diff --git a/analysis/current/analysis-report-hybrid.json b/analysis/current/analysis-report-hybrid.json index d617c5238d72ad79dcf06f033b861c3aea5d27d4..b4c1e45a67baf68f41b072bb914b0157fa308e4e 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": "20260422T180025Z", - "generated_at": "2026-04-22T18:09:32Z", + "snapshot_id": "20260423T000027Z", + "generated_at": "2026-04-23T00:09:10Z", "evidence_quality": "full", "llm_enrichment": true, "meta_bugs": [ @@ -1148,41 +1148,6 @@ } ] }, - { - "cluster_id": "cluster-44514-6", - "summary": "Cluster of 4 issues and 2 PRs centered on issue #44521.", - "status": "open", - "confidence": 0.6, - "canonical_issue_number": 44521, - "canonical_pr_number": 44535, - "issue_numbers": [ - 44514, - 44521, - 44545, - 45290 - ], - "pr_numbers": [ - 44535, - 44543 - ], - "evidence_types": [ - "closing_reference", - "soft_similarity" - ], - "pr_comparisons": [ - { - "left_pr_number": 44535, - "right_pr_number": 44543, - "code_similarity": 0.121, - "size_similarity": 0.477, - "file_overlap": 0.0, - "area_overlap": 0.0, - "patch_similarity": 0.169, - "shared_filenames": [], - "shared_file_areas": [] - } - ] - }, { "cluster_id": "cluster-43240-3", "summary": "Cluster of 1 issues and 2 PRs centered on issue #43240.", @@ -1237,7 +1202,7 @@ "cluster_id": "cluster-43824-3", "summary": "Cluster of 1 issues and 2 PRs centered on issue #43824.", "status": "open", - "confidence": 0.55, + "confidence": 0.75, "canonical_issue_number": 43824, "canonical_pr_number": 43836, "issue_numbers": [ @@ -1248,7 +1213,8 @@ 43842 ], "evidence_types": [ - "closing_reference" + "closing_reference", + "shared_issue_target" ], "pr_comparisons": [ { @@ -1695,18 +1661,7 @@ ] } ], - "duplicate_issues": [ - { - "cluster_id": "cluster-44514-6", - "canonical_issue_number": 44521, - "duplicate_issue_numbers": [ - 44514, - 44545, - 45290 - ], - "reason": "Issues in cluster-44514-6 are treated as duplicates because they share closing_reference, soft_similarity evidence." - } - ], + "duplicate_issues": [], "duplicate_prs": [ { "cluster_id": "cluster-41211-3", @@ -1742,7 +1697,7 @@ 43842 ], "target_issue_number": 43824, - "reason": "PRs in cluster-43824-3 are treated as duplicates because they converge on issue #43824 with closing_reference evidence." + "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", @@ -1770,15 +1725,6 @@ "target_issue_number": null, "reason": "PRs in cluster-44018-2 are treated as duplicates because they share soft_similarity evidence." }, - { - "cluster_id": "cluster-44514-6", - "canonical_pr_number": 44535, - "duplicate_pr_numbers": [ - 44543 - ], - "target_issue_number": 44521, - "reason": "PRs in cluster-44514-6 are treated as duplicates because they converge on issue #44521 with closing_reference, soft_similarity evidence." - }, { "cluster_id": "cluster-45081-3", "canonical_pr_number": 45317, @@ -1790,10 +1736,10 @@ } ], "best_issue": { - "cluster_id": "cluster-44514-6", - "issue_number": 44521, - "reason": "Issue #44521 is the strongest global issue candidate because it is open, belongs to a cluster with 6 artifacts, and carries 8 discussion comments plus 5 inbound references.", - "score": 181.0 + "cluster_id": "cluster-41211-3", + "issue_number": 41211, + "reason": "Issue #41211 is the strongest global issue candidate because it is open, belongs to a cluster with 3 artifacts, and carries 6 discussion comments plus 6 inbound references.", + "score": 148.0 }, "best_pr": { "cluster_id": "cluster-43979-11", diff --git a/analysis/current/analysis-report-hybrid.llm-reviews.json b/analysis/current/analysis-report-hybrid.llm-reviews.json index d7b35fefd97e01f8d7545b7138b97a8577e4e810..6166aae5918ee6694fe9ba0059022384e01d40b6 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": "20260422T180025Z", - "generated_at": "2026-04-22T18:09:32Z", + "snapshot_id": "20260423T000027Z", + "generated_at": "2026-04-23T00:09:10Z", "model": "gpt-5.4-mini?service_tier=flex", "reviews": [ { @@ -26,46 +26,46 @@ "issue:13244|issue:16618", "issue:15980|issue:16618" ], - "prepared_review_unit_hash": "dfe6a0a7ff27be9c299961655d69ea55daa906853cef8388929fbccc1533780f", + "prepared_review_unit_hash": "66bfc44967010832d2937d318ab9877350859211b88141bb644aeaced5a638a0", "component_budget": { "node_count": 3, "item_count": 3, "soft_pair_count": 2, - "serialized_chars": 2478, - "estimated_input_tokens": 620, - "estimated_eval_tokens": 1496 + "serialized_chars": 2476, + "estimated_input_tokens": 619, + "estimated_eval_tokens": 1494 }, "budget": { "node_count": 3, "item_count": 3, "soft_pair_count": 2, - "serialized_chars": 2478, - "estimated_input_tokens": 620, - "estimated_eval_tokens": 1496 + "serialized_chars": 2476, + "estimated_input_tokens": 619, + "estimated_eval_tokens": 1494 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These look like two model-loading/download complaints and one unrelated tokenization bug; the broadest and most representative issue is the general model-load failure in 16618.", - "confidence": 0.67, - "canonical_issue_reason": "Issue 16618 is the most general user-facing model-loading problem, while 15980 is specifically about private-model auth/error wording and 13244 is a separate Tapas tokenization mismatch.", + "summary": "These three issues are only loosely related by TensorFlow/Transformers usage, but they describe different bugs: tokenization mismatch, an authentication/error-message problem, and a model-loading failure. I would not merge them as duplicates.", + "confidence": 0.97, + "canonical_issue_reason": "Issue #16618 is the closest cluster anchor because it is the only one with soft similarity links to both other issues and it concerns a concrete model-loading failure, but it is still not a true duplicate of the others.", "canonical_pr_reason": null, - "best_issue_reason": "16618 best captures the cluster\u2019s shared theme of failing to load a model and is the broadest candidate for deduping nearby reports.", + "best_issue_reason": "#16618 is the best representative of this cluster\u2019s narrow overlap because it is the most central node in the similarity graph and has the clearest, self-contained bug report.