + ], + "soft_pairs": [ + "issue:44995|issue:45588", + "issue:43653|issue:44162", + "issue:45397|issue:45656", + "issue:44368|issue:44977", + "issue:44368|issue:45563", + "issue:44060|issue:44077", + "issue:45137|issue:45600", + "issue:43673|issue:44734", + "issue:45003|issue:45561", + "issue:43931|issue:44368", + "issue:44933|issue:45459", + "issue:44869|issue:45356", + "issue:44297|issue:44623", + "issue:43646|issue:44162", + "issue:44361|issue:44749", + "issue:43452|issue:45003", + "issue:43673|issue:43827", + "issue:44623|issue:45656", + "issue:43232|issue:43673", + "issue:28282|issue:30990", + "issue:42994|issue:43482", + "issue:44749|issue:45656", + "issue:44749|issue:45310", + "issue:43475|issue:43643", + "issue:45230|issue:45397", + "issue:44038|issue:44623", + "issue:42371|issue:45092", + "issue:43475|issue:45381", + "issue:45081|issue:45103", + "issue:41628|issue:44933", + "issue:42994|issue:43525", + "issue:42673|issue:43704" + ], + "prepared_review_unit_hash": "1b407101868bab82fd63f6e404e9e1b1c928d669094039fd6d901cec34c004b2", + "component_budget": { + "node_count": 387, + "item_count": 387, + "soft_pair_count": 1455, + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 32, + "serialized_chars": 37259, + "estimated_input_tokens": 9315, + "estimated_eval_tokens": 18886 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly heterogeneous and does not form a single duplicate set. The only clear duplicate pair is the repeated Qwen3ForCausalLM VRAM leak report; everything else appears to be separate bugs or feature regressions.", + "confidence": 0.84, + "canonical_issue_reason": "issue:42673 is the strongest canonical issue because it is an exact title duplicate of issue:43704, is earlier, and has more discussion around the same Qwen3ForCausalLM VRAM leak in multi-dataloader-thread usage.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:42673 is the best representative issue in this set: it is specific, actionable, and has the clearest duplicate signal among the listed issues.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44995", + "right": "issue:45588", + "accept": false, + "reason": "Different failures: stale indexer cache on second forward pass vs flash-attention crash on s_aux=None." + }, + { + "left": "issue:43653", + "right": "issue:44162", + "accept": false, + "reason": "Both involve model/tokenizer breakage, but the concrete bugs are unrelated: BigBird tokenizer special-token registration vs ESM2 loading/runtime failure." + }, + { + "left": "issue:45397", + "right": "issue:45656", + "accept": false, + "reason": "Different deepspeed-related problems: gemma-4 zero3 from_pretrained vs optimizer.step being called twice." + }, + { + "left": "issue:44368", + "right": "issue:44977", + "accept": false, + "reason": "Both mention Qwen3.5, but one is a tied-weights warning and the other is a generation failure with flash-attention; not the same bug." + }, + { + "left": "issue:44368", + "right": "issue:45563", + "accept": false, + "reason": "A repeated warning about tie_word_embeddings is not the same as a stale num_return_sequences warning in paged generate()." + }, + { + "left": "issue:44060", + "right": "issue:44077", + "accept": false, + "reason": "Both are warning-related, but one is a false tied-weights warning and the other is a patchtsmixer post_init policy change." + }, + { + "left": "issue:45137", + "right": "issue:45600", + "accept": false, + "reason": "DeepSpeed ZeRO3 deque error and auto_mappings CI breakage are unrelated code paths." + }, + { + "left": "issue:43673", + "right": "issue:44734", + "accept": false, + "reason": "Both are generation/cache bugs, but chunked_prefill cache handling and /v1/responses tensor indexing are different concrete failures." + }, + { + "left": "issue:45003", + "right": "issue:45561", + "accept": false, + "reason": "sys.modules access safety and pytest-xdist file capture races are unrelated." + }, + { + "left": "issue:43931", + "right": "issue:44368", + "accept": false, + "reason": "Qwen3-VL weight-shape mismatch and Qwen3.5 LoRA warning are different model-loading vs warning issues." + }, + { + "left": "issue:44933", + "right": "issue:45459", + "accept": false, + "reason": "Nonexistent image_utils import and tokenizer protobuf error handling are unrelated." + }, + { + "left": "issue:44869", + "right": "issue:45356", + "accept": false, + "reason": "Whisper timestamp decode crash and Kimi-K2.5 tokenizer regression are different bugs." + }, + { + "left": "issue:44297", + "right": "issue:44623", + "accept": false, + "reason": "Tokenizer save metadata mismatch and processor.save_pretrained missing files are separate serialization issues." + }, + { + "left": "issue:43646", + "right": "issue:44162", + "accept": false, + "reason": "Custom model initialization breakage and ESM2 breakage are not the same underlying bug." + }, + { + "left": "issue:44361", + "right": "issue:44749", + "accept": false, + "reason": "MLukeTokenizer task AttributeError and a performance regression after upgrading Transformers are unrelated." + }, + { + "left": "issue:43452", + "right": "issue:45003", + "accept": false, + "reason": "gguf_file load-path breakage and unsafe sys.modules access are distinct loading issues." + }, + { + "left": "issue:43673", + "right": "issue:43827", + "accept": false, + "reason": "Missing GenerationMixin cache in chunked_prefill and docs still referencing pipeline() are unrelated." + }, + { + "left": "issue:44623", + "right": "issue:45656", + "accept": false, + "reason": "Processor save_pretrained missing files and duplicate optimizer steps under deepspeed are different behaviors." + }, + { + "left": "issue:43232", + "right": "issue:43673", + "accept": false, + "reason": "sync_gpus generation kwargs update and chunked_prefill cache missing are different generation bugs." + }, + { + "left": "issue:28282", + "right": "issue:30990", + "accept": false, + "reason": "PyTorch ImportError and Sentence Transformers hanging on load are unrelated." + }, + { + "left": "issue:42994", + "right": "issue:43482", + "accept": false, + "reason": "Quantized model saving failure and Qwen2.5-GGUF loading failure are opposite-direction, different bugs." + }, + { + "left": "issue:44749", + "right": "issue:45656", + "accept": false, + "reason": "Filtering slowdown after upgrade and duplicate optimizer-step behavior are unrelated." + }, + { + "left": "issue:44749", + "right": "issue:45310", + "accept": false, + "reason": "Performance regression in data filtering and Qwen3.5-Moe from_pretrained error are different issues." + }, + { + "left": "issue:43475", + "right": "issue:43643", + "accept": false, + "reason": "SAM3 vision encoder missing attribute and trust_remote_code missing config fields are different missing-field problems." + }, + { + "left": "issue:45230", + "right": "issue:45397", + "accept": false, + "reason": "Generic bug report and gemma-4 zero3 from_pretrained are not the same bug." + }, + { + "left": "issue:44038", + "right": "issue:44623", + "accept": false, + "reason": "Qwen3-VL-Moe v5 breakage and processor.save_pretrained missing files are unrelated." + }, + { + "left": "issue:42371", + "right": "issue:45092", + "accept": false, + "reason": "TF32 API settings guidance and InternVL2 meta-init incompatibility are unrelated." + }, + { + "left": "issue:43475", + "right": "issue:45381", + "accept": false, + "reason": "SAM 3 video missing attribute and Qwen2.5-VL wrong vision_position_ids are different multimodal bugs." + }, + { + "left": "issue:45081", + "right": "issue:45103", + "accept": false, + "reason": "Mistral regex patch crash and auto_docstring future-annotations crash are unrelated." + }, + { + "left": "issue:41628", + "right": "issue:44933", + "accept": false, + "reason": "AutoImageProcessor import failure and a different nonexistent image_utils import are not the same concrete issue." + }, + { + "left": "issue:42994", + "right": "issue:43525", + "accept": false, + "reason": "Quantized saving failure and Llama4Config missing pad_token_id are different model/config problems." + }, + { + "left": "issue:42673", + "right": "issue:43704", + "accept": true, + "reason": "Exact same title and same underlying bug: Qwen3ForCausalLM VRAM leak when used in multiple dataloader threads." + }, + { + "left": "issue:43673", + "right": "issue:43827", + "accept": false, + "reason": "Repeated pair in the input set would still be unrelated generation cache bug vs docs cleanup." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "Grounded overall. The summary is conservative and matches the packet: the only clearly duplicate pair is issue:42673 / issue:43704, which have identical titles, earlier/later timestamps, and the same VRAM-leak description. The soft-edge rejections are appropriately cautious and do not overclaim duplication." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 34, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-387", + "nodes": [ + "issue:30333", + "issue:30990", + "issue:33290", + "issue:35141", + "issue:36010", + "issue:41628", + "issue:42175", + "issue:42222", + "issue:42831", + "issue:42898", + "issue:42994", + "issue:43232", + "issue:43278", + "issue:43377", + "issue:43475", + "issue:43723", + "issue:43784", + "issue:43824", + "issue:43874", + "issue:43901", + "issue:44038", + "issue:44060", + "issue:44188", + "issue:44222", + "issue:44242", + "issue:44351", + "issue:44373", + "issue:44496", + "issue:44568", + "issue:44623", + "issue:44661", + "issue:44734", + "issue:44779", + "issue:44792", + "issue:44964", + "issue:45092", + "issue:45103", + "issue:45125", + "issue:45137", + "issue:45200", + "issue:45310", + "issue:45356", + "issue:45375", + "issue:45405", + "issue:45478", + "issue:45561", + "issue:45600", + "issue:45712" + ], + "soft_pairs": [ + "issue:43475|issue:44661", + "issue:45375|issue:45478", + "issue:45310|issue:45375", + "issue:42831|issue:44222", + "issue:42222|issue:45200", + "issue:41628|issue:44351", + "issue:44222|issue:45310", + "issue:42994|issue:43824", + "issue:43377|issue:44242", + "issue:36010|issue:41628", + "issue:42175|issue:45712", + "issue:44038|issue:45310", + "issue:44964|issue:45137", + "issue:44496|issue:45561", + "issue:45561|issue:45600", + "issue:42898|issue:44568", + "issue:30990|issue:35141", + "issue:43901|issue:45561", + "issue:43232|issue:44792", + "issue:43232|issue:44188", + "issue:44188|issue:44373", + "issue:43278|issue:45137", + "issue:43784|issue:43824", + "issue:44623|issue:45310", + "issue:30333|issue:33290", + "issue:44734|issue:45103", + "issue:43723|issue:44568", + "issue:42898|issue:45356", + "issue:44060|issue:45125", + "issue:42898|issue:44779", + "issue:43874|issue:45092", + "issue:45405|issue:45712" + ], + "prepared_review_unit_hash": "8bb6a302683a11ad6074122fc6fd21451325736939ef9299c6e2a24e8086e16c", + "component_budget": { + "node_count": 387, + "item_count": 387, + "soft_pair_count": 1455, + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 32, + "serialized_chars": 37311, + "estimated_input_tokens": 9328, + "estimated_eval_tokens": 18912 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The set is mostly a collection of unrelated bug reports; none of the soft pairs look like true duplicates of the same concrete issue, so all soft edges are rejected. If one issue must represent the cluster, issue:43824 is the strongest candidate due to the highest activity and a clear reproducible import failure.", + "confidence": 0.66, + "canonical_issue_reason": "issue:43824 has the strongest support in this set (high discussion and inbound references) and is a concrete, well-scoped bug report, making it the best representative issue.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43824 is the most central and well-documented issue among otherwise unrelated reports, so it is the best issue choice overall.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43475", + "right": "issue:44661", + "accept": false, + "reason": "Different failures: a SAM 3 VisionEncoderOutput attribute error vs. a tokenizer-mapping add-new-model-like failure." + }, + { + "left": "issue:45375", + "right": "issue:45478", + "accept": false, + "reason": "Same model family, but the bugs are different: missing vision config field vs. a from_pretrained error in Qwen3.5 MoE." + }, + { + "left": "issue:45310", + "right": "issue:45375", + "accept": false, + "reason": "Both involve Qwen3.5 MoE, but one is a from_pretrained failure and the other is a missing config field in the vision config." + }, + { + "left": "issue:42831", + "right": "issue:44222", + "accept": false, + "reason": "Both mention FP8, but one is an accuracy regression and the other is a save_pretrained failure." + }, + { + "left": "issue:42222", + "right": "issue:45200", + "accept": false, + "reason": "Unrelated model families and symptoms: broken vitpose models vs. Gemma 4 token-type defaults for text-only fine-tuning." + }, + { + "left": "issue:41628", + "right": "issue:44351", + "accept": false, + "reason": "Both are import errors, but for different symbols and likely different code paths." + }, + { + "left": "issue:44222", + "right": "issue:45310", + "accept": false, + "reason": "Different concrete bugs: FP8 MoE save_pretrained vs. Qwen3.5 MoE from_pretrained loading." + }, + { + "left": "issue:42994", + "right": "issue:43824", + "accept": false, + "reason": "A quantized model save failure is not the same as an import error for a specific Qwen2.5 VL class." + }, + { + "left": "issue:43377", + "right": "issue:44242", + "accept": false, + "reason": "Different subsystems and failures: MIMI batching/padding-mask behavior vs. MoE load-balancing loss computation." + }, + { + "left": "issue:36010", + "right": "issue:41628", + "accept": false, + "reason": "Different missing imports: GenerationMixin vs. AutoImageProcessor." + }, + { + "left": "issue:42175", + "right": "issue:45712", + "accept": false, + "reason": "Both touch packaging/availability, but one is missing tensorflow in torch extras and the other is leftover dummy classes in the no-torch path." + }, + { + "left": "issue:44038", + "right": "issue:45310", + "accept": false, + "reason": "Related Qwen3 model family, but the reports describe different failures and different model variants." + }, + { + "left": "issue:44964", + "right": "issue:45137", + "accept": false, + "reason": "Completely different bugs: a multimodal model loading failure vs. a DeepSpeed ZeRO3 deque error." + }, + { + "left": "issue:44496", + "right": "issue:45561", + "accept": false, + "reason": "Unrelated: model config/model_type resolution vs. a pytest-xdist file race." + }, + { + "left": "issue:45561", + "right": "issue:45600", + "accept": false, + "reason": "Different issues: test-worker file contention vs. stale auto_mappings references breaking CI." + }, + { + "left": "issue:42898", + "right": "issue:44568", + "accept": false, + "reason": "Both are tokenizer behavior regressions in v5, but they affect different behaviors and tokenizers." + }, + { + "left": "issue:30990", + "right": "issue:35141", + "accept": false, + "reason": "Sentence-transformer loading hang and token-embedding resize reinit are unrelated." + }, + { + "left": "issue:43901", + "right": "issue:45561", + "accept": false, + "reason": "Documentation mismatch for TextClassificationPipeline is unrelated to the xdist captured_info.txt race." + }, + { + "left": "issue:43232", + "right": "issue:44792", + "accept": false, + "reason": "Different code paths: generation kwargs update after sync_gpus vs. a janus image-generation test failure." + }, + { + "left": "issue:43232", + "right": "issue:44188", + "accept": false, + "reason": "Generation bookkeeping bug and attention-kernel divergence under torch.compile are not the same defect." + }, + { + "left": "issue:44188", + "right": "issue:44373", + "accept": false, + "reason": "A kernel branching bug is unrelated to a docstring typo for position_ids." + }, + { + "left": "issue:43278", + "right": "issue:45137", + "accept": false, + "reason": "Embedding dtype drift during evaluation is unrelated to the DeepSpeed ZeRO3 deque error." + }, + { + "left": "issue:43784", + "right": "issue:43824", + "accept": false, + "reason": "Both are import-related, but they concern different missing symbols and different package interactions." + }, + { + "left": "issue:44623", + "right": "issue:45310", + "accept": false, + "reason": "Processor save_pretrained missing files is a serialization bug, not the Qwen3.5 MoE from_pretrained error." + }, + { + "left": "issue:30333", + "right": "issue:33290", + "accept": false, + "reason": "MLflow reporting on failed training jobs is unrelated to an Adafactor/DeepSpeed OOM." + }, + { + "left": "issue:44734", + "right": "issue:45103", + "accept": false, + "reason": "Different failures: serving KV-cache continuation indexing vs. auto_docstring annotation handling." + }, + { + "left": "issue:43723", + "right": "issue:44568", + "accept": false, + "reason": "Both relate to tokenizer behavior in v5, but they are different concrete regressions affecting different APIs." + }, + { + "left": "issue:42898", + "right": "issue:45356", + "accept": false, + "reason": "Different tokenizer regressions: clean_up_tokenization_spaces behavior vs. Kimi-K2.5 codec handling." + }, + { + "left": "issue:44060", + "right": "issue:45125", + "accept": false, + "reason": "Both involve Qwen3.5 MoE, but one is a tied-weights warning and the other is a missing tensor-parallel plan." + }, + { + "left": "issue:42898", + "right": "issue:44779", + "accept": false, + "reason": "Different tokenizer problems with different models and symptoms." + }, + { + "left": "issue:43874", + "right": "issue:45092", + "accept": false, + "reason": "Different multimodal/model-init issues: missing image-patch helper vs. remote-code checkpoint/meta-init incompatibility." + }, + { + "left": "issue:45405", + "right": "issue:45712", + "accept": false, + "reason": "A dependency version bump problem is unrelated to dummy classes leaking into dir(transformers)." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 35, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-387", + "nodes": [ + "issue:28282", + "issue:30333", + "issue:30990", + "issue:34689", + "issue:35141", + "issue:36331", + "issue:38175", + "issue:41628", + "issue:42175", + "issue:42757", + "issue:42831", + "issue:42898", + "issue:43295", + "issue:43479", + "issue:43644", + "issue:43723", + "issue:43873", + "issue:43976", + "issue:44079", + "issue:44188", + "issue:44206", + "issue:44222", + "issue:44261", + "issue:44263", + "issue:44295", + "issue:44297", + "issue:44360", + "issue:44373", + "issue:44448", + "issue:44483", + "issue:44484", + "issue:44485", + "issue:44514", + "issue:44623", + "issue:44869", + "issue:44871", + "issue:44913", + "issue:44933", + "issue:44945", + "issue:44964", + "issue:44977", + "issue:45092", + "issue:45103", + "issue:45310", + "issue:45357", + "issue:45406", + "issue:45656", + "issue:45712" + ], + "soft_pairs": [ + "issue:42898|issue:44448", + "issue:43479|issue:44079", + "issue:44188|issue:44964", + "issue:44295|issue:45656", + "issue:42898|issue:43295", + "issue:44933|issue:45712", + "issue:42175|issue:44933", + "issue:44222|issue:44297", + "issue:43976|issue:44933", + "issue:44297|issue:45656", + "issue:43479|issue:44869", + "issue:44206|issue:45103", + "issue:42831|issue:44623", + "issue:44222|issue:44623", + "issue:30333|issue:36331", + "issue:41628|issue:42757", + "issue:30333|issue:34689", + "issue:43723|issue:44977", + "issue:28282|issue:30333", + "issue:44297|issue:45310", + "issue:30333|issue:35141", + "issue:44222|issue:44484", + "issue:44263|issue:44360", + "issue:44483|issue:45406", + "issue:44263|issue:45092", + "issue:43873|issue:44945", + "issue:44373|issue:44485", + "issue:44514|issue:45103", + "issue:43479|issue:43644", + "issue:44295|issue:44623", + "issue:43723|issue:44871", + "issue:45092|issue:45357", + "issue:44261|issue:45092", + "issue:43479|issue:44913", + "issue:30990|issue:38175" + ], + "prepared_review_unit_hash": "eddab6edd75b86aab8539e17c8d9e87c54865fb8b511ddc4e6821c8ee3ee452d", + "component_budget": { + "node_count": 387, + "item_count": 387, + "soft_pair_count": 1455, + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 35, + "serialized_chars": 38054, + "estimated_input_tokens": 9514, + "estimated_eval_tokens": 19284 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is a loose collection of Transformers issues around loading/saving, config serialization, processor/tokenizer behavior, and a few unrelated training/runtime bugs. I don\u2019t see any soft pair that is clearly the same underlying defect, so no duplicate merges are justified.", + "confidence": 0.27, + "canonical_issue_reason": "If a single representative is needed, issue 44297 is the most central of the serialization-related subset: it is specific, reproducible, and closest to several save/load-related complaints. Still, the cluster is too heterogeneous for a true canonical duplicate.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 44297 is the best standalone anchor because it is concrete and sits at the intersection of the strongest soft-similarity links, but it should be treated as a thematic representative rather than a confirmed duplicate hub.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:42898", + "right": "issue:44448", + "accept": false, + "reason": "Both mention v5 output behavior changes, but one is tokenizer cleanup behavior and the other is Pegasus output drift; not the same code-path bug." + }, + { + "left": "issue:43479", + "right": "issue:44079", + "accept": false, + "reason": "Multimodal config initialization with None is unrelated to ModelOutput key assignment semantics." + }, + { + "left": "issue:44188", + "right": "issue:44964", + "accept": false, + "reason": "Attention-kernel branching under torch.compile is unrelated to a Phi-4 multimodal loading failure." + }, + { + "left": "issue:44295", + "right": "issue:45656", + "accept": false, + "reason": "A position_ids buffer read error is not the same as a double optimizer-step bug under deepspeed." + }, + { + "left": "issue:42898", + "right": "issue:43295", + "accept": false, + "reason": "Both are regressions, but cleanup_spaces and processor.tokenizer/image handling are different failure modes and code paths." + }, + { + "left": "issue:44933", + "right": "issue:45712", + "accept": false, + "reason": "Both touch exports/imports, but one is a missing image_utils import and the other is dummy PT stubs leaking into the public API." + }, + { + "left": "issue:42175", + "right": "issue:44933", + "accept": false, + "reason": "Backend packaging for pip extras is unrelated to an import-export issue in image_utils." + }, + { + "left": "issue:44222", + "right": "issue:44297", + "accept": false, + "reason": "Both involve save_pretrained, but one is FP8 MoE model saving and the other is tokenizer config metadata; different artifacts and bugs." + }, + { + "left": "issue:43976", + "right": "issue:44933", + "accept": false, + "reason": "A Python version compatibility issue is unrelated to a missing import from image_utils." + }, + { + "left": "issue:44297", + "right": "issue:45656", + "accept": false, + "reason": "Tokenizer config mismatch on save is unrelated to an optimizer-step duplication problem." + }, + { + "left": "issue:43479", + "right": "issue:44869", + "accept": false, + "reason": "Multimodal config defaults and Whisper word-timestamp decoding are unrelated." + }, + { + "left": "issue:44206", + "right": "issue:45103", + "accept": false, + "reason": "A feature regression in LasrFeatureExtractor is unrelated to an auto-docstring crash from future annotations." + }, + { + "left": "issue:42831", + "right": "issue:44623", + "accept": false, + "reason": "FP8 accuracy issues and processor.save_pretrained missing files are different problems." + }, + { + "left": "issue:44222", + "right": "issue:44623", + "accept": false, + "reason": "Both mention saving, but model FP8 save_pretrained and processor.save_pretrained file omission are not the same bug." + }, + { + "left": "issue:30333", + "right": "issue:36331", + "accept": false, + "reason": "MLFlow reporting on failed jobs is unrelated to a Trainer compute_loss signature regression." + }, + { + "left": "issue:41628", + "right": "issue:42757", + "accept": false, + "reason": "Missing AutoImageProcessor and missing is_offline_mode are separate import breakages in different modules." + }, + { + "left": "issue:30333", + "right": "issue:34689", + "accept": false, + "reason": "MLFlow job reporting and Llama 3.2 model-loading breakage do not share a concrete code path." + }, + { + "left": "issue:43723", + "right": "issue:44977", + "accept": false, + "reason": "Tokenizer loading in v5 and Qwen3.5 flash-attention generation are different subsystems." + }, + { + "left": "issue:28282", + "right": "issue:30333", + "accept": false, + "reason": "A missing PyTorch ImportError and MLFlow reporting for failed training jobs are unrelated." + }, + { + "left": "issue:44297", + "right": "issue:45310", + "accept": false, + "reason": "Both are serialization/load regressions, but tokenizer_class mismatch and Qwen3.5 MoE from_pretrained failure are not the same defect." + }, + { + "left": "issue:30333", + "right": "issue:35141", + "accept": false, + "reason": "Failed-job MLFlow reporting and embedding reinitialization in post_init are unrelated." + }, + { + "left": "issue:44222", + "right": "issue:44484", + "accept": false, + "reason": "FP8 model save_pretrained behavior is unrelated to the max_shard_size default question." + }, + { + "left": "issue:44263", + "right": "issue:44360", + "accept": false, + "reason": "A torch.split return-value issue in one indexer is not the same as a missing ReLU in another." + }, + { + "left": "issue:44483", + "right": "issue:45406", + "accept": false, + "reason": "An API request rejection in v5.3 is unrelated to Gemma4Processor missing _tokenizer during serve." + }, + { + "left": "issue:44263", + "right": "issue:45092", + "accept": false, + "reason": "A DSA indexer torch.split bug is unrelated to remote-code checkpoint incompatibility with meta init." + }, + { + "left": "issue:43873", + "right": "issue:44945", + "accept": false, + "reason": "Quantization offloading behavior and pipeline-parallel output correctness are different runtime problems." + }, + { + "left": "issue:44373", + "right": "issue:44485", + "accept": false, + "reason": "A docstring error for position_ids is not the same as a RoPE implementation issue." + }, + { + "left": "issue:44514", + "right": "issue:45103", + "accept": false, + "reason": "A batched chat-template crash and an auto-docstring AttributeError are unrelated." + }, + { + "left": "issue:43479", + "right": "issue:43644", + "accept": false, + "reason": "Multimodal config default handling is unrelated to non-persistent buffer initialization junk." + }, + { + "left": "issue:44295", + "right": "issue:44623", + "accept": false, + "reason": "Reading a registered buffer and missing processor save files are different serialization problems." + }, + { + "left": "issue:43723", + "right": "issue:44871", + "accept": false, + "reason": "Tokenizer loading in v5 and eos_token_id configuration inconsistency are not the same issue." + }, + { + "left": "issue:45092", + "right": "issue:45357", + "accept": false, + "reason": "Remote-code meta-init incompatibility and incorrect visual encoder keys on save_pretrained are related only loosely by model loading, not by the same bug." + }, + { + "left": "issue:44261", + "right": "issue:45092", + "accept": false, + "reason": "A precision error from missing rms_norm_eps is unrelated to meta-init incompatibility for old checkpoints." + }, + { + "left": "issue:43479", + "right": "issue:44913", + "accept": false, + "reason": "Both involve config persistence, but None-handling in a multimodal config and rotary_pct reset on reload are different fields and code paths." + }, + { + "left": "issue:30990", + "right": "issue:38175", + "accept": false, + "reason": "Sentence Transformers loading stalls and SIGLIP2 zero probabilities are unrelated model/runtime problems." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 36, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-387", + "nodes": [ + "issue:22355", + "issue:28282", + "issue:30990", + "issue:33357", + "issue:41628", + "issue:42371", + "issue:42757", + "issue:42831", + "issue:42886", + "issue:42907", + "issue:42994", + "issue:43066", + "issue:43116", + "issue:43122", + "issue:43232", + "issue:43408", + "issue:43475", + "issue:43576", + "issue:43723", + "issue:43756", + "issue:43784", + "issue:43824", + "issue:43937", + "issue:44038", + "issue:44060", + "issue:44117", + "issue:44261", + "issue:44263", + "issue:44295", + "issue:44297", + "issue:44360", + "issue:44373", + "issue:44442", + "issue:44485", + "issue:44514", + "issue:44521", + "issue:44704", + "issue:44743", + "issue:44779", + "issue:44945", + "issue:45200", + "issue:45245", + "issue:45278", + "issue:45290", + "issue:45310", + "issue:45593", + "issue:45600", + "issue:45753" + ], + "soft_pairs": [ + "issue:22355|issue:28282", + "issue:42886|issue:42994", + "issue:41628|issue:43824", + "issue:42831|issue:45310", + "issue:44263|issue:44373", + "issue:44945|issue:45753", + "issue:42831|issue:44038", + "issue:42757|issue:45278", + "issue:43576|issue:43784", + "issue:42886|issue:43066", + "issue:44117|issue:44704", + "issue:43232|issue:45593", + "issue:42371|issue:44263", + "issue:43756|issue:44060", + "issue:30990|issue:33357", + "issue:44704|issue:45290", + "issue:42907|issue:44945", + "issue:43122|issue:44514", + "issue:44295|issue:44297", + "issue:43232|issue:45245", + "issue:43723|issue:44779", + "issue:43723|issue:44442", + "issue:43784|issue:45600", + "issue:43122|issue:44521", + "issue:42907|issue:43937", + "issue:44261|issue:44373", + "issue:41628|issue:45278", + "issue:43408|issue:43475", + "issue:44360|issue:44485", + "issue:43232|issue:44743", + "issue:43116|issue:45200" + ], + "prepared_review_unit_hash": "2bae3e6c3cce5a150a630b0ed494c7cafac813c167e8de5f00a86d26a518588c", + "component_budget": { + "node_count": 387, + "item_count": 387, + "soft_pair_count": 1455, + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 36499, + "estimated_input_tokens": 9125, + "estimated_eval_tokens": 18506 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a set of transformer v5-era regressions and import/tokenizer issues, but the soft pairs are still too heterogeneous to treat as duplicates. I would keep one broad tokenizer-loading regression as the representative issue, but reject all soft edges.", + "confidence": 0.86, + "canonical_issue_reason": "issue:43723 is the best representative because it is a broad, concrete v5 tokenizer-loading failure centered on `AutoTokenizer.from_pretrained`, which sits near several of the other tokenizer/import regressions in this cluster.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43723 is the strongest cluster anchor: it is specific enough to be actionable, yet broad enough to cover the common tokenizer-loading regression theme better than the narrower one-off reports.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:22355", + "right": "issue:28282", + "accept": false, + "reason": "Different failure modes: missing `transformers.onnx` module vs missing PyTorch for `AutoModel`." + }, + { + "left": "issue:42886", + "right": "issue:42994", + "accept": false, + "reason": "Offline tokenizer cache loading and quantized model saving are unrelated code paths." + }, + { + "left": "issue:41628", + "right": "issue:43824", + "accept": false, + "reason": "Both are import errors, but for different missing symbols and different model APIs." + }, + { + "left": "issue:42831", + "right": "issue:45310", + "accept": false, + "reason": "FineGrainedFP8 accuracy issue is unrelated to the Qwen3.5-Moe `from_pretrained` error." + }, + { + "left": "issue:44263", + "right": "issue:44373", + "accept": false, + "reason": "A runtime bug in `torch.split` is not the same as a wrong docstring." + }, + { + "left": "issue:44945", + "right": "issue:45753", + "accept": false, + "reason": "Pipeline parallelism output corruption and a Qwen3_5 image-specific loop are different bugs." + }, + { + "left": "issue:42831", + "right": "issue:44038", + "accept": false, + "reason": "Different models and symptoms: FP8 accuracy vs Qwen3-VL import/runtime bug." + }, + { + "left": "issue:42757", + "right": "issue:45278", + "accept": false, + "reason": "One is a specific missing import from `huggingface_hub`; the other is a broad post-upgrade import report." + }, + { + "left": "issue:43576", + "right": "issue:43784", + "accept": false, + "reason": "`transformers env` is broken, while the other is a `NameError` during sentence-transformers import." + }, + { + "left": "issue:42886", + "right": "issue:43066", + "accept": false, + "reason": "Both involve tokenizers, but cache/offline loading is not the same as wrong decoder type in v5." + }, + { + "left": "issue:44117", + "right": "issue:44704", + "accept": false, + "reason": "`TOKENIZER_MAPPING_NAMES` returning `None` and `AutoProcessor` not forwarding kwargs are different bugs." + }, + { + "left": "issue:43232", + "right": "issue:45593", + "accept": false, + "reason": "Generation kwargs handling and D-FINE auxiliary-loss behavior are unrelated." + }, + { + "left": "issue:42371", + "right": "issue:44263", + "accept": false, + "reason": "TF32 API guidance and a `torch.split` indexing bug do not match." + }, + { + "left": "issue:43756", + "right": "issue:44060", + "accept": false, + "reason": "Smollm3 RoPE-layer issue and Qwen3-Next tied-weights warning are different model-specific problems." + }, + { + "left": "issue:30990", + "right": "issue:33357", + "accept": false, + "reason": "Sentence-transformers loading stall and a MacOS bus error with CLIP are not the same underlying bug." + }, + { + "left": "issue:44704", + "right": "issue:45290", + "accept": false, + "reason": "`AutoProcessor.from_pretrained` kwargs handling and chat-template tool-call crashes are separate code paths." + }, + { + "left": "issue:42907", + "right": "issue:44945", + "accept": false, + "reason": "Quantized model saving failure and pipeline-parallelism output errors are unrelated." + }, + { + "left": "issue:43122", + "right": "issue:44514", + "accept": false, + "reason": "Tokenizer behavior changed between versions, but the batched chat-template crash is a different failure mode." + }, + { + "left": "issue:44295", + "right": "issue:44297", + "accept": false, + "reason": "A buffer-read error on `position_ids` is distinct from a tokenizer class mismatch in saved config." + }, + { + "left": "issue:43232", + "right": "issue:45245", + "accept": false, + "reason": "Generation state update after `sync_gpus` has nothing to do with category-count limits." + }, + { + "left": "issue:43723", + "right": "issue:44779", + "accept": false, + "reason": "Both are v5 tokenizer regressions, but one is loading/import failure and the other is incorrect tokenization output." + }, + { + "left": "issue:43723", + "right": "issue:44442", + "accept": false, + "reason": "`AutoTokenizer` load failure and FastSpeech2Conformer tokenizer loading failure are different registration issues." + }, + { + "left": "issue:43784", + "right": "issue:45600", + "accept": false, + "reason": "Sentence-transformers import `NameError` and Sam3LiteText auto-mapping CI breakage are unrelated." + }, + { + "left": "issue:43122", + "right": "issue:44521", + "accept": false, + "reason": "Tokenization differences and zeroed assistant masks in multimodal chat templates are different bugs." + }, + { + "left": "issue:42907", + "right": "issue:43937", + "accept": false, + "reason": "Saving a dequantized model and invalid `GenerationConfig` validation are unrelated." + }, + { + "left": "issue:44261", + "right": "issue:44373", + "accept": false, + "reason": "Missing `rms_norm_eps` in MLA config and a wrong docstring for `position_ids` are not the same bug." + }, + { + "left": "issue:41628", + "right": "issue:45278", + "accept": false, + "reason": "A single missing import and a broad upgrade-wide import regression are not a duplicate pair." + }, + { + "left": "issue:43408", + "right": "issue:43475", + "accept": false, + "reason": "Same SAM 3 area, but different problems: model-type warning vs missing attribute on output object." + }, + { + "left": "issue:44360", + "right": "issue:44485", + "accept": false, + "reason": "DSA indexer activation choice and GLM-5 RoPE implementation are unrelated." + }, + { + "left": "issue:43232", + "right": "issue:44743", + "accept": false, + "reason": "Generation kwargs handling and recurrent-state reset in modular Qwen3.5 are different code paths." + }, + { + "left": "issue:43116", + "right": "issue:45200", + "accept": false, + "reason": "Multi-label classification example behavior and Gemma 4 token-type defaults are unrelated." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 37, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-387", + "nodes": [ + "issue:22355", + "issue:28218", + "issue:29127", + "issue:30990", + "issue:36296", + "issue:39401", + "issue:42222", + "issue:42371", + "issue:42757", + "issue:42831", + "issue:42907", + "issue:43064", + "issue:43066", + "issue:43116", + "issue:43122", + "issue:43441", + "issue:43452", + "issue:43493", + "issue:43673", + "issue:43784", + "issue:43824", + "issue:43825", + "issue:43874", + "issue:43937", + "issue:43976", + "issue:43992", + "issue:44038", + "issue:44075", + "issue:44222", + "issue:44295", + "issue:44297", + "issue:44351", + "issue:44485", + "issue:44488", + "issue:44521", + "issue:44625", + "issue:44704", + "issue:44779", + "issue:44829", + "issue:44936", + "issue:44964", + "issue:44977", + "issue:45083", + "issue:45216", + "issue:45446", + "issue:45478", + "issue:45563", + "issue:45753" + ], + "soft_pairs": [ + "issue:44625|issue:44829", + "issue:29127|issue:30990", + "issue:43122|issue:44779", + "issue:42907|issue:44964", + "issue:44038|issue:44297", + "issue:42907|issue:43452", + "issue:44038|issue:44295", + "issue:39401|issue:44488", + "issue:36296|issue:43116", + "issue:43824|issue:44351", + "issue:44038|issue:44222", + "issue:43066|issue:44779", + "issue:39401|issue:44521", + "issue:22355|issue:28218", + "issue:45216|issue:45478", + "issue:42757|issue:44351", + "issue:43122|issue:44977", + "issue:43992|issue:44704", + "issue:43784|issue:45446", + "issue:42907|issue:44222", + "issue:43937|issue:44485", + "issue:42222|issue:43441", + "issue:43976|issue:45446", + "issue:42371|issue:43874", + "issue:44075|issue:44936", + "issue:43493|issue:45753", + "issue:42831|issue:45753", + "issue:43116|issue:43825", + "issue:44038|issue:45478", + "issue:44222|issue:45478", + "issue:43116|issue:43493", + "issue:43064|issue:43673", + "issue:45083|issue:45446", + "issue:43937|issue:45563" + ], + "prepared_review_unit_hash": "b96dcd1d307b5698630cfa6f662cf9dbd5a22b3abf3a393be358a69349b2d2a5", + "component_budget": { + "node_count": 387, + "item_count": 387, + "soft_pair_count": 1455, + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 34, + "serialized_chars": 37417, + "estimated_input_tokens": 9355, + "estimated_eval_tokens": 18966 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is a loose collection of unrelated Transformers issues. The soft-similarity pairs mostly share broad themes like tokenization, loading, training, or import errors, but they do not look like the same concrete underlying bug/change.", + "confidence": 0.93, + "canonical_issue_reason": "No single canonical issue fits the cluster: the items span unrelated failures across tokenization, model loading, training, multimodal processing, and API/import regressions.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 44779 is the clearest concrete regression-style report among the set, but only as a representative issue; it is not a true duplicate hub for the whole cluster.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44625", + "right": "issue:44829", + "accept": false, + "reason": "One is a Qwen3.5 num_labels propagation bug; the other is a flash_attention_3 training degeneration issue. Different code paths and symptoms." + }, + { + "left": "issue:29127", + "right": "issue:30990", + "accept": false, + "reason": "LayoutLMv3 error-message clarity and Sentence-Transformers loading hang are unrelated problems." + }, + { + "left": "issue:43122", + "right": "issue:44779", + "accept": false, + "reason": "Both concern tokenizer behavior in v5, but one is a general version-to-version tokenization difference and the other is a DeepSeek-specific regression. Too broad to call the same bug." + }, + { + "left": "issue:42907", + "right": "issue:44964", + "accept": false, + "reason": "Saving dequantized Ministral/Devstral models is a different failure from loading Phi-4 multimodal." + }, + { + "left": "issue:44038", + "right": "issue:44297", + "accept": false, + "reason": "Qwen3-VL-Moe import/runtime issue versus tokenizer_class mismatch on save_pretrained; not the same failure." + }, + { + "left": "issue:42907", + "right": "issue:43452", + "accept": false, + "reason": "Model save/dequantization failure and gguf_file load breakage are distinct loading/saving paths." + }, + { + "left": "issue:44038", + "right": "issue:44295", + "accept": false, + "reason": "Qwen3-VL-Moe bug and position_ids buffer read error are unrelated." + }, + { + "left": "issue:39401", + "right": "issue:44488", + "accept": false, + "reason": "Qwen3 tokenizer offset_mapping bug and loading cjvt/sleng-bert are different tokenizer/model-loading issues." + }, + { + "left": "issue:36296", + "right": "issue:43116", + "accept": false, + "reason": "Tensor-parallel training bug and multi-label classification example failure are unrelated." + }, + { + "left": "issue:43824", + "right": "issue:44351", + "accept": false, + "reason": "Qwen2.5-VL import failure and HybridCache import failure are separate missing-symbol regressions." + }, + { + "left": "issue:44038", + "right": "issue:44222", + "accept": false, + "reason": "Qwen3-VL-Moe runtime issue versus FP8 save_pretrained moe issue; different subsystems." + }, + { + "left": "issue:43066", + "right": "issue:44779", + "accept": false, + "reason": "Wrong tokenizer decoder type and DeepSeek tokenizer incorrect results are both tokenizer-related, but they target different concrete defects." + }, + { + "left": "issue:39401", + "right": "issue:44521", + "accept": false, + "reason": "Wrong offset_mapping and all-zero assistant_masks are different tokenizer/template behaviors." + }, + { + "left": "issue:22355", + "right": "issue:28218", + "accept": false, + "reason": "Missing transformers.onnx module and added-token spacing bug are unrelated." + }, + { + "left": "issue:45216", + "right": "issue:45478", + "accept": false, + "reason": "Qwen3.5 save_pretrained checkpoint regression and Qwen3.5 Moe from_pretrained error are related to the same model family but not the same bug." + }, + { + "left": "issue:42757", + "right": "issue:44351", + "accept": false, + "reason": "Import errors from different missing symbols in different libraries; no shared code-path." + }, + { + "left": "issue:43122", + "right": "issue:44977", + "accept": false, + "reason": "General tokenizer regression in v5 versus Qwen3.5 flash-attention generation issue; different failures." + }, + { + "left": "issue:43992", + "right": "issue:44704", + "accept": false, + "reason": "UMT5Encoder missing embed_tokens.weight and AutoProcessor kwargs forwarding are unrelated loading bugs." + }, + { + "left": "issue:43784", + "right": "issue:45446", + "accept": false, + "reason": "Sentence-Transformers import NameError and flex_attention version-check bug are unrelated." + }, + { + "left": "issue:42907", + "right": "issue:44222", + "accept": false, + "reason": "Dequantized model save failure and FP8 moe save_pretrained bug are both saving-related, but not the same concrete problem." + }, + { + "left": "issue:43937", + "right": "issue:44485", + "accept": false, + "reason": "GenerationConfig validation error and GLM-5 RoPE implementation discussion are different issues." + }, + { + "left": "issue:42222", + "right": "issue:43441", + "accept": false, + "reason": "VitPose import breakage and Ministral-3 FlashAttention failure are unrelated model-specific regressions." + }, + { + "left": "issue:43976", + "right": "issue:45446", + "accept": false, + "reason": "Python version compatibility report and flex_attention import/version check bug do not match." + }, + { + "left": "issue:42371", + "right": "issue:43874", + "accept": false, + "reason": "TF32 API guidance and missing get_number_of_image_patches are unrelated." + }, + { + "left": "issue:44075", + "right": "issue:44936", + "accept": false, + "reason": "Optimizer SGD args ignored and trainer.evaluate() failing after train() are distinct training issues." + }, + { + "left": "issue:43493", + "right": "issue:45753", + "accept": false, + "reason": "SigLIP2 HF vs JAX discrepancy and Qwen3_5 infinite loop on a specific image are different model problems." + }, + { + "left": "issue:42831", + "right": "issue:45753", + "accept": false, + "reason": "FineGrainedFP8 accuracy issue and Qwen3_5 image-specific infinite loop are unrelated." + }, + { + "left": "issue:43116", + "right": "issue:43825", + "accept": false, + "reason": "Example script multi-label output bug and pipeline translation-task error message bug are different example/API issues." + }, + { + "left": "issue:44038", + "right": "issue:45478", + "accept": false, + "reason": "Both involve Qwen3.5/Qwen3.5-Moe, but one is a load/runtime bug and the other is a from_pretrained error; not clearly the same defect." + }, + { + "left": "issue:44222", + "right": "issue:45478", + "accept": false, + "reason": "FP8 save_pretrained moe and Qwen3.5 Moe from_pretrained error are separate failure modes." + }, + { + "left": "issue:43116", + "right": "issue:43493", + "accept": false, + "reason": "Multi-label classification example failure and SigLIP2 implementation discrepancy are unrelated." + }, + { + "left": "issue:43064", + "right": "issue:43673", + "accept": false, + "reason": "Wrong optimizer states/params under FSDP2+PEFT and missing GenerationMixin cache in chunked_prefill are different training vs generation bugs." + }, + { + "left": "issue:45083", + "right": "issue:45446", + "accept": false, + "reason": "Helper length computation in qwen3_omni_moe and flex_attention import/version check are unrelated." + }, + { + "left": "issue:43937", + "right": "issue:45563", + "accept": false, + "reason": "GenerationConfig invalid and stale warning for num_return_sequences are different generation issues with different symptoms." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 38, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-387", + "nodes": [ + "issue:22355", + "issue:29127", + "issue:33453", + "issue:38617", + "issue:39401", + "issue:41628", + "issue:42175", + "issue:42222", + "issue:42673", + "issue:43012", + "issue:43066", + "issue:43122", + "issue:43381", + "issue:43408", + "issue:43421", + "issue:43450", + "issue:43577", + "issue:43742", + "issue:43827", + "issue:43937", + "issue:43986", + "issue:44075", + "issue:44117", + "issue:44220", + "issue:44291", + "issue:44410", + "issue:44483", + "issue:44484", + "issue:44496", + "issue:44779", + "issue:44829", + "issue:44938", + "issue:44945", + "issue:44964", + "issue:44977", + "issue:45071", + "issue:45072", + "issue:45081", + "issue:45216", + "issue:45278", + "issue:45310", + "issue:45405", + "issue:45446", + "issue:45464", + "issue:45542", + "issue:45588", + "issue:45656" + ], + "soft_pairs": [ + "issue:42175|issue:44938", + "issue:38617|issue:45278", + "issue:33453|issue:39401", + "issue:39401|issue:44779", + "issue:44117|issue:45588", + "issue:22355|issue:29127", + "issue:43827|issue:45446", + "issue:43381|issue:44945", + "issue:43937|issue:44977", + "issue:45405|issue:45446", + "issue:43450|issue:43986", + "issue:39401|issue:43421", + "issue:44117|issue:45081", + "issue:42673|issue:45072", + "issue:38617|issue:41628", + "issue:44829|issue:45216", + "issue:44410|issue:45310", + "issue:44291|issue:45071", + "issue:43408|issue:43742", + "issue:42222|issue:45542", + "issue:44484|issue:45310", + "issue:42175|issue:42222", + "issue:43937|issue:44496", + "issue:44483|issue:45464", + "issue:44075|issue:45656", + "issue:44964|issue:45278", + "issue:43012|issue:43577", + "issue:43066|issue:43122", + "issue:38617|issue:44220" + ], + "prepared_review_unit_hash": "da2aab9401052791b8656bcadb5ced002e18ffc1ff965ef17eb2c7ab02c3e982", + "component_budget": { + "node_count": 387, + "item_count": 387, + "soft_pair_count": 1455, + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 29, + "serialized_chars": 35653, + "estimated_input_tokens": 8914, + "estimated_eval_tokens": 18084 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The set is heterogeneous: it contains import regressions, tokenizer bugs, model-loading failures, dtype/backend issues, and docs/problems in unrelated subsystems. None of the soft pairs look like the same underlying bug, so I reject all merges.", + "confidence": 0.96, + "canonical_issue_reason": "No single canonical issue fits the cluster; the items span unrelated bug classes and should not be collapsed into one duplicate bucket.", + "canonical_pr_reason": null, + "best_issue_reason": "If forced to pick an umbrella, issue 45278 is the broadest import-error regression after upgrade, but it is still too general to serve as a safe duplicate canonical.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:42175", + "right": "issue:44938", + "accept": false, + "reason": "Both are import-related, but one is a missing TensorFlow backend in a torch install and the other is a Python 3.14 load failure." + }, + { + "left": "issue:38617", + "right": "issue:45278", + "accept": false, + "reason": "Specific missing-symbol import vs a broad report of many import errors after upgrade; not the same concrete bug." + }, + { + "left": "issue:33453", + "right": "issue:39401", + "accept": false, + "reason": "Tokenizer loading regression and Qwen3 offset-mapping bug are different failure modes." + }, + { + "left": "issue:39401", + "right": "issue:44779", + "accept": false, + "reason": "Both involve tokenization, but they target different models and symptoms, so this is too broad to merge." + }, + { + "left": "issue:44117", + "right": "issue:45588", + "accept": false, + "reason": "TOKENIZER_MAPPING_NAMES returning None and a flash-attention crash on s_aux=None are unrelated code paths." + }, + { + "left": "issue:22355", + "right": "issue:29127", + "accept": false, + "reason": "ONNX import failure and a LayoutLMv3 error-message issue are unrelated." + }, + { + "left": "issue:43827", + "right": "issue:45446", + "accept": false, + "reason": "Docs still referencing removed pipeline APIs is unrelated to the flex_attention import version check bug." + }, + { + "left": "issue:43381", + "right": "issue:44945", + "accept": false, + "reason": "Gradient-checkpointing in eval mode and incorrect output under pipeline parallelism are different problems." + }, + { + "left": "issue:43937", + "right": "issue:44977", + "accept": false, + "reason": "Both are Qwen3.5-related, but invalid GenerationConfig and flash-attention generation failure are distinct bugs." + }, + { + "left": "issue:45405", + "right": "issue:45446", + "accept": false, + "reason": "A PEFT version pin issue and a PyTorch version check bug are separate dependency problems." + }, + { + "left": "issue:43450", + "right": "issue:43986", + "accept": false, + "reason": "Batched video-shape handling and missing torchvision during AutoProcessor loading are different failures." + }, + { + "left": "issue:39401", + "right": "issue:43421", + "accept": false, + "reason": "Offset-mapping regression and runtime special-token post-processor updates are not the same issue." + }, + { + "left": "issue:44117", + "right": "issue:45081", + "accept": false, + "reason": "A tokenizer mapping None bug and a Mistral regex patch crash are unrelated internals." + }, + { + "left": "issue:42673", + "right": "issue:45072", + "accept": false, + "reason": "VRAM leak in Qwen3ForCausalLM and dtype mismatches in inference are different bugs." + }, + { + "left": "issue:38617", + "right": "issue:41628", + "accept": false, + "reason": "Both are import errors, but they involve different missing symbols and different root causes." + }, + { + "left": "issue:44829", + "right": "issue:45216", + "accept": false, + "reason": "Flash-attention training degeneration and broken save_pretrained checkpoints are not the same concrete code path." + }, + { + "left": "issue:44410", + "right": "issue:45310", + "accept": false, + "reason": "Both concern Qwen3.5/MoE loading, but one is missing layer projections and the other is a from_pretrained failure." + }, + { + "left": "issue:44291", + "right": "issue:45071", + "accept": false, + "reason": "init_empty_weights TypeError and PretrainedConfig type-checking breakage are distinct compatibility issues." + }, + { + "left": "issue:43408", + "right": "issue:43742", + "accept": false, + "reason": "A model-type warning for sam3_* and a key error loading MobileLLM are unrelated model-loading issues." + }, + { + "left": "issue:42222", + "right": "issue:45542", + "accept": false, + "reason": "Vitpose breakage and tensorboard-without-TensorFlow backend errors are separate subsystems." + }, + { + "left": "issue:44484", + "right": "issue:45310", + "accept": false, + "reason": "Chat/completions request rejection and Qwen3.5 MoE from_pretrained loading error are different failures." + }, + { + "left": "issue:42175", + "right": "issue:42222", + "accept": false, + "reason": "TensorFlow backend installation and vitpose model breakage do not share a concrete bug." + }, + { + "left": "issue:43937", + "right": "issue:44496", + "accept": false, + "reason": "Invalid GenerationConfig for GLM-5 and unrecognized model/config loading for Olmo are different model-specific errors." + }, + { + "left": "issue:44483", + "right": "issue:45464", + "accept": false, + "reason": "Chat/completions API rejection and Qwen3.5 streaming inference failure are not the same underlying issue." + }, + { + "left": "issue:44075", + "right": "issue:45656", + "accept": false, + "reason": "SGD args being ignored and deepspeed calling optimizer.step() twice are different optimizer bugs." + }, + { + "left": "issue:44964", + "right": "issue:45278", + "accept": false, + "reason": "A Phi-4 multimodal load failure and a broad upgrade-related import-error report are not the same concrete bug." + }, + { + "left": "issue:43012", + "right": "issue:43577", + "accept": false, + "reason": "A PyTorch warning during bf16 compile and Blip2 dtype staying float32 are related only loosely by dtype, not by bug." + }, + { + "left": "issue:43066", + "right": "issue:43122", + "accept": false, + "reason": "Both are tokenizer regressions, but one is decoder-type handling in v5 and the other is a version-to-version tokenization change." + }, + { + "left": "issue:38617", + "right": "issue:44220", + "accept": false, + "reason": "Missing configuration import and fbank feature extraction are unrelated failures." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 39, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-387", + "nodes": [ + "issue:38617", + "issue:41628", + "issue:42222", + "issue:42371", + "issue:42673", + "issue:42757", + "issue:42831", + "issue:43012", + "issue:43064", + "issue:43097", + "issue:43335", + "issue:43408", + "issue:43441", + "issue:43452", + "issue:43577", + "issue:43874", + "issue:43906", + "issue:43931", + "issue:43976", + "issue:43986", + "issue:44038", + "issue:44062", + "issue:44117", + "issue:44261", + "issue:44263", + "issue:44315", + "issue:44351", + "issue:44368", + "issue:44410", + "issue:44464", + "issue:44484", + "issue:44485", + "issue:44493", + "issue:44704", + "issue:44829", + "issue:44945", + "issue:44964", + "issue:44977", + "issue:44987", + "issue:45071", + "issue:45103", + "issue:45245", + "issue:45341", + "issue:45375", + "issue:45459", + "issue:45478", + "issue:45542", + "issue:45588" + ], + "soft_pairs": [ + "issue:42371|issue:44261", + "issue:43097|issue:44493", + "issue:43874|issue:44261", + "issue:44484|issue:45478", + "issue:44410|issue:45478", + "issue:42222|issue:45459", + "issue:44263|issue:44485", + "issue:43441|issue:44829", + "issue:43012|issue:43408", + "issue:41628|issue:45071", + "issue:45071|issue:45542", + "issue:42673|issue:44977", + "issue:42371|issue:42831", + "issue:42673|issue:43906", + "issue:44117|issue:44987", + "issue:43097|issue:43452", + "issue:43986|issue:44484", + "issue:43335|issue:45071", + "issue:43931|issue:44038", + "issue:43976|issue:44945", + "issue:43408|issue:44368", + "issue:43874|issue:44263", + "issue:44829|issue:45245", + "issue:38617|issue:42757", + "issue:43064|issue:44315", + "issue:44464|issue:45071", + "issue:43097|issue:45588", + "issue:44062|issue:44117", + "issue:44410|issue:45375", + "issue:38617|issue:44351", + "issue:42673|issue:43577", + "issue:44704|issue:45103", + "issue:44038|issue:44410", + "issue:44261|issue:44485", + "issue:44964|issue:45341", + "issue:43986|issue:44117", + "issue:43931|issue:44410" + ], + "prepared_review_unit_hash": "0ee2530573861c0d660ef65316ef239599cfd141ff25d110e2a70e6419e8d31a", + "component_budget": { + "node_count": 387, + "item_count": 387, + "soft_pair_count": 1455, + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 37, + "serialized_chars": 38815, + "estimated_input_tokens": 9704, + "estimated_eval_tokens": 19664 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is highly heterogeneous: most pairs only share broad Transformers-v5/API similarity, not the same concrete bug or change. No soft edge looks safe to merge as a duplicate.", + "confidence": 0.93, + "canonical_issue_reason": "No single canonical issue cleanly subsumes this set; the items span unrelated import errors, model-specific loading bugs, config regressions, and training/runtime issues.