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:13244", "right": "issue:16618", "accept": false, - "reason": "Different underlying problem: Tapas tokenization parity with TensorFlow vs a model-loading failure for bert-base-uncased." + "reason": "Different underlying problems: Tapas tokenization parity vs TensorFlow code is unrelated to a BERT model-loading failure." }, { "left": "issue:15980", "right": "issue:16618", "accept": false, - "reason": "15980 is about private-model download/auth error messaging; 16618 is a generic inability to load a specific public model, so the concrete bugs do not clearly match." + "reason": "One is about a bad login/error message for private model downloads; the other is about failing to load bert-base-uncased. Same subsystem, but not the same bug." } ] }, @@ -188,8 +188,8 @@ "issue:41762|issue:43596", "issue:43901|issue:44512", "issue:39290|issue:41720", - "issue:43866|issue:44863", "issue:43039|issue:44560", + "issue:43866|issue:44863", "issue:43854|issue:43866", "issue:43425|issue:45070", "issue:39290|issue:43531", @@ -203,12 +203,12 @@ "issue:41093|issue:41762", "issue:43296|issue:43366" ], - "prepared_review_unit_hash": "1d45e9f2626283393ba1a8f1d44e10970986d33bdd899613e91e76a710122e52", + "prepared_review_unit_hash": "15fedc60ae7a6503aa0f525b091f17a2e02913e7b665b8cd4e7fbf2325a10fa2", "component_budget": { "node_count": 367, "item_count": 367, "soft_pair_count": 1370, - "serialized_chars": 706987, + "serialized_chars": 706986, "estimated_input_tokens": 176747, "estimated_eval_tokens": 353750 }, @@ -216,327 +216,327 @@ "node_count": 48, "item_count": 48, "soft_pair_count": 63, - "serialized_chars": 46476, - "estimated_input_tokens": 11619, - "estimated_eval_tokens": 23494 + "serialized_chars": 46483, + "estimated_input_tokens": 11621, + "estimated_eval_tokens": 23498 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly heterogeneous: model-specific runtime, export, docs, and distributed-training bugs. I only see one true duplicate pair: the two DeepSpeed ZeRO-3 Bert index-out-of-bounds reports. No PRs are present.", - "confidence": 0.91, - "canonical_issue_reason": "issue:43596 is the cleaner canonical issue: it reports the same DeepSpeed ZeRO-3 + Bert index-0 out-of-bounds failure as issue:43638, and it is the earlier report.", + "summary": "Mostly a mixed cluster of unrelated model-support, dtype/config, docs, and DeepSpeed/FSDP issues. Only one soft duplicate pair looks high-confidence: the two DeepSpeed ZeRO-3 BERT out-of-bounds reports.", + "confidence": 0.74, + "canonical_issue_reason": "issue:43596 is the cleanest anchor for the accepted duplicate pair: it states the DeepSpeed ZeRO-3 + BERT index-0 out-of-bounds failure clearly and concisely.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43596 is the best representative of the only confirmed duplicate group in this set, with the clearest same-code-path failure description.", + "best_issue_reason": "issue:43596 is the best global issue candidate because it most clearly represents the only strong duplicate pair in this set; the other items are mostly separate bugs or docs problems.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44322", "right": "issue:45464", "accept": false, - "reason": "Both are Qwen3.5-related, but they describe different failures and code paths: missing config field vs streaming chat/completions behavior." + "reason": "Different Qwen3.5 failures: config attribute error vs streaming chat/completions error; not the same code path." }, { "left": "issue:43716", "right": "issue:45237", "accept": false, - "reason": "Different model families and failure modes: Mistral-3 image/text dtype mismatch vs GPT-OSS AMD GPU failure." + "reason": "Mistral-3 image-preprocessor dtype mismatch is unrelated to GPT-OSS AMD GPU failure." }, { "left": "issue:43828", "right": "issue:45237", "accept": false, - "reason": "Different models and different bugs: autocast dtype mismatch in Phi-tiny-MoE vs GPU backend failure in GPT-OSS." + "reason": "Different models and different failure modes; both are dtype/runtime issues only at a very broad level." }, { "left": "issue:45237", "right": "issue:45507", "accept": false, - "reason": "Different model implementations and error types; no shared concrete code-path bug." + "reason": "GPT-OSS on AMD GPUs vs GraniteMoEHybrid invalid-method error are unrelated." }, { "left": "issue:44292", "right": "issue:45237", "accept": false, - "reason": "Different models and subsystems: Qwen-3 NVFP4 runtime error vs GPT-OSS AMD GPU issue." + "reason": "Qwen-3-8B-NVFP4 runtime error is a different model/path from GPT-OSS AMD GPU support." }, { "left": "issue:30277", "right": "issue:35545", "accept": false, - "reason": "Different problems: DeepSpeed collective mismatch in training vs ModernBERT ONNX export failure." + "reason": "DeepSpeed collective mismatch and ModernBERT ONNX export are unrelated bugs." }, { "left": "issue:35545", "right": "issue:42915", "accept": false, - "reason": "Different model/export bug versus Qwen3MoE FP8 config failure." + "reason": "ModernBERT ONNX export vs Qwen3Moe FP8 failure are different models and subsystems." }, { "left": "issue:35545", "right": "issue:41720", "accept": false, - "reason": "ModernBERT ONNX export and Qwen3 auto device mapping cuda assert are unrelated failures." + "reason": "ONNX export issue is unrelated to Qwen3 auto device-mapping cuda assert." }, { "left": "issue:44509", "right": "issue:44512", "accept": false, - "reason": "Both are docs cleanup issues, but they target different removed features and different pages." + "reason": "Both are docs cleanup, but they reference different removed features and different pages; too broad to treat as the same bug." }, { "left": "issue:35545", "right": "issue:36010", "accept": false, - "reason": "Different subsystems: ONNX export vs GenerationMixin import path breakage." + "reason": "ONNX export bug and GenerationMixin import regression are unrelated." }, { "left": "issue:43638", "right": "issue:44805", "accept": false, - "reason": "Both mention index errors, but the specific symptoms and contexts differ; not the same underlying bug." + "reason": "Both mention indexing errors, but the model/setup and error shape differ enough that the underlying bug is not clearly the same." }, { "left": "issue:43596", "right": "issue:44805", "accept": false, - "reason": "Different error shape/context: Bert ZeRO-3 index-0 OOB vs mask/tensor shape mismatch." + "reason": "Zero-3+BERT index-out-of-bounds vs mask/tensor shape mismatch are different concrete failures." }, { "left": "issue:44805", "right": "issue:45161", "accept": false, - "reason": "Different model and failure mode: tensor/mask shape mismatch vs GPT-OSS tensor-parallel issue." + "reason": "Shape-mismatch indexing error and GPT-OSS tensor-parallel issue are unrelated." }, { "left": "issue:44805", "right": "issue:44918", "accept": false, - "reason": "Different bugs: mask shape mismatch vs Qwen3.5 embedding unpacking with TRL." + "reason": "Different failure signatures and code paths; not the same bug." }, { "left": "issue:44661", "right": "issue:44805", "accept": false, - "reason": "Tokenizer-mapping