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43097 is the best umbrella representative because it is a widely referenced version-wide breaking change with broader downstream impact than the others, though it still does not cover the whole cluster.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:42371", + "right": "issue:44261", + "accept": false, + "reason": "Different bugs: TF32 API guidance vs a FineGrainedFP8/precision regression." + }, + { + "left": "issue:43097", + "right": "issue:44493", + "accept": false, + "reason": "Different regressions: removed config flag vs unexpected position-id key warnings." + }, + { + "left": "issue:43874", + "right": "issue:44261", + "accept": false, + "reason": "Different code paths: missing image-patch helper vs RMS eps precision issue." + }, + { + "left": "issue:44484", + "right": "issue:45478", + "accept": false, + "reason": "Unrelated: shard-size default question vs Qwen3.5 MoE loading error." + }, + { + "left": "issue:44410", + "right": "issue:45478", + "accept": false, + "reason": "Same broad model family, but different defects and fixes." + }, + { + "left": "issue:42222", + "right": "issue:45459", + "accept": false, + "reason": "VitPose breakage vs tokenizer protobuf error; no shared failure mode." + }, + { + "left": "issue:44263", + "right": "issue:44485", + "accept": false, + "reason": "Different GLM issues: torch.split handling vs RoPE implementation." + }, + { + "left": "issue:43441", + "right": "issue:44829", + "accept": false, + "reason": "Both flash-attention related, but one is inference breakage and the other is training degeneration." + }, + { + "left": "issue:43012", + "right": "issue:43408", + "accept": false, + "reason": "Float16/bfloat16 warning vs SAM3 model-type warning; unrelated." + }, + { + "left": "issue:41628", + "right": "issue:45071", + "accept": false, + "reason": "Import error for AutoImageProcessor vs PretrainedConfig type-checking regression." + }, + { + "left": "issue:45071", + "right": "issue:45542", + "accept": false, + "reason": "Config type checking vs TensorFlow backend detection; different code paths." + }, + { + "left": "issue:42673", + "right": "issue:44977", + "accept": false, + "reason": "VRAM leak in dataloader threads vs flash-attention misgeneration." + }, + { + "left": "issue:42371", + "right": "issue:42831", + "accept": false, + "reason": "TF32 settings warning and FineGrainedFP8 accuracy are unrelated." + }, + { + "left": "issue:42673", + "right": "issue:43906", + "accept": false, + "reason": "Different reports: a Qwen VRAM leak vs an isolated reproduction of another issue." + }, + { + "left": "issue:44117", + "right": "issue:44987", + "accept": false, + "reason": "Tokenizer mapping None handling vs a specific model loading failure." + }, + { + "left": "issue:43097", + "right": "issue:43452", + "accept": false, + "reason": "Config deprecation vs gguf_file loading breakage; not the same bug." + }, + { + "left": "issue:43986", + "right": "issue:44484", + "accept": false, + "reason": "AutoProcessor/torchvision crash vs save_pretrained shard-size behavior." + }, + { + "left": "issue:43335", + "right": "issue:45071", + "accept": false, + "reason": "SwitchTransformers sparse-layer creation bug vs config type-checking regression." + }, + { + "left": "issue:43931", + "right": "issue:44038", + "accept": false, + "reason": "Different Qwen-VL failures: weight-shape mismatch vs Qwen3-VL-Moe incompatibility." + }, + { + "left": "issue:43976", + "right": "issue:44945", + "accept": false, + "reason": "Python version compatibility report vs pipeline-parallelism output bug." + }, + { + "left": "issue:43408", + "right": "issue:44368", + "accept": false, + "reason": "Model type mismatch warning vs tie_word_embeddings warning; separate issues." + }, + { + "left": "issue:43874", + "right": "issue:44263", + "accept": false, + "reason": "Different GLM failures: missing helper method vs torch.split semantics." + }, + { + "left": "issue:44829", + "right": "issue:45245", + "accept": false, + "reason": "FlashAttention-3 training regression vs a category-count runtime limit." + }, + { + "left": "issue:38617", + "right": "issue:42757", + "accept": false, + "reason": "Both are import errors, but they are different missing symbols with different causes." + }, + { + "left": "issue:43064", + "right": "issue:44315", + "accept": false, + "reason": "Distributed optimizer-state bug vs Liger Kernel not applied; unrelated." + }, + { + "left": "issue:44464", + "right": "issue:45071", + "accept": false, + "reason": "Chunked generation inconsistency under compiled forward vs config type-checking regression." + }, + { + "left": "issue:43097", + "right": "issue:45588", + "accept": false, + "reason": "Config removal vs flash_attention s_aux crash; different code paths." + }, + { + "left": "issue:44062", + "right": "issue:44117", + "accept": false, + "reason": "AddedToken construction bug vs tokenizer mapping None handling." + }, + { + "left": "issue:44410", + "right": "issue:45375", + "accept": false, + "reason": "Qwen3Next missing projections vs a vision-config field being dropped; similar family, different bug." + }, + { + "left": "issue:38617", + "right": "issue:44351", + "accept": false, + "reason": "Separate import errors for different missing transformers symbols." + }, + { + "left": "issue:42673", + "right": "issue:43577", + "accept": false, + "reason": "VRAM leak vs BLIP2 dtype propagation bug." + }, + { + "left": "issue:44704", + "right": "issue:45103", + "accept": false, + "reason": "Kwargs forwarding bug vs auto-docstring crash from future annotations." + }, + { + "left": "issue:44038", + "right": "issue:44410", + "accept": false, + "reason": "Different Qwen failures: MoE loading issue vs missing attention projections." + }, + { + "left": "issue:44261", + "right": "issue:44485", + "accept": false, + "reason": "MLA q_a_layernorm precision issue vs GLM-5 RoPE implementation; unrelated." + }, + { + "left": "issue:44964", + "right": "issue:45341", + "accept": false, + "reason": "Model loading failure vs a testing_utils bug." + }, + { + "left": "issue:43986", + "right": "issue:44117", + "accept": false, + "reason": "AutoProcessor crash without torchvision vs tokenizer mapping None handling." + }, + { + "left": "issue:43931", + "right": "issue:44410", + "accept": false, + "reason": "Qwen3-VL weight mismatch vs Qwen3Next missing projections; distinct bugs." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 40, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-387", + "nodes": [ + "issue:36683", + "issue:38617", + "issue:42617", + "issue:42757", + "issue:42886", + "issue:42907", + "issue:42994", + "issue:43012", + "issue:43064", + "issue:43066", + "issue:43097", + "issue:43208", + "issue:43299", + "issue:43352", + "issue:43381", + "issue:43475", + "issue:43502", + "issue:43526", + "issue:43531", + "issue:43550", + "issue:43618", + "issue:43761", + "issue:43824", + "issue:43825", + "issue:43867", + "issue:43873", + "issue:43901", + "issue:43906", + "issue:43937", + "issue:43986", + "issue:43992", + "issue:44038", + "issue:44230", + "issue:44368", + "issue:44380", + "issue:44483", + "issue:44484", + "issue:44514", + "issue:44655", + "issue:44683", + "issue:45216", + "issue:45290", + "issue:45341", + "issue:45372", + "issue:45381", + "issue:45561", + "issue:45588", + "issue:45753" + ], + "soft_pairs": [ + "issue:43502|issue:43992", + "issue:43097|issue:43873", + "issue:43066|issue:43906", + "issue:43986|issue:45290", + "issue:43986|issue:45372", + "issue:45341|issue:45561", + "issue:43986|issue:45341", + "issue:43097|issue:43901", + "issue:43825|issue:44655", + "issue:42617|issue:45341", + "issue:43992|issue:44484", + "issue:45341|issue:45588", + "issue:42886|issue:43381", + "issue:43208|issue:43550", + "issue:43012|issue:43550", + "issue:43012|issue:44368", + "issue:42907|issue:45753", + "issue:42907|issue:44655", + "issue:43531|issue:45341", + "issue:43208|issue:43526", + "issue:44483|issue:44514", + "issue:42757|issue:43824", + "issue:43012|issue:44683", + "issue:44483|issue:45290", + "issue:44038|issue:44230", + "issue:43618|issue:43761", + "issue:43064|issue:44380", + "issue:38617|issue:43502", + "issue:44483|issue:45381", + "issue:43299|issue:43867", + "issue:43901|issue:44655", + "issue:36683|issue:43475", + "issue:43867|issue:45216", + "issue:42994|issue:44655", + "issue:43352|issue:43937", + "issue:43066|issue:45381" + ], + "prepared_review_unit_hash": "4ee419f0bc2a8a144b8d2c841f72f633908cb67c63025d7ef42f3a97a03b2f5c", + "component_budget": { + "node_count": 387, + "item_count": 387, + "soft_pair_count": 1455, + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 36, + "serialized_chars": 38197, + "estimated_input_tokens": 9550, + "estimated_eval_tokens": 19356 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "No soft pair looks like the same underlying bug/change. The items mostly split into unrelated v5 regressions, model-specific crashes, saving/loading bugs, and test-utils issues. The most representative issue for the cluster is the broad v5 API regression around removed behavior.", + "confidence": 0.55, + "canonical_issue_reason": "issue:43097 is the broadest and most representative regression here: it\u2019s explicitly tied to v5.0.0, describes a removed public API without guidance, and has inbound references, making it a better cluster anchor than the model-specific crashes.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43097 is the best single issue to surface because it captures a user-facing v5 breaking change with broader impact than the narrower model or pipeline-specific reports.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43502", + "right": "issue:43992", + "accept": false, + "reason": "Local-files-only network leakage and missing embed_tokens.weight are different loading bugs." + }, + { + "left": "issue:43097", + "right": "issue:43873", + "accept": false, + "reason": "Removed v5 API behavior vs quantized offloading failure; different code paths and symptoms." + }, + { + "left": "issue:43066", + "right": "issue:43906", + "accept": false, + "reason": "Wrong tokenizer decoder type is unrelated to the unspecified isolated reproduction issue." + }, + { + "left": "issue:43986", + "right": "issue:45290", + "accept": false, + "reason": "AutoProcessor dependency crash vs chat-template tool-call crash are distinct failures." + }, + { + "left": "issue:43986", + "right": "issue:45372", + "accept": false, + "reason": "Missing torchvision crash is unrelated to mistral_common import failure in Gemma 4 loading." + }, + { + "left": "issue:45341", + "right": "issue:45561", + "accept": false, + "reason": "Generic testing_utils bug vs xdist race on captured_info.txt are not the same defect." + }, + { + "left": "issue:43986", + "right": "issue:45341", + "accept": false, + "reason": "Processor loading crash and a testing_utils bug do not share the same underlying issue." + }, + { + "left": "issue:43097", + "right": "issue:43901", + "accept": false, + "reason": "Breaking API removal vs stale docs for return_all_scores are different problems." + }, + { + "left": "issue:43825", + "right": "issue:44655", + "accept": false, + "reason": "Pipeline task messaging issue and save_pretrained serialization failure are unrelated." + }, + { + "left": "issue:42617", + "right": "issue:45341", + "accept": false, + "reason": "3d_parallel.py execution failure is unrelated to a testing-utils report." + }, + { + "left": "issue:43992", + "right": "issue:44484", + "accept": false, + "reason": "Missing checkpoint key vs max_shard_size default question are different save/load concerns." + }, + { + "left": "issue:45341", + "right": "issue:45588", + "accept": false, + "reason": "Testing utils bug vs flash_attention AttributeError on sink-less models are unrelated." + }, + { + "left": "issue:42886", + "right": "issue:43381", + "accept": false, + "reason": "Offline tokenizer cache loading and gradient-checkpointing eval-mode behavior are different code paths." + }, + { + "left": "issue:43208", + "right": "issue:43550", + "accept": false, + "reason": "xLSTM training bugs are unrelated to Bamba torch.compile/SDPA failure." + }, + { + "left": "issue:43012", + "right": "issue:43550", + "accept": false, + "reason": "bf16 warning during compile is not the same as a compile-time SDPA crash." + }, + { + "left": "issue:43012", + "right": "issue:44368", + "accept": false, + "reason": "PyTorch precision warning and tie_word_embeddings warning are different warnings with different causes." + }, + { + "left": "issue:42907", + "right": "issue:45753", + "accept": false, + "reason": "Quantized model save failure is unrelated to a Qwen3_5 image-generation infinite loop." + }, + { + "left": "issue:42907", + "right": "issue:44655", + "accept": false, + "reason": "Saving dequantized models and saving Pipeline objects are different serialization problems." + }, + { + "left": "issue:43531", + "right": "issue:45341", + "accept": false, + "reason": "Qwen3-MoE sliding_window issue is unrelated to testing_utils." + }, + { + "left": "issue:43208", + "right": "issue:43526", + "accept": false, + "reason": "xLSTM training bugs do not match BeitImageProcessorFast reduce_labels behavior." + }, + { + "left": "issue:44483", + "right": "issue:44514", + "accept": false, + "reason": "v1 chat/completions rejection and batched Qwen2.5-VL chat_template crash are different failures." + }, + { + "left": "issue:42757", + "right": "issue:43824", + "accept": false, + "reason": "Import errors for different symbols/modules are not duplicates." + }, + { + "left": "issue:43012", + "right": "issue:44683", + "accept": false, + "reason": "bfloat16 warning and compiled flex_attention failure are distinct issues." + }, + { + "left": "issue:44483", + "right": "issue:45290", + "accept": false, + "reason": "Chat/completions endpoint rejection is unrelated to assistant tool-call template crashes." + }, + { + "left": "issue:44038", + "right": "issue:44230", + "accept": false, + "reason": "Qwen3-VL-Moe regression and Qwen3.5 fp8 inference support are different requests." + }, + { + "left": "issue:43618", + "right": "issue:43761", + "accept": false, + "reason": "Missing attentions assignment and missing hidden_states are separate CLIP regressions." + }, + { + "left": "issue:43064", + "right": "issue:44380", + "accept": false, + "reason": "FSDP2 optimizer-state corruption and GPT2 attention scaling under SDPA are unrelated." + }, + { + "left": "issue:38617", + "right": "issue:43502", + "accept": false, + "reason": "ImportError in configuration_utils and network requests despite local_files_only are different bugs." + }, + { + "left": "issue:44483", + "right": "issue:45381", + "accept": false, + "reason": "Chat/completions API rejection and Qwen2.5-VL video position_ids are unrelated." + }, + { + "left": "issue:43299", + "right": "issue:43867", + "accept": false, + "reason": "Qwen3VL loading breakage and sorted-state_dict load failure are not the same defect." + }, + { + "left": "issue:43901", + "right": "issue:44655", + "accept": false, + "reason": "Documentation mismatch and Pipeline save_pretrained failure are unrelated." + }, + { + "left": "issue:36683", + "right": "issue:43475", + "accept": false, + "reason": "Missing vocab_size on Gemma3Config and missing fpn_position_embeddings on SAM3 output are different attribute errors." + }, + { + "left": "issue:43867", + "right": "issue:45216", + "accept": false, + "reason": "State_dict sorting load error and incorrect Qwen3.5 checkpoint saving are distinct serialization issues." + }, + { + "left": "issue:42994", + "right": "issue:44655", + "accept": false, + "reason": "Quantized model saving failure is not the same as saving Pipeline objects." + }, + { + "left": "issue:43352", + "right": "issue:43937", + "accept": false, + "reason": "FlashAttention support error and invalid GenerationConfig are unrelated." + }, + { + "left": "issue:43066", + "right": "issue:45381", + "accept": false, + "reason": "Tokenizer decoder type regression and video position_ids bug are different problems." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 41, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-387", + "nodes": [ + "issue:28282", + "issue:33453", + "issue:36010", + "issue:36296", + "issue:36683", + "issue:41669", + "issue:42757", + "issue:42890", + "issue:43064", + "issue:43097", + "issue:43299", + "issue:43352", + "issue:43441", + "issue:43525", + "issue:43575", + "issue:43618", + "issue:43756", + "issue:43867", + "issue:43881", + "issue:43906", + "issue:43976", + "issue:43992", + "issue:44077", + "issue:44117", + "issue:44220", + "issue:44242", + "issue:44246", + "issue:44263", + "issue:44355", + "issue:44360", + "issue:44380", + "issue:44484", + "issue:44485", + "issue:44492", + "issue:44655", + "issue:44855", + "issue:44857", + "issue:44938", + "issue:45216", + "issue:45310", + "issue:45357", + "issue:45375", + "issue:45440", + "issue:45478", + "issue:45479", + "issue:45561", + "issue:45588", + "issue:45753" + ], + "soft_pairs": [ + "issue:36010|issue:42757", + "issue:44220|issue:44655", + "issue:44263|issue:45440", + "issue:43756|issue:44485", + "issue:36683|issue:43525", + "issue:43618|issue:45216", + "issue:43906|issue:45753", + "issue:36683|issue:44242", + "issue:36683|issue:43881", + "issue:43299|issue:45753", + "issue:43352|issue:43976", + "issue:43441|issue:44380", + "issue:43352|issue:44380", + "issue:43992|issue:45310", + "issue:44380|issue:45479", + "issue:41669|issue:44246", + "issue:43867|issue:45375", + "issue:43064|issue:44492", + "issue:43618|issue:44380", + "issue:43352|issue:45478", + "issue:44355|issue:44855", + "issue:36683|issue:44117", + "issue:42890|issue:45561", + "issue:43097|issue:43352", + "issue:43867|issue:45357", + "issue:43976|issue:44938", + "issue:43992|issue:45478", + "issue:44360|issue:45440", + "issue:43064|issue:44655", + "issue:36296|issue:43575", + "issue:44077|issue:44380", + "issue:43618|issue:45588", + "issue:28282|issue:33453", + "issue:43867|issue:44484", + "issue:43618|issue:44857" + ], + "prepared_review_unit_hash": "2b5fac5231b0b5bb7f9306640c1ab5f5f79e964e0d3e40aa312eb234c17e15f4", + "component_budget": { + "node_count": 387, + "item_count": 387, + "soft_pair_count": 1455, + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 35, + "serialized_chars": 37791, + "estimated_input_tokens": 9448, + "estimated_eval_tokens": 19152 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly heterogeneous model-specific bug reports and regressions. The only strong duplicate/soft-match is the import-time slowdown pair (`import transformers takes long sometimes` vs. the `import *` performance regression/fix); the rest are different code paths, models, or symptoms and should stay separate.", + "confidence": 0.89, + "canonical_issue_reason": "`issue:44246` is the broadest, most general issue in the set and best represents the import-performance theme; `issue:41669` reads as a specific cause/fix for that same slowdown.", + "canonical_pr_reason": null, + "best_issue_reason": "`issue:44246` is the best representative issue because it describes the user-facing symptom without narrowing to one implementation detail or model.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:36010", + "right": "issue:42757", + "accept": false, + "reason": "Both are import errors, but they involve different missing symbols in different packages and different fixes." + }, + { + "left": "issue:44220", + "right": "issue:44655", + "accept": false, + "reason": "Audio feature extraction bug vs. pipeline save_pretrained problem; unrelated code paths." + }, + { + "left": "issue:44263", + "right": "issue:45440", + "accept": false, + "reason": "Different model internals and failures; no shared underlying bug." + }, + { + "left": "issue:43756", + "right": "issue:44485", + "accept": false, + "reason": "Both mention RoPE, but they target different models and different implementation issues." + }, + { + "left": "issue:36683", + "right": "issue:43525", + "accept": false, + "reason": "Both are missing config attributes, but for different models and different fields." + }, + { + "left": "issue:43618", + "right": "issue:45216", + "accept": false, + "reason": "CLIP attentions assignment bug vs. Qwen save_pretrained regression; not the same issue." + }, + { + "left": "issue:43906", + "right": "issue:45753", + "accept": false, + "reason": "One is a reproduction of another issue; the other is an infinite-loop report on a specific image." + }, + { + "left": "issue:36683", + "right": "issue:44242", + "accept": false, + "reason": "Config attribute bug vs. MoE load-balancing-loss logic bug; different failures." + }, + { + "left": "issue:36683", + "right": "issue:43881", + "accept": false, + "reason": "Different model-loading failure and different missing field; not a duplicate." + }, + { + "left": "issue:43299", + "right": "issue:45753", + "accept": false, + "reason": "Both are Qwen3.5-related, but one is MoE loading and the other is an image-specific generation loop." + }, + { + "left": "issue:43352", + "right": "issue:43976", + "accept": false, + "reason": "FlashAttention model-support issue vs. Python-version compatibility issue." + }, + { + "left": "issue:43441", + "right": "issue:44380", + "accept": false, + "reason": "Both touch attention backends, but they concern different models and different bugs." + }, + { + "left": "issue:43352", + "right": "issue:44380", + "accept": false, + "reason": "Generic FlashAttention backend behavior vs. model-specific unsupported-feature error." + }, + { + "left": "issue:43992", + "right": "issue:45310", + "accept": false, + "reason": "Different Qwen/UMT5 loading problems with no clear shared code-path." + }, + { + "left": "issue:44380", + "right": "issue:45479", + "accept": false, + "reason": "Attention scaling bug vs. classification loss degeneration; unrelated." + }, + { + "left": "issue:41669", + "right": "issue:44246", + "accept": true, + "reason": "These describe the same import-time slowdown: one identifies the `import *` cause, the other reports the user-visible slow import symptom." + }, + { + "left": "issue:43867", + "right": "issue:45375", + "accept": false, + "reason": "Both are load/config issues, but one is about state_dict ordering and the other is a missing vision config field." + }, + { + "left": "issue:43064", + "right": "issue:44492", + "accept": false, + "reason": "Distributed training optimizer-state bug vs. a cache-strategy typo; unrelated." + }, + { + "left": "issue:43618", + "right": "issue:44380", + "accept": false, + "reason": "Different attention/output bugs in different model families." + }, + { + "left": "issue:43352", + "right": "issue:45478", + "accept": false, + "reason": "Same general FlashAttention theme, but different models and distinct breakage." + }, + { + "left": "issue:44355", + "right": "issue:44855", + "accept": false, + "reason": "Compiled-Python import error vs. TorchScript parsing error; different root causes." + }, + { + "left": "issue:36683", + "right": "issue:44117", + "accept": false, + "reason": "Missing config attribute vs. tokenizer mapping returning None; separate failures." + }, + { + "left": "issue:42890", + "right": "issue:45561", + "accept": false, + "reason": "Both are test-related, but one is missing set_seed and the other is an xdist file-write race." + }, + { + "left": "issue:43097", + "right": "issue:43352", + "accept": false, + "reason": "Tie-embeddings API removal vs. FlashAttention support error; not the same bug." + }, + { + "left": "issue:43867", + "right": "issue:45357", + "accept": false, + "reason": "Different Qwen3.5 save/load regressions with no shared underlying code-path." + }, + { + "left": "issue:43976", + "right": "issue:44938", + "accept": false, + "reason": "Python version compatibility issue vs. Python 3.14 import failure; related only loosely." + }, + { + "left": "issue:43992", + "right": "issue:45478", + "accept": false, + "reason": "Different Qwen3.5 loading failures; one is UMT5 encoder weights, the other is MoE from_pretrained." + }, + { + "left": "issue:44360", + "right": "issue:45440", + "accept": false, + "reason": "DSA indexer math bug vs. DeepseekV3MoE implementation divergence; not the same issue." + }, + { + "left": "issue:43064", + "right": "issue:44655", + "accept": false, + "reason": "Optimizer-state corruption in training vs. pipeline save_pretrained failure; unrelated." + }, + { + "left": "issue:36296", + "right": "issue:43575", + "accept": false, + "reason": "Tensor-parallel training bug vs. load-time OOM for a specific model." + }, + { + "left": "issue:44077", + "right": "issue:44380", + "accept": false, + "reason": "Optional post_init restriction vs. attention scaling ignored; different functionality." + }, + { + "left": "issue:43618", + "right": "issue:45588", + "accept": false, + "reason": "CLIP attentions assignment bug vs. flash_attention crash on sink-less models; different models and code paths." + }, + { + "left": "issue:28282", + "right": "issue:33453", + "accept": false, + "reason": "PyTorch-missing ImportError vs. tokenizer loading regression; unrelated root causes." + }, + { + "left": "issue:43867", + "right": "issue:44484", + "accept": false, + "reason": "State-dict loading issue vs. max_shard_size default question; not duplicates." + }, + { + "left": "issue:43618", + "right": "issue:44857", + "accept": false, + "reason": "Different model-specific runtime errors with no shared underlying bug." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet, and the only accepted soft edge (issue:41669 \u2194 issue:44246) is a conservative duplicate match based on the shared import-time slowdown theme. The remaining pairwise rejections are appropriately cautious and do not overclaim duplicates or shared root causes." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 42, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-387", + "nodes": [ + "issue:30333", + "issue:30990", + "issue:31795", + "issue:33453", + "issue:34689", + "issue:36296", + "issue:42548", + "issue:42757", + "issue:42831", + "issue:42890", + "issue:43010", + "issue:43299", + "issue:43502", + "issue:43576", + "issue:43618", + "issue:43723", + "issue:43824", + "issue:43827", + "issue:43874", + "issue:43901", + "issue:44062", + "issue:44075", + "issue:44112", + "issue:44222", + "issue:44230", + "issue:44315", + "issue:44351", + "issue:44355", + "issue:44360", + "issue:44368", + "issue:44387", + "issue:44485", + "issue:44556", + "issue:44655", + "issue:44749", + "issue:44938", + "issue:44961", + "issue:44964", + "issue:44977", + "issue:45071", + "issue:45125", + "issue:45278", + "issue:45341", + "issue:45375", + "issue:45468", + "issue:45478", + "issue:45600", + "issue:45656" + ], + "soft_pairs": [ + "issue:43618|issue:44977", + "issue:43901|issue:44938", + "issue:44655|issue:44938", + "issue:42890|issue:44964", + "issue:42890|issue:44112", + "issue:42890|issue:45600", + "issue:43502|issue:44062", + "issue:43010|issue:44075", + "issue:42890|issue:45341", + "issue:42548|issue:44351", + "issue:44222|issue:45125", + "issue:43502|issue:44485", + "issue:44355|issue:44749", + "issue:30990|issue:31795", + "issue:45125|issue:45478", + "issue:45125|issue:45375", + "issue:44387|issue:44556", + "issue:43010|issue:45278", + "issue:30333|issue:33453", + "issue:43723|issue:44556", + "issue:43010|issue:44368", + "issue:43502|issue:44222", + "issue:43502|issue:44961", + "issue:44315|issue:44360", + "issue:42831|issue:43874", + "issue:30333|issue:36296", + "issue:44355|issue:44655", + "issue:44355|issue:45071", + "issue:44222|issue:44230", + "issue:43827|issue:44961", + "issue:43010|issue:45656", + "issue:33453|issue:34689", + "issue:43299|issue:44556", + "issue:42757|issue:43576", + "issue:33453|issue:36296", + "issue:43576|issue:45071", + "issue:43576|issue:43824", + "issue:44961|issue:45468" + ], + "prepared_review_unit_hash": "c347ea7360e09f82d5c604cc106352e32a55a887a9531993145c9d6527cc7544", + "component_budget": { + "node_count": 387, + "item_count": 387, + "soft_pair_count": 1455, + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 38, + "serialized_chars": 38503, + "estimated_input_tokens": 9626, + "estimated_eval_tokens": 19508 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a grab bag of unrelated Transformers v5 regressions, import/load failures, and a few docs/tests issues. The broadest representative is the open umbrella report about many import errors after upgrading to 5.5.0; the rest are not close enough to treat as duplicates.", + "confidence": 0.78, + "canonical_issue_reason": "Issue 45278 is the best umbrella for this cluster because it broadly captures the recurring theme of import/load regressions after upgrading to v5, while the other issues are mostly narrower variants or unrelated docs/test failures.