model addition bug is unrelated to the mask/tensor shape error." + "reason": "Tokenizer-mapping add-new-model-like failure is unrelated to the mask/tensor shape mismatch." }, { "left": "issue:34634", "right": "issue:35545", "accept": false, - "reason": "BarkProcessor voice_preset and ModernBERT ONNX export are unrelated." + "reason": "BarkProcessor voice_preset issue is unrelated to ModernBERT ONNX export." }, { "left": "issue:30277", "right": "issue:43638", "accept": false, - "reason": "Both involve DeepSpeed, but collective mismatch and Bert index-out-of-bounds are distinct failures." + "reason": "Both involve DeepSpeed ZeRO-3, but the collective mismatch and BERT index error are different bugs." }, { "left": "issue:43596", "right": "issue:43638", "accept": true, - "reason": "Same DeepSpeed ZeRO-3 + Bert index-0 out-of-bounds error on the same code path; these look like duplicate reports." + "reason": "Near-identical DeepSpeed ZeRO-3 BERT out-of-bounds reports with the same error pattern and same underlying code path." }, { "left": "issue:43975", "right": "issue:44457", "accept": false, - "reason": "Different model/task issues: detokenization bug vs LoRA merge/save/reload output mismatch." + "reason": "Detokenization regression and LoRA merge/reload inconsistency are separate problems." }, { "left": "issue:43039", "right": "issue:45161", "accept": false, - "reason": "Liger kernel cross-entropy path and GPT-OSS tensor-parallel failure are not the same bug." + "reason": "Liger Kernel cross-entropy behavior is unrelated to GPT-OSS tensor-parallel failure." }, { "left": "issue:44292", "right": "issue:44457", "accept": false, - "reason": "Different models and different failure modes: Qwen-3 NVFP4 runtime error vs LoRA weight merge/save inconsistency." + "reason": "Qwen NVFP4 runtime error and LoRA merge/save mismatch are unrelated." }, { "left": "issue:34928", "right": "issue:44805", "accept": false, - "reason": "Activation-checkpointing/FSDP recomputed size mismatch is unrelated to the mask shape error." + "reason": "FSDP checkpointing recompute mismatch is not the same as the mask/tensor indexing error." }, { "left": "issue:24643", "right": "issue:30277", "accept": false, - "reason": "Both are DeepSpeed training issues, but the concrete errors and triggers differ." + "reason": "Both mention DeepSpeed, but the weight-dimension error and collective mismatch are distinct failures." }, { "left": "issue:43975", "right": "issue:45237", "accept": false, - "reason": "Different models and domains: detokenization vs AMD GPU runtime failure." + "reason": "Detokenization bug vs GPU backend support issue; unrelated." }, { "left": "issue:43638", "right": "issue:45161", "accept": false, - "reason": "Bert ZeRO-3 index error and GPT-OSS tensor-parallel issue are unrelated." + "reason": "BERT ZeRO-3 indexing bug and GPT-OSS tensor-parallel bug are different issues." }, { "left": "issue:34928", "right": "issue:36331", "accept": false, - "reason": "Activation checkpointing/FSDP size mismatch is unrelated to the custom trainer loss signature change." + "reason": "Activation-checkpointing/FSDP issue is unrelated to CustomTrainer compute_loss signature regression." }, { "left": "issue:43596", "right": "issue:45161", "accept": false, - "reason": "Different models and code paths; same distributed-training theme is too broad for duplicate triage." + "reason": "Different models and different failure classes." }, { "left": "issue:43716", - "right": "issue:45507", + "right": "issue:44292", "accept": false, - "reason": "Mistral-3 preprocessing dtype mismatch and GraniteMoEHybrid invalid method are unrelated." + "reason": "Mistral-3 dtype mismatch and Qwen NVFP4 runtime error do not share the same bug." }, { "left": "issue:39290", "right": "issue:45081", "accept": false, - "reason": "Gemma3 vLLM config attribute error and Mistral tokenizer patch crash are different bugs." + "reason": "Gemma3 vLLM config attribute error and Mistral tokenizer regex patch crash are unrelated." }, { "left": "issue:43828", "right": "issue:45161", "accept": false, - "reason": "Autocast dtype mismatch in Phi-tiny-MoE is unrelated to GPT-OSS tensor-parallel failure." + "reason": "Autocast dtype mismatch and GPT-OSS TP failure are not the same underlying problem." }, { "left": "issue:30277", "right": "issue:36331", "accept": false, - "reason": "Different failures: collectives mismatch in Deepspeed vs trainer API change." + "reason": "DeepSpeed collective mismatch and trainer method signature change are unrelated." }, { "left": "issue:43039", "right": "issue:44805", "accept": false, - "reason": "Liger kernel cross-entropy usage is unrelated to the mask/tensor shape mismatch." + "reason": "Liger Kernel cross-entropy path is unrelated to the indexing shape mismatch." }, { "left": "issue:34928", "right": "issue:36010", "accept": false, - "reason": "Activation checkpointing/FSDP bug and GenerationMixin import error are unrelated." + "reason": "FSDP activation checkpointing bug and GenerationMixin import error are unrelated." }, { "left": "issue:41762", "right": "issue:43596", "accept": false, - "reason": "Both are DeepSpeed-related, but Qwen/Gemma-style loading issues are not the same as the Bert ZeRO-3 index bug." + "reason": "Both are ZeRO-3 index errors, but they involve different model families and enough context difference to avoid merging." }, { "left": "issue:43901", "right": "issue:44512", "accept": false, - "reason": "Both are docs issues, but they cover different removed APIs and should not be merged as duplicates." + "reason": "Both are docs regressions, but they concern different removed APIs and different documentation targets." }, { "left": "issue:39290", "right": "issue:41720", "accept": false, - "reason": "Gemma3TextConfig missing sliding_window_pattern and Qwen3 cuda assert are different failures." + "reason": "Gemma3 vLLM config regression and Qwen3 auto-device CUDA assert are unrelated." }, { - "left": "issue:43866", - "right": "issue:44863", + "left": "issue:43039", + "right": "issue:44560", "accept": false, - "reason": "Checkpoint corruption and NemotronH loading failure are different underlying problems." + "reason": "Liger Kernel cross-entropy issue and Qwen3-vl video StopIteration are unrelated." }, { - "left": "issue:43039", - "right": "issue:44560", + "left": "issue:43866", + "right": "issue:44863", "accept": false, - "reason": "Different issues: Liger kernel cross-entropy call vs Qwen3-vl video StopIteration." + "reason": "Corrupted checkpoint report and NemotronH loading failure are not the same bug." }, { "left": "issue:43854", "right": "issue:43866", "accept": false, - "reason": "Unit-test