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 45278 is the most useful global representative: it is broad, current, and aligns with the dominant pattern in the cluster (v5 upgrade regressions causing import/load breakage).", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43618", + "right": "issue:44977", + "accept": false, + "reason": "Different model families and symptoms: CLIP attentions assignment vs Qwen3.5 flash-attention generation." + }, + { + "left": "issue:43901", + "right": "issue:44938", + "accept": false, + "reason": "Docs behavior mismatch vs Python 3.14 load failure; unrelated problems." + }, + { + "left": "issue:44655", + "right": "issue:44938", + "accept": false, + "reason": "Pipeline save_pretrained issue is not the same as interpreter compatibility/loading failure." + }, + { + "left": "issue:42890", + "right": "issue:44964", + "accept": false, + "reason": "Different model/test failure and a separate latest-transformers loading regression." + }, + { + "left": "issue:42890", + "right": "issue:44112", + "accept": false, + "reason": "Both are CI/test failures, but on different models with different causes." + }, + { + "left": "issue:42890", + "right": "issue:45600", + "accept": false, + "reason": "Different CI regressions: missing seeds in tests vs removed config references in auto_mappings." + }, + { + "left": "issue:43502", + "right": "issue:44062", + "accept": false, + "reason": "local_files_only network leakage and AddedToken kwarg duplication are different code paths." + }, + { + "left": "issue:43010", + "right": "issue:44075", + "accept": false, + "reason": "Cache update no_grad behavior is unrelated to SGD optimizer argument handling." + }, + { + "left": "issue:42890", + "right": "issue:45341", + "accept": false, + "reason": "Test seeding flake vs a separate testing_utils bug; not the same issue." + }, + { + "left": "issue:42548", + "right": "issue:44351", + "accept": false, + "reason": "Both are import errors, but for different symbols and likely different causes." + }, + { + "left": "issue:44222", + "right": "issue:45125", + "accept": false, + "reason": "FP8 save_pretrained bug and missing tensor-parallel plan are different failures." + }, + { + "left": "issue:43502", + "right": "issue:44485", + "accept": false, + "reason": "Tokenizer/network isolation bug vs GLM RoPE implementation discussion." + }, + { + "left": "issue:44355", + "right": "issue:44749", + "accept": false, + "reason": "Compiled Python file errors are unrelated to the reported slowdown when filtering data." + }, + { + "left": "issue:30990", + "right": "issue:31795", + "accept": false, + "reason": "Loading hang and forward-argument documentation confusion are not duplicates." + }, + { + "left": "issue:45125", + "right": "issue:45478", + "accept": false, + "reason": "Same model family, but one is tensor-parallel metadata and the other is a from_pretrained regression; not one concrete bug." + }, + { + "left": "issue:45125", + "right": "issue:45375", + "accept": false, + "reason": "Different missing config fields and different failure modes." + }, + { + "left": "issue:44387", + "right": "issue:44556", + "accept": false, + "reason": "Memory/OOM regression under int4 quantization is distinct from checkpoint reload compatibility after upgrading." + }, + { + "left": "issue:43010", + "right": "issue:45278", + "accept": false, + "reason": "Cache update semantics do not match the broad import-error regression report." + }, + { + "left": "issue:30333", + "right": "issue:33453", + "accept": false, + "reason": "MLflow job status reporting and tokenizer loading regression are unrelated." + }, + { + "left": "issue:43723", + "right": "issue:44556", + "accept": false, + "reason": "Tokenizer AutoTokenizer loading issue vs generic checkpoint reload breakage; different loaders." + }, + { + "left": "issue:44355", + "right": "issue:44655", + "accept": false, + "reason": "Compiled-file execution errors are unrelated to saving Pipeline objects." + }, + { + "left": "issue:44355", + "right": "issue:45071", + "accept": false, + "reason": "Runtime compiled-file issue and PretrainedConfig type-checking regression are different bugs." + }, + { + "left": "issue:44222", + "right": "issue:44230", + "accept": false, + "reason": "FP8 save_pretrained and FP8 inference support are different change requests/issues." + }, + { + "left": "issue:43827", + "right": "issue:44961", + "accept": false, + "reason": "Documentation cleanup after pipeline removals is unrelated to the placeholder issue." + }, + { + "left": "issue:43010", + "right": "issue:45656", + "accept": false, + "reason": "Cache update no_grad concern is not the same as optimizer stepping twice under deepspeed." + }, + { + "left": "issue:33453", + "right": "issue:34689", + "accept": false, + "reason": "Tokenizer loading regression and model loading regression are different concrete failures." + }, + { + "left": "issue:43299", + "right": "issue:44556", + "accept": false, + "reason": "Specific Qwen3VL MoE load breakage is not the same as generic checkpoint reload incompatibility." + }, + { + "left": "issue:42757", + "right": "issue:43576", + "accept": false, + "reason": "Dependency import error and the v5 env command breakage do not show the same code-path problem." + }, + { + "left": "issue:43576", + "right": "issue:45071", + "accept": false, + "reason": "Env command breakage and config type-checking are unrelated." + }, + { + "left": "issue:43576", + "right": "issue:43824", + "accept": false, + "reason": "Env command bug and Qwen2.5-VL import error are different failures." + }, + { + "left": "issue:44961", + "right": "issue:45468", + "accept": false, + "reason": "No meaningful relation between the two issues." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 43, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-387", + "nodes": [ + "issue:31795", + "issue:36296", + "issue:38617", + "issue:41628", + "issue:41669", + "issue:42548", + "issue:42890", + "issue:43010", + "issue:43232", + "issue:43352", + "issue:43408", + "issue:43526", + "issue:43576", + "issue:43673", + "issue:43827", + "issue:43901", + "issue:43976", + "issue:44038", + "issue:44077", + "issue:44246", + "issue:44261", + "issue:44273", + "issue:44295", + "issue:44297", + "issue:44315", + "issue:44351", + "issue:44355", + "issue:44360", + "issue:44373", + "issue:44393", + "issue:44492", + "issue:44556", + "issue:44623", + "issue:44704", + "issue:44829", + "issue:44861", + "issue:44908", + "issue:44938", + "issue:44961", + "issue:45003", + "issue:45071", + "issue:45092", + "issue:45125", + "issue:45230", + "issue:45341", + "issue:45478", + "issue:45542", + "issue:45753" + ], + "soft_pairs": [ + "issue:31795|issue:38617", + "issue:43010|issue:43976", + "issue:41628|issue:42548", + "issue:38617|issue:43576", + "issue:44908|issue:44961", + "issue:43576|issue:44938", + "issue:43010|issue:43232", + "issue:36296|issue:43901", + "issue:41628|issue:44355", + "issue:36296|issue:43526", + "issue:44261|issue:44360", + "issue:44273|issue:45003", + "issue:44623|issue:45230", + "issue:44273|issue:45542", + "issue:44393|issue:45753", + "issue:43576|issue:44556", + "issue:44393|issue:45071", + "issue:44038|issue:45125", + "issue:41669|issue:43673", + "issue:44556|issue:45092", + "issue:44373|issue:44961", + "issue:44704|issue:44908", + "issue:42890|issue:44908", + "issue:44246|issue:44273", + "issue:43408|issue:44315", + "issue:43976|issue:44393", + "issue:43352|issue:44393", + "issue:44297|issue:45230", + "issue:43576|issue:44908", + "issue:44908|issue:45071", + "issue:44295|issue:45230", + "issue:43673|issue:44492", + "issue:41669|issue:44351", + "issue:41669|issue:45341", + "issue:41669|issue:44492", + "issue:44393|issue:45478", + "issue:44315|issue:44829", + "issue:44077|issue:44861", + "issue:43827|issue:44908" + ], + "prepared_review_unit_hash": "fbab0301503245c833b30e8d235e2d597de5a6ad5a62b741b93e3a7ff3073e85", + "component_budget": { + "node_count": 387, + "item_count": 387, + "soft_pair_count": 1455, + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 39, + "serialized_chars": 38756, + "estimated_input_tokens": 9689, + "estimated_eval_tokens": 19634 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The set is mostly a grab bag of unrelated bug reports; the soft-similarity pairs are not strong enough to treat as duplicates. I rejected all soft edges and picked the most central import/performance report as the best issue representative.", + "confidence": 0.91, + "canonical_issue_reason": "Issue 44246 is the clearest broad import-performance report in this set and best represents the import/loading theme shared by several nearby issues.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 44246 has the strongest signal and widest relevance among the issues here; it is the most suitable overall issue representative.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:31795", + "right": "issue:38617", + "accept": false, + "reason": "Documentation confusion vs an ImportError for a different symbol; unrelated bugs." + }, + { + "left": "issue:43010", + "right": "issue:43976", + "accept": false, + "reason": "Cache/no_grad behavior is unrelated to Python version compatibility." + }, + { + "left": "issue:41628", + "right": "issue:42548", + "accept": false, + "reason": "Both are import errors, but for different classes/symbols and likely different causes." + }, + { + "left": "issue:38617", + "right": "issue:43576", + "accept": false, + "reason": "ImportError for layer_type_validation is unrelated to the v5 env command breakage." + }, + { + "left": "issue:44908", + "right": "issue:44961", + "accept": false, + "reason": "Scheduler kwargs bug has nothing to do with the unrelated 'racoon' issue." + }, + { + "left": "issue:43576", + "right": "issue:44938", + "accept": false, + "reason": "Broken env command and Python 3.14 load failure are different failure modes." + }, + { + "left": "issue:43010", + "right": "issue:43232", + "accept": false, + "reason": "Both touch generation internals, but one is no_grad decoration and the other is sync_gpus state update; not the same bug." + }, + { + "left": "issue:36296", + "right": "issue:43901", + "accept": false, + "reason": "Tensor parallel training bug and docs mentioning return_all_scores are unrelated." + }, + { + "left": "issue:41628", + "right": "issue:44355", + "accept": false, + "reason": "Importing AutoImageProcessor and running compiled Python files are different issues." + }, + { + "left": "issue:36296", + "right": "issue:43526", + "accept": false, + "reason": "Tensor parallel training and BeitImageProcessorFast label reduction are unrelated." + }, + { + "left": "issue:44261", + "right": "issue:44360", + "accept": false, + "reason": "Different model/math bugs with no shared concrete code path." + }, + { + "left": "issue:44273", + "right": "issue:45003", + "accept": false, + "reason": "Lazy loading is broad; sys.modules access is a separate implementation bug." + }, + { + "left": "issue:44623", + "right": "issue:45230", + "accept": false, + "reason": "Processor save_pretrained file output bug and an unspecified bug report are not the same issue." + }, + { + "left": "issue:44273", + "right": "issue:45542", + "accept": false, + "reason": "Lazy loading problems are unrelated to the TensorFlow backend error." + }, + { + "left": "issue:44393", + "right": "issue:45753", + "accept": false, + "reason": "Two different Qwen3 image-model symptoms; not enough evidence of the same root cause." + }, + { + "left": "issue:43576", + "right": "issue:44556", + "accept": false, + "reason": "Different v5 regressions: env command vs checkpoint reload compatibility." + }, + { + "left": "issue:44393", + "right": "issue:45071", + "accept": false, + "reason": "Qwen3-VL output issue and PretrainedConfig type checking are unrelated." + }, + { + "left": "issue:44038", + "right": "issue:45125", + "accept": false, + "reason": "Qwen3-VL bug and missing tensor-parallel plan for Qwen3_5Moe are different problems." + }, + { + "left": "issue:41669", + "right": "issue:43673", + "accept": false, + "reason": "Import star performance regression and missing GenerationMixin cache are different code paths." + }, + { + "left": "issue:44556", + "right": "issue:45092", + "accept": false, + "reason": "Both concern v5 checkpoint loading, but for different models and distinct failure causes." + }, + { + "left": "issue:44373", + "right": "issue:44961", + "accept": false, + "reason": "Docstring typo and unrelated placeholder issue." + }, + { + "left": "issue:44704", + "right": "issue:44908", + "accept": false, + "reason": "Passing kwargs to cached_file and scheduler kwargs are unrelated bugs." + }, + { + "left": "issue:42890", + "right": "issue:44908", + "accept": false, + "reason": "Missing set_seed in tests is unrelated to scheduler behavior." + }, + { + "left": "issue:44246", + "right": "issue:44273", + "accept": false, + "reason": "Import slowness and lazy loading malfunction are related at a high level, but not the same concrete bug." + }, + { + "left": "issue:43408", + "right": "issue:44315", + "accept": false, + "reason": "Model-type warning and Liger Kernel model_init behavior are unrelated." + }, + { + "left": "issue:43976", + "right": "issue:44393", + "accept": false, + "reason": "Python support issue and Qwen3-VL visual output bug are unrelated." + }, + { + "left": "issue:43352", + "right": "issue:44393", + "accept": false, + "reason": "Flash Attention support warning is unrelated to Qwen3-VL hallucination/output errors." + }, + { + "left": "issue:44297", + "right": "issue:45230", + "accept": false, + "reason": "Tokenizer class mismatch on save is not the same as a generic bug report." + }, + { + "left": "issue:43576", + "right": "issue:44908", + "accept": false, + "reason": "CLI env breakage and scheduler kwargs handling are unrelated." + }, + { + "left": "issue:44908", + "right": "issue:45071", + "accept": false, + "reason": "Scheduler kwargs bug and PretrainedConfig type-checking regression are unrelated." + }, + { + "left": "issue:44295", + "right": "issue:45230", + "accept": false, + "reason": "position_ids buffer access error does not match an unspecified bug report." + }, + { + "left": "issue:43673", + "right": "issue:44492", + "accept": false, + "reason": "Missing cache during generation and a typo in cache strategies are different issues." + }, + { + "left": "issue:41669", + "right": "issue:44351", + "accept": false, + "reason": "Import-star performance problem and missing HybridCache import are different symptoms with different fixes." + }, + { + "left": "issue:41669", + "right": "issue:45341", + "accept": false, + "reason": "Import performance regression and a testing_utils bug are unrelated." + }, + { + "left": "issue:41669", + "right": "issue:44492", + "accept": false, + "reason": "Import-star slowdown and a cache-strategy typo are unrelated." + }, + { + "left": "issue:44393", + "right": "issue:45478", + "accept": false, + "reason": "Different Qwen3.5/Qwen3-VL issues with no shared concrete failure path." + }, + { + "left": "issue:44315", + "right": "issue:44829", + "accept": false, + "reason": "Liger Kernel application in model_init and flash_attention_3 training degeneration are unrelated." + }, + { + "left": "issue:44077", + "right": "issue:44861", + "accept": false, + "reason": "Optional post_init policy and tied-weight-key AttributeError are different model-internals bugs." + }, + { + "left": "issue:43827", + "right": "issue:44908", + "accept": false, + "reason": "Docs still mentioning pipeline() and scheduler kwargs handling are unrelated." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 16, + "review_unit_index": 44, + "review_unit_count": 44, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-387", + "nodes": [ + "issue:30333", + "issue:31515", + "issue:31795", + "issue:33453", + "issue:34689", + "issue:36296", + "issue:37428", + "issue:38617", + "issue:41669", + "issue:42371", + "issue:42913", + "issue:43066", + "issue:43298", + "issue:43502", + "issue:43519", + "issue:43525", + "issue:43572", + "issue:43576", + "issue:43606", + "issue:43704", + "issue:43827", + "issue:44230", + "issue:44261", + "issue:44263", + "issue:44273", + "issue:44315", + "issue:44360", + "issue:44485", + "issue:44492", + "issue:44623", + "issue:44655", + "issue:44861", + "issue:44908", + "issue:44998", + "issue:45092", + "issue:45230", + "issue:45290", + "issue:45310", + "issue:45341", + "issue:45468" + ], + "soft_pairs": [ + "issue:31795|issue:33453", + "issue:41669|issue:44908", + "issue:42371|issue:44492", + "issue:44655|issue:44861", + "issue:43576|issue:44273", + "issue:44360|issue:45468", + "issue:45230|issue:45310", + "issue:43827|issue:44492", + "issue:44273|issue:44908", + "issue:43519|issue:44485", + "issue:41669|issue:43704", + "issue:43519|issue:44908", + "issue:31795|issue:34689", + "issue:31795|issue:37428", + "issue:44315|issue:45092", + "issue:43704|issue:44485", + "issue:44485|issue:45468", + "issue:42371|issue:43704", + "issue:43502|issue:43519", + "issue:43519|issue:44623", + "issue:43519|issue:43704", + "issue:44263|issue:45468", + "issue:43606|issue:45468", + "issue:44261|issue:45468", + "issue:43704|issue:45341", + "issue:42913|issue:43066", + "issue:44908|issue:44998", + "issue:44861|issue:45290", + "issue:43525|issue:44861", + "issue:43704|issue:44998", + "issue:41669|issue:44998", + "issue:43572|issue:44861", + "issue:43066|issue:44998", + "issue:31515|issue:38617", + "issue:44360|issue:44998", + "issue:44230|issue:44623", + "issue:30333|issue:31515", + "issue:31515|issue:36296", + "issue:31515|issue:31795", + "issue:44230|issue:45310", + "issue:31515|issue:33453", + "issue:43298|issue:44230" + ], + "prepared_review_unit_hash": "568487fc00857f754cbd0b5a58bc54212cfdd0bd2cd4a7021b536d64b79ece3a", + "component_budget": { + "node_count": 387, + "item_count": 387, + "soft_pair_count": 1455, + "serialized_chars": 749544, + "estimated_input_tokens": 187386, + "estimated_eval_tokens": 375028 + }, + "budget": { + "node_count": 40, + "item_count": 40, + "soft_pair_count": 42, + "serialized_chars": 34970, + "estimated_input_tokens": 8743, + "estimated_eval_tokens": 17742 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "No soft-edge pair looks like the same underlying bug/change; they mostly share only broad subsystem words. No PRs were provided, so issue triage is the only meaningful choice here.", + "confidence": 0.91, + "canonical_issue_reason": "issue:43502 is the clearest, most concrete user-facing bug: `local_files_only=True` still triggering API requests. It is narrowly scoped, reproducible, and broadly useful as the representative issue.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43502 is the best overall candidate because it describes a specific, high-signal loading/networking regression with an unambiguous expected behavior.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:31795", + "right": "issue:33453", + "accept": false, + "reason": "Different problems: documentation confusion for forward inputs vs a tokenizer loading regression." + }, + { + "left": "issue:41669", + "right": "issue:44908", + "accept": false, + "reason": "Unrelated topics: import-star performance overhead vs scheduler kwargs being ignored." + }, + { + "left": "issue:42371", + "right": "issue:44492", + "accept": false, + "reason": "Different bugs: TF32 API guidance vs a typo in cache strategies." + }, + { + "left": "issue:44655", + "right": "issue:44861", + "accept": false, + "reason": "Both involve saving/model internals, but one is pipeline save_pretrained failure and the other is a tied-weights helper AttributeError." + }, + { + "left": "issue:43576", + "right": "issue:44273", + "accept": false, + "reason": "Different features: v5 `transformers env` command breakage vs lazy loading malfunction." + }, + { + "left": "issue:44360", + "right": "issue:45468", + "accept": false, + "reason": "Distinct code paths and model components; not the same bug." + }, + { + "left": "issue:45230", + "right": "issue:45310", + "accept": false, + "reason": "Generic bug report vs a specific Qwen3.5 MoE from_pretrained failure; too vague to be the same issue." + }, + { + "left": "issue:43827", + "right": "issue:44492", + "accept": false, + "reason": "Docs still mentioning removed pipeline APIs is unrelated to a cache strategy typo." + }, + { + "left": "issue:44273", + "right": "issue:44908", + "accept": false, + "reason": "Lazy loading behavior and inverse_sqrt scheduler kwargs are unrelated." + }, + { + "left": "issue:43519", + "right": "issue:44485", + "accept": false, + "reason": "Different model-specific bugs: Qwen3VL timestamp math vs GLM-5 RoPE implementation." + }, + { + "left": "issue:41669", + "right": "issue:43704", + "accept": false, + "reason": "Import-star performance issue is unrelated to VRAM leakage in multi-threaded dataloading." + }, + { + "left": "issue:43519", + "right": "issue:44908", + "accept": false, + "reason": "Different subsystems: processor timestamp calculation vs learning-rate scheduler config handling." + }, + { + "left": "issue:31795", + "right": "issue:34689", + "accept": false, + "reason": "Documentation wording issue vs model-loading regression for Llama 3.2 Vision." + }, + { + "left": "issue:31795", + "right": "issue:37428", + "accept": false, + "reason": "Forward-argument docs confusion is unrelated to a flash-attention import error." + }, + { + "left": "issue:44315", + "right": "issue:45092", + "accept": false, + "reason": "Liger Kernel application in model_init is not the same as remote-code checkpoint incompatibility with meta init." + }, + { + "left": "issue:43704", + "right": "issue:44485", + "accept": false, + "reason": "Different model/runtime bugs: VRAM leak vs RoPE implementation." + }, + { + "left": "issue:44485", + "right": "issue:45468", + "accept": false, + "reason": "Separate model components and failures; not a single concrete bug." + }, + { + "left": "issue:42371", + "right": "issue:43704", + "accept": false, + "reason": "TF32 configuration guidance is unrelated to the dataloader-thread VRAM leak." + }, + { + "left": "issue:43502", + "right": "issue:43519", + "accept": false, + "reason": "Network access during loading vs Qwen3VL timestamp calculation are different bugs." + }, + { + "left": "issue:43519", + "right": "issue:44623", + "accept": false, + "reason": "Processor timestamp math and missing files in processor.save_pretrained are distinct issues." + }, + { + "left": "issue:43519", + "right": "issue:43704", + "accept": false, + "reason": "Different areas: processor timestamp computation vs VRAM leakage in threaded loading." + }, + { + "left": "issue:44263", + "right": "issue:45468", + "accept": false, + "reason": "torch.split return handling in a DSA indexer is unrelated to Gemma4 audio positional encoding." + }, + { + "left": "issue:43606", + "right": "issue:45468", + "accept": false, + "reason": "CPU offload device mismatch has no clear relation to Gemma4 audio positional encoding." + }, + { + "left": "issue:44261", + "right": "issue:45468", + "accept": false, + "reason": "Missing rms_norm_eps in MLA q_a_layernorm is a different numerical bug than Gemma4 audio positional encoding." + }, + { + "left": "issue:43704", + "right": "issue:45341", + "accept": false, + "reason": "VRAM leak in multi-threaded dataloading is unrelated to a testing_utils bug." + }, + { + "left": "issue:42913", + "right": "issue:43066", + "accept": false, + "reason": "Both are tokenizer/v5 related, but one is a broad behavior difference and the other is a specific wrong decoder type bug." + }, + { + "left": "issue:44908", + "right": "issue:44998", + "accept": false, + "reason": "Scheduler kwargs bug is unrelated to the unrelated 'Unemployment' issue." + }, + { + "left": "issue:44861", + "right": "issue:45290", + "accept": false, + "reason": "Tied-weights helper crash is unrelated to chat template tool-call handling." + }, + { + "left": "issue:43525", + "right": "issue:44861", + "accept": false, + "reason": "Missing pad_token_id in Llama4Config is a different config bug than tied-weights handling." + }, + { + "left": "issue:43704", + "right": "issue:44998", + "accept": false, + "reason": "VRAM leak bug and the unrelated issue titled 'Unemployment' are clearly not the same." + }, + { + "left": "issue:41669", + "right": "issue:44998", + "accept": false, + "reason": "Import-star performance issue is unrelated to the unrelated issue titled 'Unemployment'." + }, + { + "left": "issue:43572", + "right": "issue:44861", + "accept": false, + "reason": "StableLmConfig pad_token_idx removal issue is unrelated to tied-weights AttributeError." + }, + { + "left": "issue:43066", + "right": "issue:44998", + "accept": false, + "reason": "Wrong tokenizer decoder type is unrelated to the unrelated issue titled 'Unemployment'." + }, + { + "left": "issue:31515", + "right": "issue:38617", + "accept": false, + "reason": "Slow checkpoint loading and an ImportError from configuration_utils are different failure modes." + }, + { + "left": "issue:44360", + "right": "issue:44998", + "accept": false, + "reason": "DSA indexer functionality issue is unrelated to the unrelated issue titled 'Unemployment'." + }, + { + "left": "issue:44230", + "right": "issue:44623", + "accept": false, + "reason": "FP8 inference support for Qwen models is unrelated to processor.save_pretrained missing files." + }, + { + "left": "issue:30333", + "right": "issue:31515", + "accept": false, + "reason": "MLflow reporting on failed training jobs is unrelated to slow from_pretrained checkpoint loading." + }, + { + "left": "issue:31515", + "right": "issue:36296", + "accept": false, + "reason": "Checkpoint loading performance is unrelated to a tensor parallel training bug." + }, + { + "left": "issue:31515", + "right": "issue:31795", + "accept": false, + "reason": "Slow checkpoint loading and forward-argument documentation confusion are unrelated." + }, + { + "left": "issue:44230", + "right": "issue:45310", + "accept": false, + "reason": "FP8 support request and a Qwen3.5 MoE from_pretrained error are different issues." + }, + { + "left": "issue:31515", + "right": "issue:33453", + "accept": false, + "reason": "Checkpoint loading slowness is not the same as the tokenizer loading regression." + }, + { + "left": "issue:43298", + "right": "issue:44230", + "accept": false, + "reason": "Continuous batching audio-model support and FP8 inference support are different feature gaps, not the same bug." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 3, + "component_count": 16, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-26350-2", + "nodes": [ + "issue:26350", + "issue:33260" + ], + "soft_pairs": [ + "issue:26350|issue:33260" + ], + "prepared_review_unit_hash": "010abed66cc8c1d5a28f2876a9b7a7d8dc8445c25b6616c5d45cf08051f29b41", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1589, + "estimated_input_tokens": 398, + "estimated_eval_tokens": 1052 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1589, + "estimated_input_tokens": 398, + "estimated_eval_tokens": 1052 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are two unrelated community contribution issues: one requests Flash Attention 2 support for more architectures, the other requests GGUF support for more architectures. They do not describe the same underlying bug or change.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": "No canonical issue exists here because the requests target different features; if one had to be picked as a representative issue, #26350 is the older and more discussed feature request.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:26350", + "right": "issue:33260", + "accept": false, + "reason": "Different feature requests with no shared code path or explicit linkage: Flash Attention 2 support vs GGUF support." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 4, + "component_count": 16, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-27088-4", + "nodes": [ + "issue:27088", + "issue:43834", + "issue:44365", + "issue:44370" + ], + "soft_pairs": [ + "issue:27088|issue:44370", + "issue:27088|issue:44365", + "issue:27088|issue:43834", + "issue:43834|issue:44370", + "issue:43834|issue:44365", + "issue:44365|issue:44370" + ], + "prepared_review_unit_hash": "2f2d8e8a2ec10bccf707ea15462303fc48908d959146f7e0b89e4ca9dd1da267", + "component_budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 6, + "serialized_chars": 4454, + "estimated_input_tokens": 1114, + "estimated_eval_tokens": 2484 + }, + "budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 6, + "serialized_chars": 4454, + "estimated_input_tokens": 1114, + "estimated_eval_tokens": 2484 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "All four items are translation-tracking issues for docs, but they are not the same underlying request: one is Turkish-specific and the others are generic placeholder issues for different languages. No PRs are present.", + "confidence": 0.92, + "canonical_issue_reason": "Issue 27088 is the only concrete, language-specific report ([i18n-TR] Turkish); the others are template-style placeholders and not a better canonical anchor.", + "canonical_pr_reason": null, + "best_issue_reason": "27088 is the most specific and actionable issue, while the others are generic duplicates/templates rather than a concrete unique request.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:27088", + "right": "issue:44370", + "accept": false, + "reason": "Same broad translation-tracking template, but not the same language/request; different underlying issue." + }, + { + "left": "issue:27088", + "right": "issue:44365", + "accept": false, + "reason": "These are different translation targets, so they are not duplicates of the same bug/request." + }, + { + "left": "issue:27088", + "right": "issue:43834", + "accept": false, + "reason": "Turkish-specific issue vs a generic language placeholder; related topic but not the same underlying issue." + }, + { + "left": "issue:43834", + "right": "issue:44370", + "accept": false, + "reason": "Both are generic translation placeholders, but they do not identify the same concrete language or request." + }, + { + "left": "issue:43834", + "right": "issue:44365", + "accept": false, + "reason": "Template similarity only; the issues are for different language translations, not one shared request." + }, + { + "left": "issue:44365", + "right": "issue:44370", + "accept": false, + "reason": "They share the same template wording, but no evidence they refer to the same translation target or change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 5, + "component_count": 16, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-36817-14", + "nodes": [ + "issue:36817", + "issue:37015", + "issue:41115", + "issue:41211", + "issue:42503", + "issue:42738", + "issue:43671", + "issue:43837", + "issue:43909", + "issue:44121", + "issue:44205", + "issue:45027", + "issue:45177", + "issue:45322" + ], + "soft_pairs": [ + "issue:36817|issue:41211", + "issue:36817|issue:43671", + "issue:37015|issue:43837", + "issue:37015|issue:44121", + "issue:37015|issue:43909", + "issue:37015|issue:42503", + "issue:37015|issue:45027", + "issue:41115|issue:43837", + "issue:41115|issue:44121", + "issue:41115|issue:42503", + "issue:41115|issue:45027", + "issue:41211|issue:43671", + "issue:42503|issue:43837", + "issue:42503|issue:43671", + "issue:42503|issue:42738", + "issue:42738|issue:45027", + "issue:43671|issue:43837", + "issue:43671|issue:45027", + "issue:43671|issue:45177", + "issue:43671|issue:44121", + "issue:43837|issue:45322", + "issue:43837|issue:43909", + "issue:43909|issue:44205", + "issue:43671|issue:43909", + "issue:44121|issue:45322", + "issue:43837|issue:44121", + "issue:43837|issue:45027", + "issue:43837|issue:45177" + ], + "prepared_review_unit_hash": "7ab19ac4bf0378c06c5dae49fbf63d38d86575ae7e6d809da951d59571ca8159", + "component_budget": { + "node_count": 14, + "item_count": 14, + "soft_pair_count": 28, + "serialized_chars": 17650, + "estimated_input_tokens": 4413, + "estimated_eval_tokens": 9082 + }, + "budget": { + "node_count": 14, + "item_count": 14, + "soft_pair_count": 28, + "serialized_chars": 17650, + "estimated_input_tokens": 4413, + "estimated_eval_tokens": 9082 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "This cluster is a set of separate model-addition requests, not duplicate reports of the same bug/change. The soft edges are mostly name-similarity across different model families or modalities (text, audio, vision), so none should be merged.", + "confidence": 0.97, + "canonical_issue_reason": "No single canonical issue fits the cluster: these are distinct model requests for different architectures and modalities, not one underlying defect or feature.", + "canonical_pr_reason": "No PRs are present in this cluster.", + "best_issue_reason": "No single issue is a good global representative; any choice would omit unrelated requests in other model families/modalities.", + "best_pr_reason": "No PRs are present in this cluster.", + "soft_edge_verdicts": [ + { + "left": "issue:36817", + "right": "issue:41211", + "accept": false, + "reason": "EuroBert config support and DEIMv2 are unrelated model additions." + }, + { + "left": "issue:36817", + "right": "issue:43671", + "accept": false, + "reason": "EuroBert is a text-model config request; Qwen3-TTS is an audio generation request." + }, + { + "left": "issue:37015", + "right": "issue:43837", + "accept": false, + "reason": "NeoBERT and Qwen3-ASR are different model families and different tasks." + }, + { + "left": "issue:37015", + "right": "issue:44121", + "accept": false, + "reason": "NeoBERT and circuit-sparsity/OpenAI weight-sparse transformer are unrelated model requests." + }, + { + "left": "issue:37015", + "right": "issue:43909", + "accept": false, + "reason": "NeoBERT is a text model; LFM2.5 Audio is an audio model." + }, + { + "left": "issue:37015", + "right": "issue:42503", + "accept": false, + "reason": "NeoBERT and ModernVBERT are different model architectures, not the same change." + }, + { + "left": "issue:37015", + "right": "issue:45027", + "accept": false, + "reason": "NeoBERT and Voxtral TTS support are unrelated model additions." + }, + { + "left": "issue:41115", + "right": "issue:43837", + "accept": false, + "reason": "MiniCPM3 architecture support and Qwen3-ASR support are different concrete changes." + }, + { + "left": "issue:41115", + "right": "issue:44121", + "accept": false, + "reason": "MiniCPM3 and the weight-sparse transformer request are not the same model or code path." + }, + { + "left": "issue:41115", + "right": "issue:42503", + "accept": false, + "reason": "MiniCPM3 and ModernVBERT are separate model additions." + }, + { + "left": "issue:41115", + "right": "issue:45027", + "accept": false, + "reason": "MiniCPM3 support and Voxtral TTS support are distinct requests." + }, + { + "left": "issue:41211", + "right": "issue:43671", + "accept": false, + "reason": "DEIMv2 and Qwen3-TTS are unrelated model families with different implementations." + }, + { + "left": "issue:42503", + "right": "issue:43837", + "accept": false, + "reason": "ModernVBERT and Qwen3-ASR are different models and different tasks." + }, + { + "left": "issue:42503", + "right": "issue:43671", + "accept": false, + "reason": "ModernVBERT and Qwen3-TTS are unrelated additions." + }, + { + "left": "issue:42503", + "right": "issue:42738", + "accept": false, + "reason": "ModernVBERT and BERT-like RoPE support are not the same concrete model change." + }, + { + "left": "issue:42738", + "right": "issue:45027", + "accept": false, + "reason": "BERT-like RoPE support and Voxtral TTS support address different model classes." + }, + { + "left": "issue:43671", + "right": "issue:43837", + "accept": false, + "reason": "TTS and ASR are related audio domains but separate model requests and code paths." + }, + { + "left": "issue:43671", + "right": "issue:45027", + "accept": false, + "reason": "Both are TTS-related, but they request different models and likely different integrations." + }, + { + "left": "issue:43671", + "right": "issue:45177", + "accept": false, + "reason": "Qwen3-TTS and DeepSeek-OCR2 are different modalities and unrelated changes." + }, + { + "left": "issue:43837", + "right": "issue:45322", + "accept": false, + "reason": "ASR support and EUPE (a perception encoder) are not the same model addition." + }, + { + "left": "issue:43837", + "right": "issue:43909", + "accept": false, + "reason": "Qwen3-ASR and LFM2.5 Audio are different audio models with different support needs." + }, + { + "left": "issue:43909", + "right": "issue:44205", + "accept": false, + "reason": "LFM2.5 Audio and SAM3-LiteText are different model requests with no shared underlying bug." + }, + { + "left": "issue:43671", + "right": "issue:43909", + "accept": false, + "reason": "Qwen3-TTS and LFM2.5 Audio are separate audio-model integrations." + }, + { + "left": "issue:44121", + "right": "issue:45322", + "accept": false, + "reason": "Weight-sparse transformer support and EUPE are unrelated model requests." + }, + { + "left": "issue:43837", + "right": "issue:44121", + "accept": false, + "reason": "Qwen3-ASR and the weight-sparse transformer request are different concrete code changes." + }, + { + "left": "issue:43837", + "right": "issue:45027", + "accept": false, + "reason": "Qwen3-ASR and Voxtral TTS are different audio tasks and integrations." + }, + { + "left": "issue:43837", + "right": "issue:45177", + "accept": false, + "reason": "ASR support and DeepSeek-OCR2 are different modalities and model families." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 6, + "component_count": 16, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-37934-6", + "nodes": [ + "issue:37934", + "issue:42936", + "issue:43240", + "issue:43668", + "issue:43813", + "issue:44031" + ], + "soft_pairs": [ + "issue:37934|issue:43668", + "issue:42936|issue:43240", + "issue:42936|issue:44031", + "issue:42936|issue:43813", + "issue:42936|issue:43668", + "issue:43240|issue:43668", + "issue:43668|issue:44031", + "issue:43668|issue:43813" + ], + "prepared_review_unit_hash": "3a21da2f46fbef8e76b79c47ec7b9c8b2646d1585ebe7dc1bf347aeffa8541a8", + "component_budget": { + "node_count": 6, + "item_count": 6, + "soft_pair_count": 8, + "serialized_chars": 6291, + "estimated_input_tokens": 1573, + "estimated_eval_tokens": 3402 + }, + "budget": { + "node_count": 6, + "item_count": 6, + "soft_pair_count": 8, + "serialized_chars": 6291, + "estimated_input_tokens": 1573, + "estimated_eval_tokens": 3402 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These issues are not duplicates; they span unrelated bugs across different models, config typing, tokenizer warnings, and a typo. All soft edges should be rejected.", + "confidence": 0.97, + "canonical_issue_reason": "No clear canonical issue: the cluster is heterogeneous, with unrelated topics such as normalization semantics, loss-plumbing, config type hints, a typo, and tokenizer warnings.", + "canonical_pr_reason": null, + "best_issue_reason": "No issue is a strong cluster representative; 44031 is the broadest title, but it is still unrelated to the other reports.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:37934", + "right": "issue:43668", + "accept": false, + "reason": "Both mention normalization-related terms, but one is about Llama4TextL2Norm semantics while the other is a ModernBERTConfig type hint; different bug classes." + }, + { + "left": "issue:42936", + "right": "issue:43240", + "accept": false, + "reason": "Mask2former ignore_value handling and loss kwargs plumbing are different code paths and different failure modes." + }, + { + "left": "issue:42936", + "right": "issue:44031", + "accept": false, + "reason": "A model loss-argument bug and a tokenizer regex warning after 4.57.3 are unrelated." + }, + { + "left": "issue:42936", + "right": "issue:43813", + "accept": false, + "reason": "Ignore_value usage in Mask2former is not the same underlying issue as a quantization-attribute typo." + }, + { + "left": "issue:42936", + "right": "issue:43668", + "accept": false, + "reason": "One is a functional bug in Mask2former loss handling; the other is an incorrect type hint in a config class." + }, + { + "left": "issue:43240", + "right": "issue:43668", + "accept": false, + "reason": "Passing kwargs into a loss function and fixing a config type hint are unrelated changes." + }, + { + "left": "issue:43668", + "right": "issue:44031", + "accept": false, + "reason": "A ModernBERTConfig annotation error and a tokenizer regex warning are different subsystems and different fixes." + }, + { + "left": "issue:43668", + "right": "issue:43813", + "accept": false, + "reason": "Incorrect type hints and a typo in orig_conversion.quantization_operation are distinct issues." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 7, + "component_count": 16, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-41084-2", + "nodes": [ + "issue:41084", + "issue:43037" + ], + "soft_pairs": [ + "issue:41084|issue:43037" + ], + "prepared_review_unit_hash": "4be7d4c65e5d2f7b80f1a57a96bd17fa29b8d56328e91c23d5ffb197dbee0c76", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1480, + "estimated_input_tokens": 370, + "estimated_eval_tokens": 996 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1480, + "estimated_input_tokens": 370, + "estimated_eval_tokens": 996 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These two issues are not duplicates: one is a low-level block decoding request, while the other is a model-specific DeepSeek v3.2 support request.", + "confidence": 0.97, + "canonical_issue_reason": "Neither issue is a clear canonical duplicate target for the other; they describe different problem types and different expected changes.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 43037 is the better standalone tracking issue because it is more specific and actionable than the vague 'Set Block Decoding' request.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:41084", + "right": "issue:43037", + "accept": false, + "reason": "The titles and likely scopes differ substantially: one concerns block decoding mechanics, the other concerns adding support for a specific model version. They do not appear to describe the same underlying bug or change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 8, + "component_count": 16, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43598-2", + "nodes": [ + "issue:43598", + "issue:43600" + ], + "soft_pairs": [ + "issue:43598|issue:43600" + ], + "prepared_review_unit_hash": "df977e7412363678a5c4d3a2f4b760f1e1c75ee16ed995b979db0031df6fc483", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1584, + "estimated_input_tokens": 396, + "estimated_eval_tokens": 1048 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1584, + "estimated_input_tokens": 396, + "estimated_eval_tokens": 1048 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are related to Trainer customization, but they describe different problems: one is about `remove_unused_column` being too rigid, the other is about FLOPS estimation for custom trainers. They are not duplicates.", + "confidence": 0.91, + "canonical_issue_reason": "Issue 43598 is the broader, more central Trainer customizability request; 43600 is a separate FLOPS-estimation edge case.", + "canonical_pr_reason": null, + "best_issue_reason": "43598 is the better representative issue for this cluster because it captures the broader customization concern and is less narrowly scoped than 43600.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43598", + "right": "issue:43600", + "accept": false, + "reason": "Different underlying bugs: one changes how Trainer drops unused columns, the other concerns FLOPS accounting for custom trainers. Shared theme only, not the same fix." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 9, + "component_count": 16, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43684-2", + "nodes": [ + "issue:43684", + "issue:44405" + ], + "soft_pairs": [ + "issue:43684|issue:44405" + ], + "prepared_review_unit_hash": "45a4006c06bab89496ef3cc59191cc57f9b29428dee00755a44cf3ea26366afa", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1612, + "estimated_input_tokens": 403, + "estimated_eval_tokens": 1062 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1612, + "estimated_input_tokens": 403, + "estimated_eval_tokens": 1062 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The two issues are about different Qwen variants and different AutoModel heads, so they are not duplicates and should not be merged.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": "Issue 43684 is the broader registration request, but it does not represent the same bug/feature as 44405; there is no true canonical issue here.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43684", + "right": "issue:44405", + "accept": false, + "reason": "Different model families and different heads: Qwen3-Omni registration for AutoModel/AutoModelForConditionalGeneration vs Qwen3.5 sequence-classification support. Not the same underlying change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 10, + "component_count": 16, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43835-4", + "nodes": [ + "issue:43835", + "issue:43845", + "issue:43846", + "issue:43859" + ], + "soft_pairs": [ + "issue:43835|issue:43845", + "issue:43835|issue:43846", + "issue:43835|issue:43859", + "issue:43845|issue:43846", + "issue:43845|issue:43859", + "issue:43846|issue:43859" + ], + "prepared_review_unit_hash": "8e38dbeff14c4affdcb488e1c9d23498e08297ab0fd021c4e3b8c3c5041653ea", + "component_budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 6, + "serialized_chars": 4265, + "estimated_input_tokens": 1067, + "estimated_eval_tokens": 2390 + }, + "budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 6, + "serialized_chars": 4265, + "estimated_input_tokens": 1067, + "estimated_eval_tokens": 2390 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Three closed issues share the exact title \"huggingface \" and look like duplicates; the \"Fett\" issue does not look related from the available metadata.", + "confidence": 0.77, + "canonical_issue_reason": "Issue 43845 is the earliest of the identical-title cluster \"huggingface \" and is a clean representative for the duplicate set.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 43845 is the best overall issue candidate because it is part of the clear duplicate cluster and was created first; 43859 has slightly more discussion but no stronger signal of being distinct.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43835", + "right": "issue:43845", + "accept": false, + "reason": "Different title and no shared targets/files or body detail to indicate the same underlying issue." + }, + { + "left": "issue:43835", + "right": "issue:43846", + "accept": false, + "reason": "Different title and no concrete evidence that this is the same bug report." + }, + { + "left": "issue:43835", + "right": "issue:43859", + "accept": false, + "reason": "Different title with no supporting metadata tying it to the huggingface duplicates." + }, + { + "left": "issue:43845", + "right": "issue:43846", + "accept": true, + "reason": "Exact same title, very close creation times, and no evidence of distinct scope." + }, + { + "left": "issue:43845", + "right": "issue:43859", + "accept": true, + "reason": "Exact same title and strong duplicate signal from the repeated report pattern." + }, + { + "left": "issue:43846", + "right": "issue:43859", + "accept": true, + "reason": "Exact same title and same apparent issue report as the other huggingface entries." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet: 43845/43846/43859 all share the exact title \"huggingface \", while 43835 is clearly different. The issue-order and canonical choice are also supported by the timestamps. The soft-edge accepts are based on limited metadata, but they do not overstate beyond title/timing similarity." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 11, + "component_count": 16, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-44371-5", + "nodes": [ + "issue:44371", + "issue:44418", + "issue:44572", + "issue:44573", + "issue:44937" + ], + "soft_pairs": [ + "issue:44371|issue:44572", + "issue:44371|issue:44573", + "issue:44371|issue:44418", + "issue:44371|issue:44937", + "issue:44418|issue:44572", + "issue:44418|issue:44573", + "issue:44572|issue:44573", + "issue:44572|issue:44937", + "issue:44573|issue:44937" + ], + "prepared_review_unit_hash": "fa113db7967fa440bbc1575ba422726461f06ac3a2c10b53d727cf97d195ba35", + "component_budget": { + "node_count": 5, + "item_count": 5, + "soft_pair_count": 9, + "serialized_chars": 5794, + "estimated_input_tokens": 1449, + "estimated_eval_tokens": 3154 + }, + "budget": { + "node_count": 5, + "item_count": 5, + "soft_pair_count": 9, + "serialized_chars": 5794, + "estimated_input_tokens": 1449, + "estimated_eval_tokens": 3154 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The cluster appears to be a set of unrelated low-quality/spam issues with superficial title similarity. No pair has enough substantive content to justify duplicate merging as the same underlying bug/change.", + "confidence": 0.97, + "canonical_issue_reason": "Issue 44371 is the earliest of the clearly spam/placeholder issues and is the best available anchor for this cluster, though it is not a meaningful product bug report.", + "canonical_pr_reason": null, + "best_issue_reason": "44371 is the most suitable issue anchor because it is an early exact-match placeholder spam item; the rest are similarly non-substantive and do not provide a better canonical signal.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44371", + "right": "issue:44572", + "accept": false, + "reason": "Both are just '' placeholders; there is no concrete bug or change description to establish the same underlying issue." + }, + { + "left": "issue:44371", + "right": "issue:44573", + "accept": false, + "reason": "Both are just '' placeholders; identical titles alone are not enough to confirm a real duplicate bug/change." + }, + { + "left": "issue:44371", + "right": "issue:44418", + "accept": false, + "reason": "One is spam and the other is a generic documentation suggestion; the overlap is only superficial wording, not the same underlying issue." + }, + { + "left": "issue:44371", + "right": "issue:44937", + "accept": false, + "reason": "The titles are unrelated spam-like phrases and do not indicate the same concrete bug or request." + }, + { + "left": "issue:44418", + "right": "issue:44572", + "accept": false, + "reason": "Generic docs-suggestion wording does not match a placeholder spam report well enough to call them the same issue." + }, + { + "left": "issue:44418", + "right": "issue:44573", + "accept": false, + "reason": "Generic docs-suggestion wording does not match a placeholder spam report well enough to call them the same issue." + }, + { + "left": "issue:44572", + "right": "issue:44573", + "accept": false, + "reason": "Although both are '', there is no substantive report content to confirm they describe the same underlying bug/change." + }, + { + "left": "issue:44572", + "right": "issue:44937", + "accept": false, + "reason": "A placeholder spam title and a specific promotional title are not evidence of the same underlying issue." + }, + { + "left": "issue:44573", + "right": "issue:44937", + "accept": false, + "reason": "A placeholder spam title and a specific promotional title are not evidence of the same underlying issue." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 12, + "component_count": 16, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45145-2", + "nodes": [ + "issue:45145", + "issue:45151" + ], + "soft_pairs": [ + "issue:45145|issue:45151" + ], + "prepared_review_unit_hash": "c87c66c34e215718d3d6c1e4eacb20c2cfc1241fdbc8fdd809ce4a9ff21272bf", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1646, + "estimated_input_tokens": 412, + "estimated_eval_tokens": 1080 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1646, + "estimated_input_tokens": 412, + "estimated_eval_tokens": 1080 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These two closed issues are strong near-duplicates: they share the same title, have substantial soft similarity, and #45145 is earlier with an inbound reference, making it the best canonical representative.", + "confidence": 0.92, + "canonical_issue_reason": "#45145 is the better canonical issue because it was created first and already has an inbound reference, while #45151 appears to be a later duplicate with no stronger distinguishing signal.", + "canonical_pr_reason": null, + "best_issue_reason": "#45145 is the strongest issue for the cluster: it is the earliest instance, has the same title as #45151, and has the only inbound reference activity.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45145", + "right": "issue:45151", + "accept": true, + "reason": "Accept as a near-duplicate: the titles match exactly and the soft-similarity signal is strong, with #45145 also predating #45151 and carrying the only inbound reference. We do not have body text here, so the judgment is based on title and available similarity signals rather than body-level evidence." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "The overall duplicate/near-duplicate conclusion is plausible, but the soft-edge rationale overstates what is in the packet: we do not have the issue bodies, so \"highly similar bodies\" is not directly grounded, and \"same closed issue type\" is not a meaningful distinction here. The verdict should be phrased more conservatively based on same title plus soft similarity evidence and earlier timestamp/inbound reference, without claiming stronger body-level evidence than provided." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 13, + "component_count": 16, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45295-2", + "nodes": [ + "issue:45295", + "issue:45373" + ], + "soft_pairs": [], + "prepared_review_unit_hash": "f5fd95fada2f530ab0999e343c17c16266cde0c7182bb22509972ca44827cd49", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 0, + "serialized_chars": 1286, + "estimated_input_tokens": 322, + "estimated_eval_tokens": 900 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 0, + "serialized_chars": 1286, + "estimated_input_tokens": 322, + "estimated_eval_tokens": 900 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These two issues describe the same missing Gemma 4 sequence-classification support; the later, more specific open issue is the better cluster representative.", + "confidence": 0.94, + "canonical_issue_reason": "issue #45373 is the clearest canonical tracker: it is open, explicitly names the missing class `Gemma4ForSequenceClassification`, and states the module-level gap relative to Gemma 2/3.", + "canonical_pr_reason": null, + "best_issue_reason": "issue #45373 best captures the concrete request and likely remaining work, while #45295 is a shorter earlier duplicate with less detail and is already closed.", + "best_pr_reason": null, + "soft_edge_verdicts": [] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 14, + "component_count": 16, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45672-2", + "nodes": [ + "issue:45672", + "issue:45674" + ], + "soft_pairs": [ + "issue:45672|issue:45674" + ], + "prepared_review_unit_hash": "19b5752beca17e4e02fb32bbef0c5c56e57bd37c447436c2a8168588232d9592", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1668, + "estimated_input_tokens": 417, + "estimated_eval_tokens": 1090 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1668, + "estimated_input_tokens": 417, + "estimated_eval_tokens": 1090 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are two different 4-bit inference bugs: one in Gemma4/audio-module handling with uint8 weights and torch.finfo(), the other in BitsAndBytesConfig skip-module handling clearing the default lm_head exclusion. They are related by quantized inference but not the same underlying defect.", + "confidence": 0.96, + "canonical_issue_reason": "No single canonical issue fits both reports; they describe separate failure modes. If forced to pick the clearer representative of its own bug, issue #45674 is the more general BitsAndBytesConfig regression report.