model load failure and checkpoint corruption are not the same bug." + "reason": "Unit-test loading failure vs corrupted checkpoint; too different to duplicate." }, { "left": "issue:43425", "right": "issue:45070", "accept": false, - "reason": "Torch-version incompatibility and pydantic PretrainedConfig breakage are unrelated." + "reason": "Torch version incompatibility and pydantic PretrainedConfig field breakage are unrelated." }, { "left": "issue:39290", "right": "issue:43531", "accept": false, - "reason": "Different Qwen3/Gemma sliding-window problems with different configs and symptoms." + "reason": "Different Qwen/Gemma sliding-window issues and different configs; not clearly the same bug." }, { "left": "issue:41093", "right": "issue:43638", "accept": false, - "reason": "Mask shape/index tensor mismatch is unrelated to Bert ZeRO-3 index-out-of-bounds." + "reason": "Both are indexing/shape errors, but they occur in different model/setup contexts and are not clearly duplicates." }, { "left": "issue:43404", "right": "issue:43866", "accept": false, - "reason": "Mistral3 lm_head tying bug and Ovis2 checkpoint corruption are unrelated." + "reason": "Mistral3 weight-tie bug and Ovis2 corrupted checkpoint are unrelated." }, { "left": "issue:45362", "right": "issue:45464", "accept": false, - "reason": "Both are Qwen3.5, but one is transformers chat crash and the other is streaming chat/completions failure." + "reason": "Qwen3.5 chat crash and streaming API failure are different symptoms with different likely causes." }, { "left": "issue:43792", "right": "issue:45070", "accept": false, - "reason": "Whisper runtime failure and pydantic model breakage are unrelated." + "reason": "Whisper load failure and pydantic model regression are unrelated." }, { "left": "issue:43541", "right": "issue:43828", "accept": false, - "reason": "Mixtral grouped_mm tracing error and Phi-tiny-MoE autocast mismatch are different bugs." + "reason": "Grouped_mm tracing error on Mixtral and autocast dtype mismatch on Phi-tiny-MoE are different code paths." }, { "left": "issue:36010", "right": "issue:39290", "accept": false, - "reason": "GenerationMixin import error and Gemma3 sliding-window config issue are unrelated." + "reason": "GenerationMixin import regression and Gemma3 sliding-window config error are unrelated." }, { "left": "issue:44841", "right": "issue:45084", "accept": false, - "reason": "Voxtral processor failure and template-node compilation error are different code paths." + "reason": "Voxtral processor failure and template-node compile error are unrelated." }, { - "left": "issue:39290", - "right": "issue:41720", + "left": "issue:41093", + "right": "issue:41762", "accept": false, - "reason": "Gemma3 vLLM config error and Qwen3 auto device mapping cuda assert are unrelated." + "reason": "Both are mask/index errors, but they are on different models and not clearly the same root cause." }, { "left": "issue:43296", "right": "issue:43366", "accept": false, - "reason": "PaddleOCR-VL load failure and GGUF gpt-oss support request are different issues." + "reason": "PaddleOCR-VL loading failure and GGUF gpt-oss support request are unrelated." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is grounded in the packet, and the lone accepted duplicate pair (issue:43596 / issue:43638) is supported by the titles and the high similarity score. The other soft-edge verdicts are conservatively rejected rather than overstated. No PR-pair concerns apply because there are no PRs in the packet." + "feedback": "The summary is grounded in the packet and stays conservative overall. The only accepted soft pair, issue:43596 vs issue:43638, is supported by very similar DeepSpeed ZeRO-3 + index-out-of-bounds reports. The remaining rejections are generally cautious and do not overclaim duplication from broad similarity alone." }, "evaluator_used": true, "retried": false, @@ -642,12 +642,12 @@ "issue:36331|issue:41762", "issue:44560|issue:45507" ], - "prepared_review_unit_hash": "da7f744214f9226c785285d5928d31b8e78187e6a1702bc70686eaba85ed9e29", + "prepared_review_unit_hash": "8ef8a5bf83a138b19ad11e4fbf89972cf8ba6404e8221ff3813f09ee968e3c06", "component_budget": { "node_count": 367, "item_count": 367, "soft_pair_count": 1370, - "serialized_chars": 706987, + "serialized_chars": 706986, "estimated_input_tokens": 176747, "estimated_eval_tokens": 353750 }, @@ -664,225 +664,225 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Most soft similarities are too weak to treat as duplicates from the packet alone. The only clearly duplicate-like chain is the non-persistent-buffer regression cluster around issue:43950; the SigLIP2 pair is not strong enough to merge conservatively.", - "confidence": 0.78, - "canonical_issue_reason": "issue:43950 is the clearest representative of the non-persistent-buffer regression: it directly names the v5 loading corruption and is the most specific statement of the bug in this group.", + "summary": "The cluster is mostly heterogeneous, but it contains two clear duplicate subclusters: non-persistent buffer corruption in Transformers v5, and a SigLIP2 output-quality regression. Most other soft edges are only broad topical similarity and should be rejected.", + "confidence": 0.82, + "canonical_issue_reason": "issue:43950 is the clearest representative of the non-persistent-buffer regression and has direct duplicates in the set (issue:43644, issue:44534).", "canonical_pr_reason": null, - "best_issue_reason": "issue:43950 is the most actionable standalone report here: it is specific, clearly described, and captures the core regression without overgeneralizing.", + "best_issue_reason": "issue:43950 is the most actionable and well-scoped issue here: concrete regression, clear symptom, and it anchors the strongest duplicate cluster.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43572", "right": "issue:45070", "accept": false, - "reason": "Both are v5 config-related regressions, but the concrete failures differ: missing StableLmConfig field vs a PretrainedConfig/pydantic field breakage." + "reason": "Different bugs: one is a missing config field, the other is a pydantic/PretrainedConfig regression." }, { "left": "issue:44387", "right": "issue:45005", "accept": false, - "reason": "Both touch v5 and model weights, but one is an int4 memory/OOM regression while the other is a tied-weights issue on translation models." + "reason": "Both mention v5 regressions, but the symptoms and code paths are unrelated (int4 OOM vs tied weights)." }, { "left": "issue:43541", "right": "issue:45161", "accept": false, - "reason": "Both involve MoE models, but the failures are different: torch dynamo grouped_mm tracing vs tensor-parallel support not working." + "reason": "Different MoE failures: torch dynamo grouped_mm tracing error vs tensor-parallel support issue." }, { "left": "issue:43828", "right": "issue:45464", "accept": false, - "reason": "Different models and different failure modes: autocast dtype mismatch in Phi-tiny-MoE versus streaming chat/completions failure on Qwen3.5." + "reason": "Unrelated runtime failures: autocast dtype mismatch vs streaming inference API failure." }, { "left": "issue:44291", "right": "issue:45070", "accept": false, - "reason": "init_empty_weights/_is_hf_initialized is a model-loading API break; the other report is a pydantic/PretrainedConfig regression." + "reason": "One is init_empty_weights/_is_hf_initialized, the other is a pydantic model field breakage." }, { "left": "issue:43366", "right": "issue:45464", "accept": false, - "reason": "GGUF support for gpt-oss and Qwen3.5 streaming inference are unrelated problems." + "reason": "Different domains: GGUF architecture support vs Qwen3.5 streaming inference." }, { "left": "issue:43531", "right": "issue:45070", "accept": false, - "reason": "Qwen3-MoE sliding-window handling is unrelated to the PretrainedConfig/pydantic break." + "reason": "Qwen3-MoE sliding_window behavior is unrelated to PretrainedConfig/pydantic issues." }, { "left": "issue:44661", "right": "issue:45464", "accept": false, - "reason": "Tokenizer mapping/add-new-model-like failure is a different code path from the Qwen3.5 streaming issue." + "reason": "Tokenizer-mapping model addition bug is unrelated to streaming inference." }, { "left": "issue:39290", "right": "issue:43742", "accept": false, - "reason": "Both are loading/config errors, but they concern different models and different missing attributes/key errors." + "reason": "Different model-specific loading errors with no clear shared code path." }, { "left": "issue:43054", "right": "issue:43994", - "accept": false, - "reason": "Both are about SigLIP2, but the packet only shows a docs/example plus quantization-failure report versus a broader bad-output/embedding-quality report; that is not enough to call the same bug." + "accept": true, + "reason": "Both describe SigLIP2 producing bad/nonsensical outputs, likely the same model regression." }, { "left": "issue:43541", "right": "issue:44589", "accept": false, - "reason": "Grouped_mm tracing on Mixtral and missing Float8 storage are unrelated runtime failures." + "reason": "Different failures: grouped_mm tracing vs missing Float8 storage type." }, { "left": "issue:41093", "right": "issue:43596", "accept": false, - "reason": "Both are index/shape errors, but one is a mask-length mismatch and the other is a DeepSpeed ZeRO-3 init/index-out-of-bounds problem." + "reason": "Both are shape/index errors, but the model/setup and likely root cause differ." }, { "left": "issue:42915", "right": "issue:43541", "accept": false, - "reason": "Different MoE-related failures: FineGrainedFP8Config vs torch dynamo grouped_mm tracing." + "reason": "Different MoE-specific problems and different failure modes." }, { "left": "issue:40990", "right": "issue:44960", "accept": false, - "reason": "Perplexity on gpt-oss-20b and a GLM5 issue are not the same bug." + "reason": "No meaningful overlap beyond being model-quality/perplexity complaints." }, { "left": "issue:39186", "right": "issue:44918", "accept": false, - "reason": "FSDP weight-dimension error and TRL unpacking failure are different training/runtime issues." + "reason": "Different training/runtime failures: FSDP weight-shape error vs TRL unpacking embeddings." }, { "left": "issue:43404", "right": "issue:45005", "accept": false, - "reason": "Both mention tied weights, but one is a specific Mistral3 lm_head tying bug and the other is a broader translation-model v5 report." + "reason": "Related theme of tied weights, but the model scope and failure mode are not clearly the same bug." }, { "left": "issue:43645", "right": "issue:45399", "accept": false, - "reason": "Notebook custom-model initialization and flash-attn2 fallback checks are unrelated." + "reason": "Notebook/custom model init issue is unrelated to flash-attn fallback checks." }, { "left": "issue:44530", "right": "issue:45005", "accept": false, - "reason": "PagedAttentionCache/Qwen3.5 linear_attention crash is unrelated to tied-weight handling." + "reason": "PagedAttentionCache/group type crash is unrelated to tied-weight regressions." }, { "left": "issue:44863", "right": "issue:45005", "accept": false, - "reason": "NemotronH checkpoint loading is a different model-loading bug from tied weights." + "reason": "NemotronH checkpoint loading is a different model-loading problem." }, { "left": "issue:43873", "right": "issue:45005", "accept": false, - "reason": "Quantization/offloading behavior does not match the tied-weights regression." + "reason": "Quantization/offloading behavior is unrelated to tied weights." }, { "left": "issue:44360", "right": "issue:44512", "accept": false, - "reason": "A DSA indexer ReLU discussion is unrelated to a docs command reference issue." + "reason": "Completely different issues: DSA indexer internals vs outdated docs." }, { "left": "issue:42915", "right": "issue:45399", "accept": false, - "reason": "FineGrainedFP8Config on Qwen3MoE and flash-attn2 fallback blocking are different subsystems and symptoms." + "reason": "No shared concrete bug; MoE FP8 failure vs flash-attn fallback gating." }, { "left": "issue:44918", "right": "issue:45507", "accept": false, - "reason": "TRL SFT unpacking on Qwen3.5 is unrelated to GraniteMoEHybrid calling an invalid method." + "reason": "Different model families and unrelated runtime errors." }, { "left": "issue:44492", "right": "issue:44512", "accept": false, - "reason": "A cache-strategy typo and removed run-command docs are not the same issue." + "reason": "A typo in cache strategies is not the same as stale CLI docs." }, { "left": "issue:43975", "right": "issue:44292", "accept": false, - "reason": "DeepSeek detokenization and Qwen-3-8B-NVFP4 runtime errors are different model-specific bugs." + "reason": "Different inference bugs with unrelated symptoms and models." }, { "left": "issue:43950", "right": "issue:44534", "accept": true, - "reason": "These describe the same v5 regression where non-persistent buffers are corrupted or filled with junk during loading." + "reason": "Both report the same non-persistent-buffer corruption regression in Transformers v5." }, { "left": "issue:43644", "right": "issue:43950", "accept": true, - "reason": "Same underlying non-persistent-buffer