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue #45674 is the stronger standalone issue: it names the exact configuration regression, states the observed AssertionError, and has a clearer path to reproduction/fix than the Gemma4-specific torch.finfo() error.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45672", + "right": "issue:45674", + "accept": false, + "reason": "Different bugs and code paths: Gemma4/audio uint8 torch.finfo() TypeError vs. BitsAndBytesConfig llm_int8_skip_modules clearing lm_head exclusion and triggering an AssertionError. They are not mergeable as one fix." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 15, + "component_count": 16, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45762-2", + "nodes": [ + "issue:45762", + "issue:45763" + ], + "soft_pairs": [ + "issue:45762|issue:45763" + ], + "prepared_review_unit_hash": "25fba1d63b8fab9cd7d84219496f61f159fc149e6e4640904ae244d4c18fbba1", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1477, + "estimated_input_tokens": 370, + "estimated_eval_tokens": 996 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1477, + "estimated_input_tokens": 370, + "estimated_eval_tokens": 996 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The two issues look like unrelated name-and-date entries with no shared description, references, or code context. They should not be clustered as duplicates.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45762", + "right": "issue:45763", + "accept": false, + "reason": "The titles only share a surname and date-like text; there is no evidence of the same underlying bug, request, or actionable issue." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 16, + "component_count": 16, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45764-2", + "nodes": [ + "issue:45764", + "issue:45765" + ], + "soft_pairs": [ + "issue:45764|issue:45765" + ], + "prepared_review_unit_hash": "fe409b76adca8244431eda270b257edc5801e5988bc6364f3a25952ac74ae71c", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1481, + "estimated_input_tokens": 371, + "estimated_eval_tokens": 998 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1481, + "estimated_input_tokens": 371, + "estimated_eval_tokens": 998 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The two closed issues appear to be unrelated name/date entries rather than the same bug or request, so this is not a duplicate cluster.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45764", + "right": "issue:45765", + "accept": false, + "reason": "Titles only contain different personal names and dates with no shared problem statement, code path, or explicit linkage; they do not look like duplicates." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 1, + "component_count": 26, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-39895-13", + "nodes": [ + "pull_request:39895", + "pull_request:41855", + "pull_request:42668", + "pull_request:43251", + "pull_request:43254", + "pull_request:43424", + "pull_request:43823", + "pull_request:43995", + "pull_request:44827", + "pull_request:45170", + "pull_request:45213", + "pull_request:45655", + "pull_request:45795" + ], + "soft_pairs": [ + "pull_request:39895|pull_request:43424", + "pull_request:41855|pull_request:43424", + "pull_request:41855|pull_request:43823", + "pull_request:42668|pull_request:43823", + "pull_request:43251|pull_request:43823", + "pull_request:43251|pull_request:44827", + "pull_request:43424|pull_request:44827", + "pull_request:43424|pull_request:43823", + "pull_request:43823|pull_request:45655", + "pull_request:43823|pull_request:43995", + "pull_request:43995|pull_request:45213", + "pull_request:43995|pull_request:44827", + "pull_request:44827|pull_request:45170", + "pull_request:44827|pull_request:45795", + "pull_request:44827|pull_request:45655", + "pull_request:44827|pull_request:45213", + "pull_request:43823|pull_request:45213", + "pull_request:43823|pull_request:45795", + "pull_request:43251|pull_request:43254" + ], + "prepared_review_unit_hash": "335bfacae392884c7a669f7690c251829a1be6d44468f0d7f741390a8612b531", + "component_budget": { + "node_count": 13, + "item_count": 13, + "soft_pair_count": 19, + "serialized_chars": 35199, + "estimated_input_tokens": 8800, + "estimated_eval_tokens": 17856 + }, + "budget": { + "node_count": 13, + "item_count": 13, + "soft_pair_count": 19, + "serialized_chars": 35199, + "estimated_input_tokens": 8800, + "estimated_eval_tokens": 17856 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The cluster is mostly heterogeneous and the soft similarities are superficial. The only clear duplicate-like pair is the two fixed_cross_entropy PRs, which address the same loss-wrapper bug and could plausibly be merged into one change.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 43251 is the better canonical representative: it directly fixes the reported issue, has the explicit issue link, and implements the kwargs forwarding in the loss wrapper in a complete way.", + "best_issue_reason": null, + "best_pr_reason": "PR 43251 is the best global representative because it is the clearest standalone bug fix in the cluster and the only one with an explicit issue target; 43254 is a narrower variant of the same change.", + "soft_edge_verdicts": [ + { + "left": "pull_request:39895", + "right": "pull_request:43424", + "accept": false, + "reason": "Unrelated model-addition vs export test changes; same broad ecosystem only, not the same bug/change." + }, + { + "left": "pull_request:41855", + "right": "pull_request:43424", + "accept": false, + "reason": "Tokenizer API additions and export-shape testing are unrelated changes." + }, + { + "left": "pull_request:41855", + "right": "pull_request:43823", + "accept": false, + "reason": "Different models and different workstreams; no shared concrete code-path problem." + }, + { + "left": "pull_request:42668", + "right": "pull_request:43823", + "accept": false, + "reason": "Processor robustness fixes for existing models versus adding a new model; not the same underlying change." + }, + { + "left": "pull_request:43251", + "right": "pull_request:43823", + "accept": false, + "reason": "Loss-function kwargs fix is unrelated to adding MobileLLM." + }, + { + "left": "pull_request:43251", + "right": "pull_request:44827", + "accept": false, + "reason": "Loss wrapper kwargs handling is unrelated to Mistral4 test/model changes." + }, + { + "left": "pull_request:43424", + "right": "pull_request:44827", + "accept": false, + "reason": "Executorch dynamic-shape test coverage is unrelated to Mistral4 fixes." + }, + { + "left": "pull_request:43424", + "right": "pull_request:43823", + "accept": false, + "reason": "Different functionality and code paths; only both are framework PRs." + }, + { + "left": "pull_request:43823", + "right": "pull_request:45655", + "accept": false, + "reason": "New model addition and class config resolution order fix do not share the same underlying bug/change." + }, + { + "left": "pull_request:43823", + "right": "pull_request:43995", + "accept": false, + "reason": "New model addition versus Falcon refactoring; different targets and problems." + }, + { + "left": "pull_request:43995", + "right": "pull_request:45213", + "accept": false, + "reason": "Falcon output-interface refactor and a model-creation skill are not the same change." + }, + { + "left": "pull_request:43995", + "right": "pull_request:44827", + "accept": false, + "reason": "Different models and different issues; no mergeable shared fix." + }, + { + "left": "pull_request:44827", + "right": "pull_request:45170", + "accept": false, + "reason": "Mistral4 test fixes and CLIP-style layernorm renaming are unrelated." + }, + { + "left": "pull_request:44827", + "right": "pull_request:45795", + "accept": false, + "reason": "Mistral4 test fixes and docs for audio/video processors are unrelated." + }, + { + "left": "pull_request:44827", + "right": "pull_request:45655", + "accept": false, + "reason": "Different code paths: model tests versus config-resolution behavior." + }, + { + "left": "pull_request:44827", + "right": "pull_request:45213", + "accept": false, + "reason": "Only a shared auto/modeling file appears in a broad cluster sense; the actual changes are unrelated." + }, + { + "left": "pull_request:43823", + "right": "pull_request:45213", + "accept": false, + "reason": "Adding a model and adding an internal creation skill are not the same underlying change." + }, + { + "left": "pull_request:43823", + "right": "pull_request:45795", + "accept": false, + "reason": "Model addition versus docs-only processor guide; not the same change." + }, + { + "left": "pull_request:43251", + "right": "pull_request:43254", + "accept": true, + "reason": "Both PRs patch the same fixed_cross_entropy wrapper to support passing through cross-entropy kwargs, so they address the same concrete bug in the same code path and could be merged into one PR." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet: PR 43251 and PR 43254 share the same file and explicit issue target, and the change in both is the same fixed_cross_entropy kwargs support. The other pair verdicts stay conservative by rejecting broad ecosystem similarity without claiming a shared bugfix. " + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 2, + "component_count": 26, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-40546-2", + "nodes": [ + "pull_request:40546", + "pull_request:42781" + ], + "soft_pairs": [ + "pull_request:40546|pull_request:42781" + ], + "prepared_review_unit_hash": "48e29c877547d507c890160961185154419cd433d2329b245c04ffb9b850f5bb", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 7226, + "estimated_input_tokens": 1807, + "estimated_eval_tokens": 3870 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 7226, + "estimated_input_tokens": 1807, + "estimated_eval_tokens": 3870 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These two pull requests are related but not duplicates: PR 40546 adds the base VibeVoice model integration, while PR 42781 extends the family with VibeVoice Realtime and an acoustic tokenizer. They share several VibeVoice files, but the change goals and code paths are not the same.", + "confidence": 0.94, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 40546 is the better canonical representative because it introduces the core VibeVoice implementation and broad Transformers integration; PR 42781 is a later, narrower realtime extension with additional tokenizer and generation changes.", + "best_issue_reason": null, + "best_pr_reason": "PR 40546 is the best single representative of the cluster because it is the original base VibeVoice PR and covers the main model import/integration work rather than the later realtime variant.", + "soft_edge_verdicts": [ + { + "left": "pull_request:40546", + "right": "pull_request:42781", + "accept": false, + "reason": "Reject: they overlap on shared VibeVoice scaffolding, but 42781 is a distinct realtime/tokenizer extension rather than the same underlying bug or one mergeable change set." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 3, + "component_count": 26, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-40755-2", + "nodes": [ + "pull_request:40755", + "pull_request:41224" + ], + "soft_pairs": [ + "pull_request:40755|pull_request:41224" + ], + "prepared_review_unit_hash": "a4f0dd266deaf10ea8b584c35b2e4d4bdc9f4b1b34c671751053549431a5fa2d", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4675, + "estimated_input_tokens": 1169, + "estimated_eval_tokens": 2594 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4675, + "estimated_input_tokens": 1169, + "estimated_eval_tokens": 2594 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The two pull requests are unrelated feature additions in different model families: TimesFM covariate forecasting vs DINOv3 image classification support. They do not describe the same bug or change.", + "confidence": 0.99, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "pull_request:40755", + "right": "pull_request:41224", + "accept": false, + "reason": "Different code paths and domains: one adds TimesFM covariate forecasting utilities/docs/tests, the other adds DINOv3 image classification support and related auto/model changes. They are not the same underlying change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 4, + "component_count": 26, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-41356-2", + "nodes": [ + "pull_request:41356", + "pull_request:44339" + ], + "soft_pairs": [ + "pull_request:41356|pull_request:44339" + ], + "prepared_review_unit_hash": "aead1308c167b0bae6ff11bf9358340baa5e76f816618383418b505a10a9cd01", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 6050, + "estimated_input_tokens": 1513, + "estimated_eval_tokens": 3282 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 6050, + "estimated_input_tokens": 1513, + "estimated_eval_tokens": 3282 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Two DEIMv2 pull requests target the same model family, but they are not the same change: PR 41356 is an early/basic model + image processor add, while PR 44339 is a much broader follow-up that adds loss code, auto-mappings, conversion tooling, modular files, and more integration work.", + "confidence": 0.91, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 44339 is the more complete and representative DEIMv2 implementation; PR 41356 looks like an earlier partial version of the same model bring-up.", + "best_issue_reason": null, + "best_pr_reason": "PR 44339 is the strongest single artifact for the cluster because it subsumes the basic model work and includes the broader integration needed for a full Transformers addition.", + "soft_edge_verdicts": [ + { + "left": "pull_request:41356", + "right": "pull_request:44339", + "accept": false, + "reason": "Same model area, but not the same concrete change: 44339 expands the implementation substantially and adds new code paths beyond the basic PR 41356. They look like related iterations, not mergeable duplicates." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 5, + "component_count": 26, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-41797-3", + "nodes": [ + "pull_request:41797", + "pull_request:45186", + "pull_request:45630" + ], + "soft_pairs": [ + "pull_request:41797|pull_request:45186", + "pull_request:45186|pull_request:45630" + ], + "prepared_review_unit_hash": "565e46cba748ec0ccc128fc81d8213c1930f227e3cfdd2973297dc84f2da1457", + "component_budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 10428, + "estimated_input_tokens": 2607, + "estimated_eval_tokens": 5470 + }, + "budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 10428, + "estimated_input_tokens": 2607, + "estimated_eval_tokens": 5470 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are three distinct model-addition pull requests (DeepSeek OCR, Isaac, and Kimi2-6). They share framework plumbing and docs/auto-registration patterns, but they do not fix the same concrete code path or represent the same change.", + "confidence": 0.96, + "canonical_issue_reason": null, + "canonical_pr_reason": "No canonical duplicate PR: each changes a different model integration, with different files, configs, and model-specific implementations.", + "best_issue_reason": null, + "best_pr_reason": "PR 45186 is the most representative of the shared pattern (a new model integration), but it is still not a duplicate of the others.", + "soft_edge_verdicts": [ + { + "left": "pull_request:41797", + "right": "pull_request:45186", + "accept": false, + "reason": "Different models and codepaths: DeepSeek OCR vs Isaac. Shared registration/docs scaffolding is not enough to treat them as the same change." + }, + { + "left": "pull_request:45186", + "right": "pull_request:45630", + "accept": false, + "reason": "Different models and integrations: Isaac vs Kimi2-6. Similar onboarding pattern, but not the same underlying bug or feature." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 6, + "component_count": 26, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-42975-2", + "nodes": [ + "pull_request:42975", + "pull_request:42976" + ], + "soft_pairs": [ + "pull_request:42975|pull_request:42976" + ], + "prepared_review_unit_hash": "86b0f308da69d8582d76e0bbb3b9e0b24ca7192adfe20e25b5e6f15b35eb6f66", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5935, + "estimated_input_tokens": 1484, + "estimated_eval_tokens": 3224 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5935, + "estimated_input_tokens": 1484, + "estimated_eval_tokens": 3224 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Both PRs are broad GitHub Actions dependency bumps, but they fix different things: #42975 upgrades several workflow actions for Node 24 compatibility, while #42976 mainly bumps docker/build-push-action in Docker image workflows. They overlap in files, but not in underlying change or failure mode.", + "confidence": 0.95, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #42975 is the better canonical representative because it is the broader, explicitly stated Node 24 compatibility upgrade across many workflows.", + "best_issue_reason": null, + "best_pr_reason": "PR #42975 is the strongest overall representative of the cluster: it has the clearer problem statement and broader workflow-action modernization scope.", + "soft_edge_verdicts": [ + { + "left": "pull_request:42975", + "right": "pull_request:42976", + "accept": false, + "reason": "They are related workflow maintenance changes, but not the same concrete bug/change. #42975 upgrades assorted GitHub Actions for Node 24 compatibility; #42976 specifically updates docker/build-push-action versions in Docker build workflows. Different actions, different code-path risk, so do not merge as duplicates." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 7, + "component_count": 26, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-42978-4", + "nodes": [ + "pull_request:42978", + "pull_request:43448", + "pull_request:43451", + "pull_request:45490" + ], + "soft_pairs": [ + "pull_request:42978|pull_request:43448", + "pull_request:42978|pull_request:43451", + "pull_request:42978|pull_request:45490", + "pull_request:43448|pull_request:43451", + "pull_request:43448|pull_request:45490", + "pull_request:43451|pull_request:45490" + ], + "prepared_review_unit_hash": "2770b640ce9fae6d15117d28f3fd7fae7902094fee9d82c798399bbdc54c133f", + "component_budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 6, + "serialized_chars": 14074, + "estimated_input_tokens": 3519, + "estimated_eval_tokens": 7294 + }, + "budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 6, + "serialized_chars": 14074, + "estimated_input_tokens": 3519, + "estimated_eval_tokens": 7294 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are four separate model-addition PRs (ViT NEPA, Molmo, Molmo2, and CTSM) that only overlap at the pattern level: docs, auto-registration, and model scaffolding. They do not appear to fix the same bug or implement the same change, so no soft-duplicate edges should be accepted.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 43451 (Molmo2) is the most substantial and best-documented representative of the cluster, with the broadest set of files and the highest review activity, but it is still a distinct model addition rather than a duplicate of the others.", + "best_issue_reason": null, + "best_pr_reason": "PR 43451 is the strongest stand-alone PR in the set: it adds a full Molmo2 stack including model, processors, video processing, auto mappings, docs, and tests.", + "soft_edge_verdicts": [ + { + "left": "pull_request:42978", + "right": "pull_request:43448", + "accept": false, + "reason": "Different model families and code paths: ViT NEPA vs Molmo. Shared docs/auto files are just boilerplate for adding a new model." + }, + { + "left": "pull_request:42978", + "right": "pull_request:43451", + "accept": false, + "reason": "Different implementations for different models (ViT NEPA vs Molmo2); overlap is only in generic integration files like docs and auto registration." + }, + { + "left": "pull_request:42978", + "right": "pull_request:45490", + "accept": false, + "reason": "CTSM is a separate time-series model addition, not the same bug/change as ViT NEPA." + }, + { + "left": "pull_request:43448", + "right": "pull_request:43451", + "accept": false, + "reason": "Molmo and Molmo2 are related by naming but are distinct model releases with different files, processors, and model code; not the same concrete change." + }, + { + "left": "pull_request:43448", + "right": "pull_request:45490", + "accept": false, + "reason": "Molmo vs CTSM are unrelated model additions; similarity is limited to scaffolding and documentation." + }, + { + "left": "pull_request:43451", + "right": "pull_request:45490", + "accept": false, + "reason": "Molmo2 and CTSM touch different model stacks and solve different integration tasks, so they are not duplicates." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 8, + "component_count": 26, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43291-3", + "nodes": [ + "pull_request:43291", + "pull_request:43665", + "pull_request:44815" + ], + "soft_pairs": [ + "pull_request:43291|pull_request:43665", + "pull_request:43291|pull_request:44815", + "pull_request:43665|pull_request:44815" + ], + "prepared_review_unit_hash": "92f252785ab5953aa569048892289f192753151ab51023b67ad90ee2add91997", + "component_budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 3, + "serialized_chars": 7093, + "estimated_input_tokens": 1774, + "estimated_eval_tokens": 3804 + }, + "budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 3, + "serialized_chars": 7093, + "estimated_input_tokens": 1774, + "estimated_eval_tokens": 3804 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "All three PRs are unrelated: two are test-only changes for different models, and one is a substantial FP8/dequantization fix across loading/quantization code. None look like duplicates of the same underlying bug or change.", + "confidence": 0.95, + "canonical_issue_reason": null, + "canonical_pr_reason": "pull_request:44815 is the only substantive code-fix PR; it targets FP8 dequant/padding/loading behavior across core loading and finegrained FP8 code, unlike the other two test-only PRs.", + "best_issue_reason": null, + "best_pr_reason": "pull_request:44815 is the best representative PR because it addresses a concrete implementation bug and has the broadest real code impact; the others are not comparable as change candidates.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43291", + "right": "pull_request:43665", + "accept": false, + "reason": "Both are test-focused, but they touch different models (Whisper vs CLIP/ViT) and do not describe the same bug or code-path fix." + }, + { + "left": "pull_request:43291", + "right": "pull_request:44815", + "accept": false, + "reason": "Whisper tokenization/test cleanup is unrelated to the FP8 dequant/loading fix in 44815; they do not share a concrete underlying issue." + }, + { + "left": "pull_request:43665", + "right": "pull_request:44815", + "accept": false, + "reason": "43665 is just noisy test edits for CLIP/ViT, while 44815 is a real FP8 quantization/dequantization fix; no same-bug overlap." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 9, + "component_count": 26, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43488-2", + "nodes": [ + "pull_request:43488", + "pull_request:43532" + ], + "soft_pairs": [ + "pull_request:43488|pull_request:43532" + ], + "prepared_review_unit_hash": "23b02b32c3859e2edf7cd896c98172c5aa0fbab223c1cc6d8ed403b12551559f", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4563, + "estimated_input_tokens": 1141, + "estimated_eval_tokens": 2538 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4563, + "estimated_input_tokens": 1141, + "estimated_eval_tokens": 2538 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The two PRs are not duplicates: #43488 is a small packaging/version-bump and test-bot check, while #43532 is a CI/workflow and testing-utils refactor. The only overlap is a shared test file, but the underlying changes are different.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": "No single canonical PR: #43488 and #43532 do not address the same code-path problem, and their shared filename is incidental rather than evidence of duplication.", + "best_issue_reason": null, + "best_pr_reason": "#43532 is the more substantial and representative PR in this cluster, but only as a cluster representative\u2014not because it duplicates #43488.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43488", + "right": "pull_request:43532", + "accept": false, + "reason": "Reject: the overlap is only a shared test file. #43488 changes setup/versioning and a dependency table; #43532 changes GitHub workflows, testing utilities, and related test behavior. They are not the same underlying bug or change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 10, + "component_count": 26, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43612-2", + "nodes": [ + "pull_request:43612", + "pull_request:43613" + ], + "soft_pairs": [ + "pull_request:43612|pull_request:43613" + ], + "prepared_review_unit_hash": "44e9e754f2f01dbcc382f42f9207e83aeb8a0979a3dbf0e6e35d09c0cf0d0221", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 6082, + "estimated_input_tokens": 1521, + "estimated_eval_tokens": 3298 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 6082, + "estimated_input_tokens": 1521, + "estimated_eval_tokens": 3298 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are two related but distinct PRs: one adds a Promptable Concept Segmentation pipeline, the other adds a Promptable Visual Segmentation pipeline. They share documentation and pipeline plumbing, but they target different tasks and model integrations, so they are not duplicates.", + "confidence": 0.96, + "canonical_issue_reason": null, + "canonical_pr_reason": "No duplicate PR exists here. If a single representative is needed, PR 43613 is the broader integration, but it is still a different feature from PR 43612.", + "best_issue_reason": null, + "best_pr_reason": "PR 43613 is the better standalone representative because it covers the broader visual segmentation stack and more related integrations/tests, while PR 43612 is a separate concept-segmentation feature.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43612", + "right": "pull_request:43613", + "accept": false, + "reason": "Shared docs and pipeline scaffolding do not make these the same change. They implement different segmentation pipelines with different targets and supporting model/test files, so they should not be merged as duplicates." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 11, + "component_count": 26, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43651-2", + "nodes": [ + "pull_request:43651", + "pull_request:43663" + ], + "soft_pairs": [ + "pull_request:43651|pull_request:43663" + ], + "prepared_review_unit_hash": "379982c43366d9f992a3139f94113da4913e4e668bfc6c291405d06a324add00", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4291, + "estimated_input_tokens": 1073, + "estimated_eval_tokens": 2402 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4291, + "estimated_input_tokens": 1073, + "estimated_eval_tokens": 2402 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Both PRs touch `src/transformers/trainer.py` and add customization hooks, but they address different trainer behaviors: gradient-accumulation loss scaling vs. signature-column filtering. They are not the same underlying change.", + "confidence": 0.95, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": "PR #43651 is the cleaner representative of the cluster because it makes a specific training-step behavior change with a clear code-path impact, while #43663 is a separate override hook for dataset column filtering.