regression: one report states the corruption directly and the other describes the same junk-filled buffer behavior." + "reason": "Same bug: v5 corrupts non-persistent buffers during load/init." }, { "left": "issue:44292", "right": "issue:45507", "accept": false, - "reason": "Qwen-3-8B-NVFP4 runtime failure and GraniteMoEHybrid invalid-method calls are unrelated." + "reason": "Qwen NVFP4 runtime error and GraniteMoE invalid method are unrelated." }, { "left": "issue:43828", "right": "issue:44928", "accept": false, - "reason": "Both mention Qwen/MoE-style runtime problems, but autocast dtype mismatch is not the same as the 3D position_ids/BF16-collapse report." + "reason": "Both involve dtype/numerics, but the actual failures and code paths differ." }, { "left": "issue:39692", "right": "issue:43054", "accept": false, - "reason": "The packet does not establish the same bug: one is a docs/example plus quantization-failure report, the other is a broader poor-embedding result report." + "reason": "Both mention SigLIP2, but docs example errors and embedding-quality regression are different problems." }, { "left": "issue:41553", "right": "issue:43883", "accept": false, - "reason": "AutoTokenizer loading error for Voxtral and missing all_tied_weights_keys on MolmoForCausalLM are different loader/model-class bugs." + "reason": "Bad tokenizer error message vs missing all_tied_weights_keys are unrelated." }, { "left": "issue:43866", "right": "issue:44403", "accept": false, - "reason": "Ovis2 checkpoint corruption and generic loading noise are unrelated." + "reason": "Checkpoint corruption and loading noise are not the same bug." }, { "left": "issue:36331", "right": "issue:41762", "accept": false, - "reason": "Trainer compute_loss kwarg mismatch and DeepSpeed ZeRO-3 index-out-of-bounds are unrelated training bugs." + "reason": "Different training/loading failures with no shared root cause." }, { "left": "issue:44560", "right": "issue:45507", "accept": false, - "reason": "Qwen3-vl embedding video StopIteration and GraniteMoEHybrid invalid-method calls are different failures." + "reason": "Qwen3-vl video StopIteration and GraniteMoE invalid method are unrelated." } ] }, "evaluator_result": { - "accept": false, - "feedback": "The summary is slightly overstated. The non-persistent-buffer pair (43950/44534 and 43644/43950) is well supported by the titles, but the SigLIP2 pair (43054/43994) is not clearly the same bug from the packet alone: one is a docs/example + quantization-failure report, the other is a broader bad-output/embedding-quality report. Also, the canonical/best-issue wording leans on a duplicate linkage that is not established in the packet. Please keep the soft-edge verdicts more conservative." + "accept": true, + "feedback": "Grounded in the packet and appropriately conservative. The two supported duplicate clusters are the non-persistent-buffer regression (43950/43644/44534) and the SigLIP2 bad-output pair (43054/43994). The other soft-edge pairs are rejected with cautious, issue-title-based reasoning rather than overclaiming evidence." }, "evaluator_used": true, - "retried": true, + "retried": false, "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null @@ -987,12 +987,12 @@ "issue:42915|issue:43872", "issue:36331|issue:41093" ], - "prepared_review_unit_hash": "aad49452eeb67c998348888deac397242ad1f42295b7ce51458d3f8635e85508", + "prepared_review_unit_hash": "7204b8a2d4879c591ed731bf4a7cc7098cdfe1756b9cc484c1c7a3834e3f68bd", "component_budget": { "node_count": 367, "item_count": 367, "soft_pair_count": 1370, - "serialized_chars": 706987, + "serialized_chars": 706986, "estimated_input_tokens": 176747, "estimated_eval_tokens": 353750 }, @@ -1009,228 +1009,222 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Mostly unrelated issue set with a few clear duplicates in the same sub-bug families: SigLIP2 quality regression, docs regressions around removed pipeline tasks, Qwen3 ZeRO-3 weight loading, and the _is_hf_initialized / init_empty_weights loading bug.", - "confidence": 0.77, - "canonical_issue_reason": "Issue 45020 is the broadest representative of the set: it captures a current, user-facing loading regression class around remote_code, and is more general than the model-specific symptom reports.", + "summary": "The set is mostly unrelated issues, but there are two clear duplicate pairs: the transformers 5.x `_is_hf_initialized` loading TypeError, and the docs regression about removed pipeline tasks still being referenced. Everything else looks too broad or symptom-level to merge as duplicates.", + "confidence": 0.78, + "canonical_issue_reason": "issue:44291 is the clearest root-cause report for the `_is_hf_initialized` TypeError during `init_empty_weights` loading, and it covers the broader failure mode better than the narrower bitsandbytes-specific report in issue:43872.", "canonical_pr_reason": null, - "best_issue_reason": "45020 is the best overall issue candidate because it is open, broadly scoped, and describes a concrete regression class rather than a narrow one-off model failure.", + "best_issue_reason": "issue:44291 is the strongest representative issue overall: it has the most explicit technical cause, a clear reproduction path, and spans the general loading regression rather than one library-specific manifestation.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43872", "right": "issue:44589", "accept": false, - "reason": "Both are loading/type errors, but they point to different underlying failures: an unexpected _is_hf_initialized argument vs a missing Float8 storage type." + "reason": "Different failures: bitsandbytes `_is_hf_initialized` construction error vs a float8 storage lookup error." }, { "left": "issue:41553", "right": "issue:43881", "accept": false, - "reason": "Different models and different failure modes; one is a bad error message, the other is a load failure." + "reason": "Different models and different failure modes; only loosely related around model loading." }, { "left": "issue:38175", "right": "issue:43054", - "accept": true, - "reason": "Both describe SigLIP2 output regressions and are consistent with the same underlying model implementation discrepancy." + "accept": false, + "reason": "Both mention SigLIP2, but the reported symptoms are different enough that they do not look like the same bug." }, { "left": "issue:42947", "right": "issue:44387", "accept": false, - "reason": "Both are training/memory complaints, but one is LoRA gradient checkpointing and the other is int4 quantization OOM; not the same bug." + "reason": "Unrelated regressions: LoRA gradient checkpointing vs int4 quantization memory growth." }, { "left": "issue:43054", "right": "issue:43493", - "accept": true, - "reason": "Same SigLIP2 regression family; 43493 is the broader implementation discrepancy, and 43054 is a symptom of it." + "accept": false, + "reason": "Related to SigLIP2 quality/discrepancy, but not enough evidence they are the same concrete bug." }, { "left": "issue:43881", "right": "issue:43883", "accept": false, - "reason": "Different model-specific loading bugs with no clear shared code-path." + "reason": "Different architectures and different errors; no shared code-path is evident." }, { "left": "issue:44403", "right": "issue:44451", "accept": false, - "reason": "Generic loading noise vs a specific model load failure; not the same underlying issue." + "reason": "Generic loading noise vs a specific inability to load a particular model." }, { "left": "issue:42947", "right": "issue:43856", "accept": false, - "reason": "Both mention training/memory, but they involve different subsystems and failure mechanisms." + "reason": "Different training concerns: ineffective checkpointing vs MoE memory usage." }, { "left": "issue:43975", "right": "issue:45520", "accept": false, - "reason": "Detokenization regression and flash_attn import lookup are unrelated." + "reason": "Tokenizer detokenization bug and a Python 3.13 flash-attn import key error are unrelated." }, { "left": "issue:43643", "right": "issue:45020", "accept": false, - "reason": "Both touch remote_code, but one is missing fields from AutoConfig and the other is a broader loading regression; not enough to treat as the same bug." + "reason": "Both involve remote-code loading, but one is a specific AutoConfig field-loss bug and the other is a broader regression claim." }, { "left": "issue:44403", "right": "issue:44488", "accept": false, - "reason": "Both are model loading complaints, but about different models and different root causes." + "reason": "Same broad area of model loading, but different models and no concrete shared bug." }, { "left": "issue:43782", "right": "issue:44960", "accept": false, - "reason": "Different model families and different errors." + "reason": "Different models and different load-time failures; no direct duplicate evidence." }, { "left": "issue:42491", "right": "issue:44368", "accept": false, - "reason": "Both involve Qwen3/MoE configuration warnings or incompatibilities, but they are not the same concrete failure." + "reason": "LoRA compatibility across versions vs a tie_word_embeddings warning; not the same issue." }, { "left": "issue:41762", "right": "issue:42915", "accept": false, - "reason": "Both are DeepSpeed-related, but one is Gemma3 ZeRO-3 loading and the other is Qwen3 MoE weight conversion; different bugs." + "reason": "Different model families and different DeepSpeed/FP8 failure modes." }, { "left": "issue:43975", "right": "issue:44451", "accept": false, - "reason": "Detokenization bug vs a model loading failure; unrelated." + "reason": "Tokenizer detokenization regression vs model loading failure; not the same bug." }, { "left": "issue:45020", "right": "issue:45356", "accept": false, - "reason": "Remote_code loading regressions and a Kimi tokenizer regression are different code paths." + "reason": "Both involve recent regressions, but one is general remote-code breakage and the other is a tokenizer codec/fix_mistral_regex issue." }, { "left": "issue:44661", "right": "issue:45020", "accept": false, - "reason": "Tokenizer mapping edge case vs general remote_code loading regressions; not the same bug." + "reason": "Different problems: tokenizer mapping guardrail vs remote-code loading regressions." }, { "left": "issue:43054", "right": "issue:43278", "accept": false, - "reason": "Both mention embeddings, but one is a quality regression and the other is a dtype change between train/eval." + "reason": "Both mention embeddings, but one is model quality and the other is dtype drift between train/eval." }, { "left": "issue:43824", "right": "issue:43872", "accept": false, - "reason": "Import error for a VL class and bitsandbytes _is_hf_initialized incompatibility are unrelated." + "reason": "ImportError for a missing class vs the `_is_hf_initialized` construction TypeError." }, { "left": "issue:44991", "right": "issue:45399", "accept": false, - "reason": "Tokenizer loading failure and flash-attn fallback gating are unrelated." + "reason": "Tokenizer loading failure vs flash-attn fallback gating; unrelated code paths." }, { "left": "issue:43827", "right": "issue:44509", "accept": true, - "reason": "Same docs regression: both say removed pipeline tasks are still documented after v5 removals." + "reason": "Same docs regression: both say v5 docs still reference removed pipeline tasks like summarization/translation/text2text-generation." }, { "left": "issue:43792", "right": "issue:44220", - "accept": true, - "reason": "Likely the same audio feature-extraction bug: the Whisper model failure is plausibly caused by _torch_extract_fbank_features()." + "accept": false, + "reason": "Possible Whisper-related overlap, but the titles are too vague to conclude the same underlying bug." }, { "left": "issue:45081", "right": "issue:45399", "accept": false, - "reason": "Mistral tokenizer regex crash and flash-attn fallback gating are different problems." + "reason": "Tokenizer regex crash vs flash-attn fallback blocking are different failure paths." }, { "left": "issue:43404", "right": "issue:43950", "accept": false, - "reason": "lm_head tying bug and non-persistent buffer corruption are distinct initialization/loading regressions." + "reason": "Tied lm_head weights and non-persistent buffer corruption are different regressions." }, { "left": "issue:43257", "right": "issue:43866", "accept": false, - "reason": "Different model families and different loading/data integrity problems." + "reason": "Different Qwen3 MoE issues: weight conversion/loading vs a corrupted checkpoint." }, { "left": "issue:42915", "right": "issue:45020", "accept": false, - "reason": "Qwen3 MoE weight conversion under accelerate+deepspeed is not the same as generic remote_code loading breakage." + "reason": "Different symptoms and different model families; only broad version/regression context overlaps." }, { "left": "issue:43742", "right": "issue:45399", "accept": false, - "reason": "Key error while loading a specific model and flash-attn fallback checks are unrelated." - }, - { - "left": "issue:43742", - "right": "issue:45020", - "accept": false, - "reason": "A model-specific key error is not the same as the broader remote_code regression." + "reason": "Model-specific key error vs flash-attn fallback gating; not the same