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43651", + "right": "pull_request:43663", + "accept": false, + "reason": "They only share the same file and general Trainer customization theme. One changes gradient accumulation loss scaling in `training_step`, the other extracts signature-column logic for feature filtering. These are different code paths and not plausibly a single merged fix/change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 12, + "component_count": 26, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43656-3", + "nodes": [ + "pull_request:43656", + "pull_request:43836", + "pull_request:43842" + ], + "soft_pairs": [ + "pull_request:43656|pull_request:43842", + "pull_request:43836|pull_request:43842" + ], + "prepared_review_unit_hash": "be7c77d81a19ce46f69580ae4a1a1e7b79143bfbc777eef0f74401e5f901aed7", + "component_budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 5993, + "estimated_input_tokens": 1499, + "estimated_eval_tokens": 3254 + }, + "budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 5993, + "estimated_input_tokens": 1499, + "estimated_eval_tokens": 3254 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Two PRs appear to target the same TypeAdapter import/annotation NameError in `src/transformers/cli/serve.py`; the third PR is a broader typing cleanup that is not the same concrete fix.", + "confidence": 0.73, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #43842 is the most focused and explicit fix: it directly addresses the `TypeAdapter` NameError when `pydantic` is absent by moving the import under `TYPE_CHECKING` and quoting the annotation.", + "best_issue_reason": null, + "best_pr_reason": "PR #43842 is the best representative because it is narrowly scoped to the runtime NameError and is less speculative than the broader typing changes in the other PRs.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43656", + "right": "pull_request:43842", + "accept": false, + "reason": "Both touch `serve.py` and mention `TypeAdapter`, but #43656 looks like a broader annotation/type-hint cleanup, while #43842 specifically fixes the `TypeAdapter` import-time NameError when `pydantic` is not installed. Different concrete fixes, so not a duplicate pair." + }, + { + "left": "pull_request:43836", + "right": "pull_request:43842", + "accept": true, + "reason": "Both PRs are working around the same `TypeAdapter` NameError in the CLI serve path, with overlapping changes to quoting/type-checking annotations in the same file. They look like two versions of the same underlying fix and could plausibly be merged into one PR." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet: #43842 is the most focused fix for the `TypeAdapter` NameError, #43656 is broader annotation cleanup in the same file, and #43836 shares the same issue target and file with #43842. The soft-edge verdicts are conservative overall: rejecting #43656 vs #43842 is justified by the broader/different-looking change, and accepting #43836 vs #43842 is reasonable because they appear to be two implementations of the same underlying `TypeAdapter` bugfix in `src/transformers/cli/serve.py`." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 13, + "component_count": 26, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43779-2", + "nodes": [ + "pull_request:43779", + "pull_request:43816" + ], + "soft_pairs": [ + "pull_request:43779|pull_request:43816" + ], + "prepared_review_unit_hash": "e7c674adcc031aa8883109a5dfc678b4407fb20b10764302092e384d031ba83a", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4362, + "estimated_input_tokens": 1091, + "estimated_eval_tokens": 2438 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4362, + "estimated_input_tokens": 1091, + "estimated_eval_tokens": 2438 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Both pull requests appear to implement the same SwanLab integration change: adding run id/resume support in `integration_utils.py` for issue #43698, with one variant also expanding the docs. They are strong duplicate PR candidates.", + "confidence": 0.94, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #43816 is the cleaner representative: it carries the same SwanLab id/resume support plus the accompanying documentation updates, making it the more complete version of the shared change.", + "best_issue_reason": null, + "best_pr_reason": "PR #43816 is the best global representative because it appears to include the full feature change and docs for SwanLab run id/resume support, while #43779 is the narrower implementation-focused variant.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43779", + "right": "pull_request:43816", + "accept": true, + "reason": "Same underlying SwanLab callback enhancement for id/resume support in the same file and same issue target; the doc differences do not change the core code-path change, so these are duplicate PRs." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet: both PRs target issue #43698, touch the same file, and the previews indicate the same SwanLab id/resume support with #43816 adding docs. The duplicate/mergeable verdict is conservative enough given the shared target and overlapping code change." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 14, + "component_count": 26, + "review_unit_index": 1, + "review_unit_count": 2, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43996-28", + "nodes": [ + "pull_request:43996", + "pull_request:43997", + "pull_request:44007", + "pull_request:44010", + "pull_request:44013", + "pull_request:44017", + "pull_request:44018", + "pull_request:44019", + "pull_request:44024", + "pull_request:44044", + "pull_request:44056", + "pull_request:44059", + "pull_request:44066", + "pull_request:44068", + "pull_request:44071", + "pull_request:44072", + "pull_request:44073", + "pull_request:44074", + "pull_request:44076", + "pull_request:44085", + "pull_request:44086", + "pull_request:44098", + "pull_request:44101", + "pull_request:44116", + "pull_request:44129", + "pull_request:44154", + "pull_request:44161", + "pull_request:44722" + ], + "soft_pairs": [ + "pull_request:43996|pull_request:44085", + "pull_request:43996|pull_request:44044", + "pull_request:44066|pull_request:44085", + "pull_request:44007|pull_request:44072", + "pull_request:44072|pull_request:44722", + "pull_request:44066|pull_request:44072", + "pull_request:44013|pull_request:44044", + "pull_request:44066|pull_request:44086", + "pull_request:44066|pull_request:44071", + "pull_request:44018|pull_request:44068", + "pull_request:44066|pull_request:44068", + "pull_request:44018|pull_request:44066", + "pull_request:44019|pull_request:44722", + "pull_request:44086|pull_request:44722", + "pull_request:44019|pull_request:44085", + "pull_request:44018|pull_request:44019", + "pull_request:44019|pull_request:44071", + "pull_request:44068|pull_request:44116", + "pull_request:44068|pull_request:44722", + "pull_request:44071|pull_request:44722", + "pull_request:44018|pull_request:44116", + "pull_request:44018|pull_request:44086", + "pull_request:44018|pull_request:44085", + "pull_request:44068|pull_request:44086", + "pull_request:44071|pull_request:44085", + "pull_request:44059|pull_request:44086", + "pull_request:44085|pull_request:44086", + "pull_request:44098|pull_request:44101", + "pull_request:44059|pull_request:44722", + "pull_request:44018|pull_request:44071", + "pull_request:44017|pull_request:44722", + "pull_request:44017|pull_request:44161", + "pull_request:44017|pull_request:44073", + "pull_request:44068|pull_request:44071", + "pull_request:44071|pull_request:44086", + "pull_request:43997|pull_request:44073", + "pull_request:44017|pull_request:44071", + "pull_request:44073|pull_request:44101", + "pull_request:44059|pull_request:44071", + "pull_request:44059|pull_request:44068", + "pull_request:44018|pull_request:44059", + "pull_request:44056|pull_request:44101", + "pull_request:44073|pull_request:44161", + "pull_request:44101|pull_request:44161", + "pull_request:43997|pull_request:44017", + "pull_request:43997|pull_request:44056", + "pull_request:44071|pull_request:44076", + "pull_request:44074|pull_request:44101", + "pull_request:44056|pull_request:44161", + "pull_request:43997|pull_request:44018", + "pull_request:43997|pull_request:44161", + "pull_request:44074|pull_request:44161", + "pull_request:43997|pull_request:44071", + "pull_request:44010|pull_request:44071", + "pull_request:44010|pull_request:44073", + "pull_request:44018|pull_request:44129", + "pull_request:43997|pull_request:44074", + "pull_request:43997|pull_request:44129", + "pull_request:44010|pull_request:44018", + "pull_request:44071|pull_request:44129", + "pull_request:43996|pull_request:44161", + "pull_request:43996|pull_request:44017", + "pull_request:44010|pull_request:44044", + "pull_request:44044|pull_request:44074", + "pull_request:44066|pull_request:44722", + "pull_request:44085|pull_request:44722", + "pull_request:44085|pull_request:44129", + "pull_request:43996|pull_request:44007", + "pull_request:43996|pull_request:44129", + "pull_request:44044|pull_request:44129", + "pull_request:44044|pull_request:44154", + "pull_request:44024|pull_request:44076" + ], + "prepared_review_unit_hash": "f2efc4714add0223714f7bdae4365e9624d6f01c309f19753206e571ae5cc251", + "component_budget": { + "node_count": 28, + "item_count": 28, + "soft_pair_count": 77, + "serialized_chars": 84508, + "estimated_input_tokens": 21127, + "estimated_eval_tokens": 42510 + }, + "budget": { + "node_count": 28, + "item_count": 28, + "soft_pair_count": 72, + "serialized_chars": 76940, + "estimated_input_tokens": 19235, + "estimated_eval_tokens": 38726 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a wave of model-specific PRs migrating output tracing to `capture_outputs`/`can_return_tuple`. Only the same-model GPT-Neo and GPT-J pairs look like true duplicates; most other pairs are parallel refactors across different architectures and should stay separate.", + "confidence": 0.82, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 44007 is the broadest, clearest representative of the initiative: it has an explicit issue link, touches multiple related vision backbones, and shows the standardized output-tracing refactor in a concrete, reusable form.", + "best_issue_reason": null, + "best_pr_reason": "PR 44007 is the best single anchor for the cluster because it is issue-linked, relatively broad, and most representative of the model-by-model output-tracing migration pattern.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43996", + "right": "pull_request:44085", + "accept": false, + "reason": "Different model families and code paths; same umbrella refactor theme only." + }, + { + "left": "pull_request:43996", + "right": "pull_request:44044", + "accept": false, + "reason": "Different architectures (FNet/CVT vs DeBERTa); not the same concrete change." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44085", + "accept": true, + "reason": "Same GPT-J output-tracing refactor in the same model file; these look like overlapping versions of the same change." + }, + { + "left": "pull_request:44007", + "right": "pull_request:44072", + "accept": false, + "reason": "Both are output-tracing migrations, but for different vision models and code paths." + }, + { + "left": "pull_request:44072", + "right": "pull_request:44722", + "accept": false, + "reason": "EfficientNet and GPT-J/CodeGen are unrelated implementations; same initiative is not enough." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44072", + "accept": false, + "reason": "GPT-J/CodeGen vs EfficientNet are separate model refactors, not one duplicate bug fix." + }, + { + "left": "pull_request:44013", + "right": "pull_request:44044", + "accept": false, + "reason": "MobileNetV2 and DeBERTa are different model stacks with unrelated output-tracing code." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44086", + "accept": false, + "reason": "GPT-J/CodeGen and MGP-STR are different code paths; same pattern only." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44071", + "accept": false, + "reason": "GPT-J/CodeGen vs MPT are distinct model implementations, so not duplicates." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44068", + "accept": true, + "reason": "Same GPT-Neo file and same decorator-based output-tracing migration; one is effectively a rewrite of the other." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44068", + "accept": false, + "reason": "GPT-J/CodeGen vs GPT-Neo are different model families and should not be merged." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44066", + "accept": false, + "reason": "Both are output-tracing refactors, but for different models (GPT-Neo vs GPT-J/CodeGen)." + }, + { + "left": "pull_request:44019", + "right": "pull_request:44722", + "accept": false, + "reason": "ResNet and GPT-J/CodeGen are unrelated code paths." + }, + { + "left": "pull_request:44086", + "right": "pull_request:44722", + "accept": false, + "reason": "MGP-STR and GPT-J/CodeGen are different model implementations." + }, + { + "left": "pull_request:44019", + "right": "pull_request:44085", + "accept": false, + "reason": "ResNet and GPT-J are different model families; not a duplicate change." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44019", + "accept": false, + "reason": "GPT-Neo and ResNet are unrelated; only the refactor style overlaps." + }, + { + "left": "pull_request:44019", + "right": "pull_request:44071", + "accept": false, + "reason": "ResNet and MPT are different code paths with no shared bug/change." + }, + { + "left": "pull_request:44068", + "right": "pull_request:44116", + "accept": false, + "reason": "GPT-Neo and Flaubert are separate model refactors, not one mergeable PR." + }, + { + "left": "pull_request:44068", + "right": "pull_request:44722", + "accept": false, + "reason": "GPT-Neo vs GPT-J/CodeGen are different model families despite similar output-tracing mechanics." + }, + { + "left": "pull_request:44071", + "right": "pull_request:44722", + "accept": false, + "reason": "MPT and GPT-J/CodeGen are unrelated model implementations." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44116", + "accept": false, + "reason": "GPT-Neo and Flaubert are different models; no same concrete code-path problem." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44086", + "accept": false, + "reason": "GPT-Neo and MGP-STR are different model refactors." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44085", + "accept": false, + "reason": "GPT-Neo and GPT-J are distinct model families." + }, + { + "left": "pull_request:44068", + "right": "pull_request:44086", + "accept": false, + "reason": "GPT-Neo and MGP-STR are unrelated implementations." + }, + { + "left": "pull_request:44098", + "right": "pull_request:44101", + "accept": false, + "reason": "ViLT and Flaubert are different architectures; same output-handling theme only." + }, + { + "left": "pull_request:44059", + "right": "pull_request:44722", + "accept": false, + "reason": "GPT-2 and GPT-J/CodeGen are separate refactors with different code paths." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44071", + "accept": false, + "reason": "GPT-Neo and MPT are unrelated model families." + }, + { + "left": "pull_request:44017", + "right": "pull_request:44722", + "accept": false, + "reason": "Segformer and GPT-J/CodeGen are different models; not one duplicate change." + }, + { + "left": "pull_request:44017", + "right": "pull_request:44161", + "accept": false, + "reason": "Segformer and LongT5 are separate output-tracing migrations." + }, + { + "left": "pull_request:44017", + "right": "pull_request:44073", + "accept": false, + "reason": "Segformer and VisualBert are different model code paths." + }, + { + "left": "pull_request:44068", + "right": "pull_request:44071", + "accept": false, + "reason": "GPT-Neo and MPT are unrelated despite similar decorator usage." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44073", + "accept": false, + "reason": "RegNet and VisualBert are different model implementations." + }, + { + "left": "pull_request:44017", + "right": "pull_request:44018", + "accept": false, + "reason": "Segformer and GPT-Neo are different architectures." + }, + { + "left": "pull_request:44073", + "right": "pull_request:44101", + "accept": false, + "reason": "VisualBert and Flaubert are different model stacks." + }, + { + "left": "pull_request:44059", + "right": "pull_request:44071", + "accept": false, + "reason": "GPT-2 and MPT are unrelated code paths." + }, + { + "left": "pull_request:44059", + "right": "pull_request:44068", + "accept": false, + "reason": "GPT-2 and GPT-Neo are different model families; same refactor style is not enough." + }, + { + "left": "pull_request:44056", + "right": "pull_request:44101", + "accept": false, + "reason": "MPNet and Flaubert are different models with separate output-tracing code." + }, + { + "left": "pull_request:44073", + "right": "pull_request:44161", + "accept": false, + "reason": "VisualBert and LongT5 are unrelated model refactors." + }, + { + "left": "pull_request:44101", + "right": "pull_request:44161", + "accept": false, + "reason": "Flaubert and LongT5 are different model families." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44017", + "accept": false, + "reason": "RegNet and Segformer are different vision backbones; not the same change." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44056", + "accept": false, + "reason": "RegNet and MPNet are different model code paths." + }, + { + "left": "pull_request:44071", + "right": "pull_request:44076", + "accept": false, + "reason": "MPT and ImageGPT are separate model implementations." + }, + { + "left": "pull_request:44074", + "right": "pull_request:44101", + "accept": false, + "reason": "TextNet and Flaubert are unrelated refactors." + }, + { + "left": "pull_request:44056", + "right": "pull_request:44161", + "accept": false, + "reason": "MPNet and LongT5 are different models; same initiative only." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44018", + "accept": false, + "reason": "RegNet and GPT-Neo are unrelated code paths." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44161", + "accept": false, + "reason": "RegNet and LongT5 are different model families." + }, + { + "left": "pull_request:44074", + "right": "pull_request:44161", + "accept": false, + "reason": "TextNet and LongT5 are different model implementations." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44071", + "accept": false, + "reason": "RegNet and MPT are unrelated models." + }, + { + "left": "pull_request:44010", + "right": "pull_request:44071", + "accept": false, + "reason": "SqueezeBert and MPT are different architectures." + }, + { + "left": "pull_request:44010", + "right": "pull_request:44073", + "accept": false, + "reason": "SqueezeBert and VisualBert are separate model refactors." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44129", + "accept": false, + "reason": "GPT-Neo and SpeechT5 are different code paths." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44074", + "accept": false, + "reason": "RegNet and TextNet are unrelated model families." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44129", + "accept": false, + "reason": "RegNet and SpeechT5 are different implementations." + }, + { + "left": "pull_request:44071", + "right": "pull_request:44129", + "accept": false, + "reason": "MPT and SpeechT5 are unrelated code paths." + }, + { + "left": "pull_request:43996", + "right": "pull_request:44161", + "accept": false, + "reason": "FNet/CVT and LongT5 are different model families." + }, + { + "left": "pull_request:43996", + "right": "pull_request:44017", + "accept": false, + "reason": "FNet/CVT and Segformer are separate vision model refactors." + }, + { + "left": "pull_request:44010", + "right": "pull_request:44044", + "accept": false, + "reason": "SqueezeBert and DeBERTa are different model code paths." + }, + { + "left": "pull_request:44044", + "right": "pull_request:44074", + "accept": false, + "reason": "DeBERTa and TextNet are unrelated refactors." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44722", + "accept": true, + "reason": "Same GPT-J output-tracing refactor in the same file; 44722 looks like the broader follow-on version." + }, + { + "left": "pull_request:44085", + "right": "pull_request:44722", + "accept": true, + "reason": "Both target GPT-J output tracing in the same file, so they are likely overlapping variants of the same change." + }, + { + "left": "pull_request:44085", + "right": "pull_request:44129", + "accept": false, + "reason": "GPT-J and SpeechT5 are different model families." + }, + { + "left": "pull_request:43996", + "right": "pull_request:44007", + "accept": false, + "reason": "FNet/CVT and RegNet/ResNet/RT-DETR are different model refactors." + }, + { + "left": "pull_request:43996", + "right": "pull_request:44129", + "accept": false, + "reason": "FNet/CVT and SpeechT5 are unrelated implementations." + }, + { + "left": "pull_request:44044", + "right": "pull_request:44129", + "accept": false, + "reason": "DeBERTa and SpeechT5 are different model code paths." + }, + { + "left": "pull_request:44044", + "right": "pull_request:44154", + "accept": false, + "reason": "DeBERTa and VITS are separate refactors." + }, + { + "left": "pull_request:44024", + "right": "pull_request:44076", + "accept": false, + "reason": "FocalNet and ImageGPT are different model families; same output-tracing style only." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet: it correctly characterizes this as a broad output-tracing migration wave and avoids claiming cross-architecture pairs are duplicates. The two accepted soft edges are also conservative because they involve the same model/file family and look like overlapping implementations of the same refactor, while the rejected pairs are mostly different architectures or unrelated code paths." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 14, + "component_count": 26, + "review_unit_index": 2, + "review_unit_count": 2, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43996-28", + "nodes": [ + "pull_request:43997", + "pull_request:44010", + "pull_request:44056", + "pull_request:44073", + "pull_request:44074", + "pull_request:44129" + ], + "soft_pairs": [ + "pull_request:44073|pull_request:44074", + "pull_request:44056|pull_request:44073", + "pull_request:44056|pull_request:44074", + "pull_request:43997|pull_request:44010", + "pull_request:44010|pull_request:44129" + ], + "prepared_review_unit_hash": "e4e07bf558766a13c786a35a4daed48b3e2eb9b0576517dc5f570736d88125e0", + "component_budget": { + "node_count": 28, + "item_count": 28, + "soft_pair_count": 77, + "serialized_chars": 84508, + "estimated_input_tokens": 21127, + "estimated_eval_tokens": 42510 + }, + "budget": { + "node_count": 6, + "item_count": 6, + "soft_pair_count": 5, + "serialized_chars": 13068, + "estimated_input_tokens": 3267, + "estimated_eval_tokens": 6790 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These PRs are all model-specific refactors toward standardized output tracing/capture, but they target different architectures and code paths. They are similar in style, not duplicates of the same fix, so the soft edges should be rejected.", + "confidence": 0.95, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 44073 is the cleanest representative of the shared refactor pattern: it introduces the standardized tracing helpers in a single model file with a focused change set.", + "best_issue_reason": null, + "best_pr_reason": "PR 44073 is the best cluster representative because it is self-contained, clearly about the standardized output-tracing migration, and is not entangled with unrelated model changes.", + "soft_edge_verdicts": [ + { + "left": "pull_request:44073", + "right": "pull_request:44074", + "accept": false, + "reason": "Both are output-tracing migrations, but for different models (VisualBert vs TextNet). They touch different concrete code paths and are not mergeable as one fix." + }, + { + "left": "pull_request:44056", + "right": "pull_request:44073", + "accept": false, + "reason": "Same general refactor theme, but MPNet and VisualBert are separate implementations with different tracing hooks and output paths." + }, + { + "left": "pull_request:44056", + "right": "pull_request:44074", + "accept": false, + "reason": "Both migrate output capture, but they are model-specific refactors in different files with different layer types and behaviors." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44010", + "accept": false, + "reason": "RegNet and SqueezeBert are unrelated model implementations; this is only a shared refactoring pattern, not the same underlying bug or change." + }, + { + "left": "pull_request:44010", + "right": "pull_request:44129", + "accept": false, + "reason": "SqueezeBert and SpeechT5 both adjust output tracing, but they address different model code paths and SpeechT5 has its own explicit issue context." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 15, + "component_count": 26, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43998-13", + "nodes": [ + "pull_request:43998", + "pull_request:43999", + "pull_request:44000", + "pull_request:44001", + "pull_request:44002", + "pull_request:44003", + "pull_request:44004", + "pull_request:44025", + "pull_request:44026", + "pull_request:44027", + "pull_request:44028", + "pull_request:44029", + "pull_request:44030" + ], + "soft_pairs": [ + "pull_request:43998|pull_request:44002", + "pull_request:43998|pull_request:44000", + "pull_request:43998|pull_request:44001", + "pull_request:43998|pull_request:44003", + "pull_request:43998|pull_request:44029", + "pull_request:43999|pull_request:44001", + "pull_request:43999|pull_request:44002", + "pull_request:43999|pull_request:44025", + "pull_request:43999|pull_request:44026", + "pull_request:43999|pull_request:44027", + "pull_request:44000|pull_request:44001", + "pull_request:44000|pull_request:44002", + "pull_request:44000|pull_request:44003", + "pull_request:44000|pull_request:44025", + "pull_request:44001|pull_request:44002", + "pull_request:44001|pull_request:44003", + "pull_request:44001|pull_request:44025", + "pull_request:44002|pull_request:44029", + "pull_request:44002|pull_request:44003", + "pull_request:44003|pull_request:44028", + "pull_request:44003|pull_request:44004", + "pull_request:43998|pull_request:44004", + "pull_request:44001|pull_request:44004", + "pull_request:44002|pull_request:44004", + "pull_request:44000|pull_request:44004", + "pull_request:44025|pull_request:44026", + "pull_request:44025|pull_request:44027", + "pull_request:44025|pull_request:44028", + "pull_request:44025|pull_request:44029", + "pull_request:44025|pull_request:44030", + "pull_request:44026|pull_request:44029", + "pull_request:44026|pull_request:44027", + "pull_request:44026|pull_request:44028", + "pull_request:44026|pull_request:44030", + "pull_request:44027|pull_request:44028", + "pull_request:44027|pull_request:44029", + "pull_request:44027|pull_request:44030", + "pull_request:44028|pull_request:44029", + "pull_request:44028|pull_request:44030", + "pull_request:44029|pull_request:44030" + ], + "prepared_review_unit_hash": "e37b764106e27d2d7955ac1b6ff583f46ea3a8b926ae1e28b5fbe76e06bd880b", + "component_budget": { + "node_count": 13, + "item_count": 13, + "soft_pair_count": 40, + "serialized_chars": 40164, + "estimated_input_tokens": 10041, + "estimated_eval_tokens": 20338 + }, + "budget": { + "node_count": 13, + "item_count": 13, + "soft_pair_count": 40, + "serialized_chars": 40164, + "estimated_input_tokens": 10041, + "estimated_eval_tokens": 20338 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "This cluster is mostly a set of similarly named, model-specific output-tracing refactors. They share implementation patterns, but each PR targets a different architecture and code path, so they are not duplicates of the same underlying change.", + "confidence": 0.93, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 44025 is the strongest representative: it spans two closely related files in the same family and clearly implements the newer tuple-capture/output-tracing pattern, making it the most central example in this loose cluster.", + "best_issue_reason": null, + "best_pr_reason": "44025 is the best single PR to represent the cluster, but only as a theme anchor; it is not a safe duplicate target for the others because the actual model code paths differ.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43998", + "right": "pull_request:44002", + "accept": false, + "reason": "Same refactor theme, but different model families and forward paths." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44000", + "accept": false, + "reason": "Different architectures; the similarity is only in the output-tracing refactor pattern." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44001", + "accept": false, + "reason": "Separate model code paths, not the same concrete bug or change." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44003", + "accept": false, + "reason": "Different model implementations; too broad to treat as one mergeable PR." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44029", + "accept": false, + "reason": "Unrelated model families despite similar tracing cleanup." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44001", + "accept": false, + "reason": "Both are output-tracing refactors, but for different models and call stacks." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44002", + "accept": false, + "reason": "Different architecture-specific implementations; not the same change." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44025", + "accept": false, + "reason": "Different model families and files; only the refactor style matches." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44026", + "accept": false, + "reason": "Separate encoder/decoder code paths; similarity is superficial." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44027", + "accept": false, + "reason": "Different model-specific tracing refactors, not one concrete fix." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44001", + "accept": false, + "reason": "Same broad refactor theme, but distinct model code paths." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44002", + "accept": false, + "reason": "Different models and outputs; not a duplicate change." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44003", + "accept": false, + "reason": "Too broad; these are separate model implementations." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44025", + "accept": false, + "reason": "Different model families; no shared concrete bug." + }, + { + "left": "pull_request:44001", + "right": "pull_request:44002", + "accept": false, + "reason": "Same tracing-refactor pattern, but unrelated model code." + }, + { + "left": "pull_request:44001", + "right": "pull_request:44003", + "accept": false, + "reason": "Different forward paths and modules; not mergeable as one PR." + }, + { + "left": "pull_request:44001", + "right": "pull_request:44025", + "accept": false, + "reason": "Different architectures; only the refactor motif is shared." + }, + { + "left": "pull_request:44002", + "right": "pull_request:44029", + "accept": false, + "reason": "Different subsystems and model code paths." + }, + { + "left": "pull_request:44002", + "right": "pull_request:44003", + "accept": false, + "reason": "Separate model families; same theme is not enough." + }, + { + "left": "pull_request:44003", + "right": "pull_request:44028", + "accept": false, + "reason": "Different architectures and output formats; not the same bug/change." + }, + { + "left": "pull_request:44003", + "right": "pull_request:44004", + "accept": false, + "reason": "Different concrete code paths; one is Mamba-related, the other CodeGen." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44004", + "accept": false, + "reason": "Unrelated models; only broad refactoring similarity." + }, + { + "left": "pull_request:44001", + "right": "pull_request:44004", + "accept": false, + "reason": "Different model code paths and behavior." + }, + { + "left": "pull_request:44002", + "right": "pull_request:44004", + "accept": false, + "reason": "Different implementations; no shared underlying defect." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44004", + "accept": false, + "reason": "Both touch tracing/output handling, but in different models." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44026", + "accept": false, + "reason": "Different model families; not the same concrete fix." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44027", + "accept": false, + "reason": "Separate encoder-decoder and depth-model changes; too broad." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44028", + "accept": false, + "reason": "Different architectures and output contracts." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44029", + "accept": false, + "reason": "Different model families and code paths." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44030", + "accept": false, + "reason": "Different models; similar refactor wording only." + }, + { + "left": "pull_request:44026", + "right": "pull_request:44029", + "accept": false, + "reason": "Different model implementations and tracing logic." + }, + { + "left": "pull_request:44026", + "right": "pull_request:44027", + "accept": false, + "reason": "Separate model families; no single underlying bug." + }, + { + "left": "pull_request:44026", + "right": "pull_request:44028", + "accept": false, + "reason": "Different architectures and output handling." + }, + { + "left": "pull_request:44026", + "right": "pull_request:44030", + "accept": false, + "reason": "Different model code paths; not mergeable as one PR." + }, + { + "left": "pull_request:44027", + "right": "pull_request:44028", + "accept": false, + "reason": "Different model-specific refactors, not duplicates." + }, + { + "left": "pull_request:44027", + "right": "pull_request:44029", + "accept": false, + "reason": "Separate architectures and output contracts." + }, + { + "left": "pull_request:44027", + "right": "pull_request:44030", + "accept": false, + "reason": "Different code paths; only the tracing refactor theme matches." + }, + { + "left": "pull_request:44028", + "right": "pull_request:44029", + "accept": false, + "reason": "Different model families and behavior." + }, + { + "left": "pull_request:44028", + "right": "pull_request:44030", + "accept": false, + "reason": "Separate models; no shared concrete bug or change." + }, + { + "left": "pull_request:44029", + "right": "pull_request:44030", + "accept": false, + "reason": "Different model implementations; too broad for a duplicate." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 16, + "component_count": 26, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-44662-3", + "nodes": [ + "pull_request:44662", + "pull_request:45075", + "pull_request:45401" + ], + "soft_pairs": [ + "pull_request:44662|pull_request:45401", + "pull_request:44662|pull_request:45075", + "pull_request:45075|pull_request:45401" + ], + "prepared_review_unit_hash": "2ba87f1db6b33bae991de182f8e5ac8815bbf8e9e34da8034b4d5b349e639d43", + "component_budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 3, + "serialized_chars": 10911, + "estimated_input_tokens": 2728, + "estimated_eval_tokens": 5712 + }, + "budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 3, + "serialized_chars": 10911, + "estimated_input_tokens": 2728, + "estimated_eval_tokens": 5712 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "All three pull requests add different model integrations (PenguinVL, DeepSeek-OCR-2, Voxtral TTS). They share boilerplate files because model additions in Transformers follow the same template, but they are not the same underlying change and should not be merged as duplicates.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": "No single canonical PR: these are separate model-addition PRs for different architectures and codepaths, not duplicate fixes or the same feature work.", + "best_issue_reason": null, + "best_pr_reason": "PR 45075 is the most mature/active implementation PR, but it only represents DeepSeek-OCR-2 and does not subsume the other model additions.", + "soft_edge_verdicts": [ + { + "left": "pull_request:44662", + "right": "pull_request:45401", + "accept": false, + "reason": "Different models and different implementation targets (PenguinVL vs Voxtral TTS); shared framework files are expected boilerplate, not evidence of the same change." + }, + { + "left": "pull_request:44662", + "right": "pull_request:45075", + "accept": false, + "reason": "Both are model-add PRs, but for different architectures and codepaths (PenguinVL vs DeepSeek-OCR-2). Similar auto-registry/docs edits are generic template overlap." + }, + { + "left": "pull_request:45075", + "right": "pull_request:45401", + "accept": false, + "reason": "Distinct model integrations for DeepSeek-OCR-2 and Voxtral TTS; they do not fix the same concrete bug or implement the same feature." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 17, + "component_count": 26, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-44979-2", + "nodes": [ + "pull_request:44979", + "pull_request:45363" + ], + "soft_pairs": [ + "pull_request:44979|pull_request:45363" + ], + "prepared_review_unit_hash": "06ec9e1a76dfbf95b769f72b6cb1bb318068b04ca92ed9ef03727052eb02cd61", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4407, + "estimated_input_tokens": 1102, + "estimated_eval_tokens": 2460 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4407, + "estimated_input_tokens": 1102, + "estimated_eval_tokens": 2460 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These two pull requests are related only at a broad theme level (module/kernel fusion) but they do not look like the same underlying change. PR #44979 proposes a general Module Fusion API with a new `module_fusion.py` abstraction, while PR #45363 implements n-to-1 kernel fusion through `KernelConfig` and runtime integration points. They could coexist rather than be merged as duplicates.", + "confidence": 0.93, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #45363 is the better canonical PR for this cluster because it presents a concrete end-to-end fusion mechanism with integration into model loading/execution, whereas #44979 is a more generic API proposal with a narrower, earlier-looking abstraction.", + "best_issue_reason": null, + "best_pr_reason": "PR #45363 is the strongest standalone candidate: it has a specific feature goal, multiple integration files, and a clearer implementation path. PR #44979 is more of a foundational API sketch and does not appear to target the same concrete code path.", + "soft_edge_verdicts": [ + { + "left": "pull_request:44979", + "right": "pull_request:45363", + "accept": false, + "reason": "Do not merge as duplicates: they address different fusion designs and different code paths. #44979 introduces a generic module-fusion API, while #45363 adds kernel-fusion via `KernelConfig` and related integration hooks." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 18, + "component_count": 26, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45086-2", + "nodes": [ + "pull_request:45086", + "pull_request:45317" + ], + "soft_pairs": [ + "pull_request:45086|pull_request:45317" + ], + "prepared_review_unit_hash": "2d118832712d15e2e1c24bf44c0540ee85adc704eea035846db7f0c9f9dc3922", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4414, + "estimated_input_tokens": 1104, + "estimated_eval_tokens": 2464 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4414, + "estimated_input_tokens": 1104, + "estimated_eval_tokens": 2464 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Both pull requests implement the same Mistral regex fix in the same code path; the second PR adds test coverage, so they are duplicates with the later PR being the better canonical representative.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 45317 is the better canonical PR because it contains the same core fix as 45086 plus a test file, making it the more complete and verifiable change.", + "best_issue_reason": null, + "best_pr_reason": "PR 45317 is the strongest representative: same underlying AttributeError fix, same target issue, and additional tests for validation.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45086", + "right": "pull_request:45317", + "accept": true, + "reason": "Same underlying bug fix in _patch_mistral_regex, same file, same issue target, and the later PR is just the fix plus tests." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 19, + "component_count": 26, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45115-2", + "nodes": [ + "pull_request:45115", + "pull_request:45569" + ], + "soft_pairs": [ + "pull_request:45115|pull_request:45569" + ], + "prepared_review_unit_hash": "a010488c94232a807cc2392fd56dca7dbcb9256503b4b633df6289ce3ff60fd8", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 6066, + "estimated_input_tokens": 1517, + "estimated_eval_tokens": 3290 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 6066, + "estimated_input_tokens": 1517, + "estimated_eval_tokens": 3290 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are related NemotronH refactors, but they are not the same change: one is a conversion-mapping/model-inheritance refactor, the other is a broader model split with docs, auto-mapping, and tests. They should not be merged as duplicates.", + "confidence": 0.88, + "canonical_issue_reason": null, + "canonical_pr_reason": "pull_request:45569 is the better canonical PR because it is the broader, more complete NemotronH architecture change, covering the new dense/sparse model split, auto mappings, docs, and tests; pull_request:45115 is a narrower conversion/mapping refactor.", + "best_issue_reason": null, + "best_pr_reason": "pull_request:45569 is the best representative of the cluster since it touches the core model packaging and registration changes rather than only checkpoint conversion mappings.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45115", + "right": "pull_request:45569", + "accept": false, + "reason": "Both are about NemotronH, but 45115 is mainly conversion-mapping and inheritance refactoring, while 45569 is a broader model split into dense/sparse variants with docs, auto-mapping, and tests. They do not appear to fix the same concrete code-path problem and would not plausibly merge into one PR." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 20, + "component_count": 26, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45332-2", + "nodes": [ + "pull_request:45332", + "pull_request:45333" + ], + "soft_pairs": [ + "pull_request:45332|pull_request:45333" + ], + "prepared_review_unit_hash": "29d3bbe9ff0ca8230d72b8b8cedc208dfddbfdc9543a96ba9220ac7b1d82c809", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5398, + "estimated_input_tokens": 1350, + "estimated_eval_tokens": 2956 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5398, + "estimated_input_tokens": 1350, + "estimated_eval_tokens": 2956 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The two PRs are related pieces of the same heterogeneous-model feature, but they are not duplicates: one adds per-layer config plumbing, while the other extends that plumbing into cache/masking/modeling paths.", + "confidence": 0.91, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 45332 is the broader umbrella change: it implements heterogeneous model support across cache/masking/modeling and can incorporate the narrower per-layer config groundwork from 45333.", + "best_issue_reason": null, + "best_pr_reason": "PR 45332 is the best representative of the cluster because it covers the end-to-end feature rather than only the config layer, making it the more comprehensive change to keep.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45332", + "right": "pull_request:45333", + "accept": false, + "reason": "Reject as a duplicate pair: 45333 adds per-layer configuration support, while 45332 uses that groundwork to change cache/modeling behavior. They are adjacent feature PRs, not the same concrete bug or change, and would not plausibly be merged into one PR without becoming a broader umbrella feature." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 21, + "component_count": 26, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45415-2", + "nodes": [ + "pull_request:45415", + "pull_request:45425" + ], + "soft_pairs": [ + "pull_request:45415|pull_request:45425" + ], + "prepared_review_unit_hash": "2ce4e6e36b81f041833bcaa1d30aef7a849fb0424cef41e080136cf23179395c", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5513, + "estimated_input_tokens": 1379, + "estimated_eval_tokens": 3014 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5513, + "estimated_input_tokens": 1379, + "estimated_eval_tokens": 3014 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Two typing-related pull requests, but they are different changes: one is a broad type-checking sweep across many transformers files, the other adds/adjusts typing protocols and helpers for specific models and integrations.", + "confidence": 0.87, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #45415 is the clearest cluster anchor: it is a broad, repo-wide type-checking cleanup with many affected files and an explicit pyproject type-checking config change.", + "best_issue_reason": null, + "best_pr_reason": "PR #45415 is the strongest representative of the cluster overall because it is the most comprehensive and clearly scoped typing-focused change.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45415", + "right": "pull_request:45425", + "accept": false, + "reason": "Both are about typing, but they address different work: #45415 is a large type-checking pass, while #45425 adds specific typing protocols/helpers for modeling_utils and related code paths. They are not the same underlying bug/change and would not naturally merge into one duplicate PR." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 22, + "component_count": 26, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45487-2", + "nodes": [ + "pull_request:45487", + "pull_request:45615" + ], + "soft_pairs": [ + "pull_request:45487|pull_request:45615" + ], + "prepared_review_unit_hash": "52320edc30d152c884e1300110e352c4bb91f510be95feb3909f2886c0ea9dee", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5157, + "estimated_input_tokens": 1290, + "estimated_eval_tokens": 2836 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5157, + "estimated_input_tokens": 1290, + "estimated_eval_tokens": 2836 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "45487 is a substantive cross-model device-placement bug fix; 45615 is only test expectation updates for XPU in qianfan_ocr. They are not the same underlying change.", + "confidence": 0.95, + "canonical_issue_reason": null, + "canonical_pr_reason": "45487 fixes the concrete code-path bug: buffered token_type_ids are moved to the position_ids device before gather, addressing model-parallel/device mismatch across several text-embedding paths. It is the substantive implementation change here.", + "best_issue_reason": null, + "best_pr_reason": "45487 is the best PR candidate because it contains the actual bug fix in model code, while 45615 only adjusts integration test expectations and does not fix the underlying runtime issue.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45487", + "right": "pull_request:45615", + "accept": false, + "reason": "Different artifacts and problems: 45487 changes model code to fix a device-placement bug in AltCLIP/related models, while 45615 only updates qianfan_ocr test outputs for XPU. They would not plausibly merge into one PR." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 23, + "component_count": 26, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45548-2", + "nodes": [ + "pull_request:45548", + "pull_request:45662" + ], + "soft_pairs": [ + "pull_request:45548|pull_request:45662" + ], + "prepared_review_unit_hash": "3d4f19c0aaf88280f68877fdeeaaef1367a375f817d5dbef8716321c05cab30d", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4508, + "estimated_input_tokens": 1127, + "estimated_eval_tokens": 2510 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4508, + "estimated_input_tokens": 1127, + "estimated_eval_tokens": 2510 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Two PRs touch expert-parallel/modeling_utils, but they fix different loading failures: one is config serialization and EP detection for accelerate+ZeRO-3, the other is expert parameters being overwritten during FSDP2 loading.", + "confidence": 0.89, + "canonical_issue_reason": null, + "canonical_pr_reason": "No clear duplicate relationship between the PRs; if one must anchor the cluster, 45662 is the more concrete code-path fix for distributed expert loading, while 45548 is a narrower compatibility patch.", + "best_issue_reason": null, + "best_pr_reason": "45662 is the stronger representative because it addresses a specific runtime bug in expert parameter loading/wrapping across distributed execution, with broader code-path coverage.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45548", + "right": "pull_request:45662", + "accept": false, + "reason": "Different underlying bugs and fixes: 45548 removes distributed_config from serialization and adds has_ep for accelerate/ZeRO-3 compatibility, while 45662 fixes expert tensors being overwritten by rank-0 broadcast during FSDP2 loading. Shared EP context is too broad to treat as the same change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 24, + "component_count": 26, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45639-2", + "nodes": [ + "pull_request:45639", + "pull_request:45645" + ], + "soft_pairs": [ + "pull_request:45639|pull_request:45645" + ], + "prepared_review_unit_hash": "0b1a21c859ca0ae2e186591cca9b0b123c87c74e43762a5ccbd60a3b4186a745", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4589, + "estimated_input_tokens": 1148, + "estimated_eval_tokens": 2552 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4589, + "estimated_input_tokens": 1148, + "estimated_eval_tokens": 2552 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Both PRs address the same xdist-related captured_info logging problem in testing_utils, but 45645 is the broader, more complete change set because it also updates CI handling of multiple log artifacts. 45639 looks like a narrower precursor/slice of the same fix.", + "confidence": 0.87, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 45645 is the better canonical representative because it contains the core xdist-safe testing_utils fix and the additional workflow changes needed to consume multiple captured_info files in CI, making it the more complete implementation.", + "best_issue_reason": null, + "best_pr_reason": "PR 45645 is the strongest overall candidate: it directly targets the xdist collision problem, includes the shared testing_utils change, and extends the fix to downstream CI/debug-log handling.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45639", + "right": "pull_request:45645", + "accept": true, + "reason": "Same underlying xdist captured_info bug/change, with 45645 effectively subsuming 45639 by carrying the same testing_utils fix plus related CI artifact handling." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 25, + "component_count": 26, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45683-2", + "nodes": [ + "pull_request:45683", + "pull_request:45734" + ], + "soft_pairs": [ + "pull_request:45683|pull_request:45734" + ], + "prepared_review_unit_hash": "eba4ca0f26b2d3be0c00e316270a9341d31af2208b44dbab31b1d3a8e81e7538", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4469, + "estimated_input_tokens": 1118, + "estimated_eval_tokens": 2492 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4469, + "estimated_input_tokens": 1118, + "estimated_eval_tokens": 2492 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Two open PRs touch the same quantization helper, but they fix different bugs: one special-cases audio submodules, the other changes how default skip lists are combined with user-supplied skips. They should not be clustered as duplicates.", + "confidence": 0.93, + "canonical_issue_reason": null, + "canonical_pr_reason": "45734 is the better representative PR because it fixes the core skip-module composition logic in `get_modules_to_not_convert`, which is broader and more central than the narrow audio-module special case in 45683.", + "best_issue_reason": null, + "best_pr_reason": "45734 is the strongest standalone PR in the set: it addresses a concrete quantization failure mode in the main code path and its fix is directly about the helper\u2019s behavior rather than a model-specific workaround.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45683", + "right": "pull_request:45734", + "accept": false, + "reason": "Both touch quantization skip logic, but 45683 adds audio-tower/embed_audio exclusions while 45734 changes additive handling of `skip_modules`; these are different underlying bugs and would not naturally merge into one PR." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 26, + "component_count": 26, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45737-2", + "nodes": [ + "pull_request:45737", + "pull_request:45738" + ], + "soft_pairs": [ + "pull_request:45737|pull_request:45738" + ], + "prepared_review_unit_hash": "cc2ade73a49ec6a8883a7a24c433988645588bcd049e8aad2bce56b1b24daa1c", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4131, + "estimated_input_tokens": 1033, + "estimated_eval_tokens": 2322 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4131, + "estimated_input_tokens": 1033, + "estimated_eval_tokens": 2322 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Both PRs address the same MusicgenMelody cache initialization bug in `forward()`: replacing `EncoderDecoderCache(DynamicCache, DynamicCache)` with a single `DynamicCache`. PR #45737 is the more complete duplicate because it bundles the fix with a regression test, while #45738 is the fix-only variant.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #45737 is the best canonical PR because it contains the same code fix as #45738 plus a regression test, making it the most complete representation of the bug fix.", + "best_issue_reason": null, + "best_pr_reason": "PR #45737 is the strongest overall candidate: it fixes the concrete cache bug and adds a targeted regression test, whereas #45738 only changes the implementation.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45737", + "right": "pull_request:45738", + "accept": true, + "reason": "Same underlying code-path bug in `musicgen_melody` cache setup; the diffs match on the core fix and the test-only addition in #45737 could plausibly live in the same PR as #45738." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The assessment is grounded in the packet: both PRs make the same `musicgen_melody` cache initialization change, and #45737 only adds a regression test on top of that fix. The soft-edge accept is conservative enough because the diffs are mergeable into one PR for the same bugfix." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + } + ] +} diff --git a/snapshots/20260506T180045Z/analysis-runs/hybrid-model-20260506t180045z/manifest.json b/snapshots/20260506T180045Z/analysis-runs/hybrid-model-20260506t180045z/manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..3f9f8d710d45459d85eb1bca17cbc5682d693011 --- /dev/null +++ b/snapshots/20260506T180045Z/analysis-runs/hybrid-model-20260506t180045z/manifest.json @@ -0,0 +1,14 @@ +{ + "analysis_id": "hybrid-model-20260506t180045z", + "artifacts": { + "hybrid": "snapshots/20260506T180045Z/analysis-runs/hybrid-model-20260506t180045z/analysis-report-hybrid.json", + "hybrid_reviews": "snapshots/20260506T180045Z/analysis-runs/hybrid-model-20260506t180045z/analysis-report-hybrid.llm-reviews.json" + }, + "channel": "canonical", + "model": null, + "published_at": "2026-05-06T18:15:25Z", + "repo": "huggingface/transformers", + "schema_version": 1, + "snapshot_id": "20260506T180045Z", + "variant": "hybrid" +} diff --git a/snapshots/20260506T180045Z/manifest.json b/snapshots/20260506T180045Z/manifest.json index dba99a9deb13037c70e478771c70436f6c25f5ba..619b4712b074ab50abf9419e65d4b98e50f809b4 100644 --- a/snapshots/20260506T180045Z/manifest.json +++ b/snapshots/20260506T180045Z/manifest.json @@ -32,6 +32,24 @@ "timeline_events": 652 }, "extracted_at": "2026-05-06T18:00:45Z", + "published_analysis": { + "canonical_analysis_id": "hybrid-model-20260506t180045z", + "runs": { + "hybrid-model-20260506t180045z": { + "analysis_id": "hybrid-model-20260506t180045z", + "artifacts": { + "hybrid": "snapshots/20260506T180045Z/analysis-runs/hybrid-model-20260506t180045z/analysis-report-hybrid.json", + "hybrid_reviews": "snapshots/20260506T180045Z/analysis-runs/hybrid-model-20260506t180045z/analysis-report-hybrid.llm-reviews.json" + }, + "channel": "canonical", + "manifest_path": "snapshots/20260506T180045Z/analysis-runs/hybrid-model-20260506t180045z/manifest.json", + "model": null, + "published_at": "2026-05-06T18:15:25Z", + "variant": "hybrid" + } + }, + "schema_version": 1 + }, "repo": "huggingface/transformers", "snapshot_id": "20260506T180045Z", "watermark": {