bug." }, { "left": "issue:41553", "right": "issue:44488", "accept": false, - "reason": "Different models and different load errors." + "reason": "Different models and different loading outcomes." }, { "left": "issue:43940", "right": "issue:45313", - "accept": true, - "reason": "Both report Qwen3-family ZeRO-3 weight-loading failures; same underlying DeepSpeed loading path is likely broken." + "accept": false, + "reason": "Both are ZeRO-3 load failures, but for different Qwen model variants and different affected tensors." }, { "left": "issue:43872", "right": "issue:44291", "accept": true, - "reason": "Same unexpected _is_hf_initialized argument regression during empty-weight/model initialization." + "reason": "Same `_is_hf_initialized` TypeError during `init_empty_weights`/model loading; 43872 is a bitsandbytes-specific manifestation of the same incompatibility." }, { "left": "issue:43854", "right": "issue:44863", "accept": false, - "reason": "Different model load failures with no clear shared root cause." + "reason": "Both are model loading failures, but they concern different models with no shared code path evident." }, { "left": "issue:43257", "right": "issue:45305", "accept": false, - "reason": "Qwen3 MoE weight conversion failure and GAS gradient averaging regression are separate DeepSpeed issues." + "reason": "Different DeepSpeed/Qwen training regressions with different symptoms and causes." }, { "left": "issue:41553", "right": "issue:44488", "accept": false, - "reason": "Both are loading-related, but they concern different models and different symptoms." + "reason": "Different model-loading reports without a concrete shared bug." }, { "left": "issue:36331", "right": "issue:41093", "accept": false, - "reason": "Trainer signature mismatch and mask/tensor shape mismatch are unrelated errors." + "reason": "Unrelated TypeErrors/IndexErrors; the shared surface of an exception is too generic to merge." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is grounded in the packet, and the soft-edge verdicts stay conservative overall. The accepted pairs are plausibly same-bug families based on their titles/body context, while the rejections avoid over-connecting superficially similar but distinct failures. No clear overstatement stands out." + "feedback": "The summary is grounded in the packet, and the two accepted duplicate pairs are well supported by the issue titles. The soft-edge verdicts are generally conservative and avoid over-merging symptom-level similarities." }, "evaluator_used": true, "retried": false, @@ -1302,8 +1296,8 @@ ], "soft_pairs": [ "issue:41093|issue:44805", - "issue:44479|issue:44560", "issue:43872|issue:45538", + "issue:44479|issue:44560", "issue:44530|issue:44863", "issue:43366|issue:43828", "issue:44155|issue:45584", @@ -1332,12 +1326,12 @@ "issue:44960|issue:44977", "issue:44451|issue:44488" ], - "prepared_review_unit_hash": "7045a11735340c7f06cf4a00a94c1575ea90f6c824a96e126016cd4cfa510a59", + "prepared_review_unit_hash": "3bca589c0d67b30538434f8d1195859e9608541fc6ea0c3d1a847f7188cbcd34", "component_budget": { "node_count": 367, "item_count": 367, "soft_pair_count": 1370, - "serialized_chars": 706987, + "serialized_chars": 706986, "estimated_input_tokens": 176747, "estimated_eval_tokens": 353750 }, @@ -1345,210 +1339,210 @@ "node_count": 48, "item_count": 48, "soft_pair_count": 30, - "serialized_chars": 36658, - "estimated_input_tokens": 9165, - "estimated_eval_tokens": 18586 + "serialized_chars": 36654, + "estimated_input_tokens": 9164, + "estimated_eval_tokens": 18584 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Mostly a heterogeneous set of transformer/model-loading bugs with a few clear duplicate pairs. Only a small number of soft edges look like the same underlying issue; most others are related only at a broad subsystem level and should stay separate.", - "confidence": 0.77, - "canonical_issue_reason": "For the clearly duplicated non-persistent-buffer regression, issue #44534 is the cleanest representative: it states the bug plainly and matches the earlier report #43644.", + "summary": "A mostly loose cluster with only two convincing duplicate pairs: the Transformers 5.x non-persistent-buffer regression (43644\u219444534) and the custom-model initialization regression (43645\u219443646). The other soft links are mostly title/model-family similarity without enough shared body evidence, so they should stay separate.", + "confidence": 0.82, + "canonical_issue_reason": "issue:43644 is the cleanest canonical anchor: it describes the Transformers 5.0.0 non-persistent-buffer corruption crisply and has a near-exact duplicate in 44534.", "canonical_pr_reason": null, - "best_issue_reason": "Issue #43950 is the strongest standalone report overall because it most precisely describes the from_pretrained corruption bug and provides the most actionable framing of the regression.", + "best_issue_reason": "issue:43644 is the strongest overall issue candidate because it is narrowly scoped, clearly reproducible, and backed by the closest duplicate match in the packet.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:41093", "right": "issue:44805", - "accept": true, - "reason": "Same IndexError mask/tensor shape mismatch symptom; the titles are near-identical and point to the same off-by-one-style bug." - }, - { - "left": "issue:44479", - "right": "issue:44560", "accept": false, - "reason": "Both are Qwen video-related, but one is a v5.3.0 input regression and the other is a StopIteration during embedding handling; not the same concrete bug." + "reason": "Same generic IndexError wording, but the packet does not provide enough shared body evidence to prove the same underlying bug." }, { "left": "issue:43872", "right": "issue:45538", "accept": false, - "reason": "Unrelated failures: bitsandbytes Int8Params constructor incompatibility vs CLIPTokenizer max_length behavior." + "reason": "Unrelated failures: bitsandbytes initialization incompatibility vs tokenizer max-length behavior." + }, + { + "left": "issue:44479", + "right": "issue:44560", + "accept": false, + "reason": "Both are Qwen video-related, but the reported failures differ and the packet lacks evidence they are the same code-path bug." }, { "left": "issue:44530", "right": "issue:44863", "accept": false, - "reason": "Different code paths and symptoms: PagedAttentionCache linear_attention crash vs NemotronH checkpoint loading failure." + "reason": "Different problems: PagedAttentionCache linear_attention crash vs NemotronH checkpoint loading." }, { "left": "issue:43366", "right": "issue:43828", "accept": false, - "reason": "Different model families and different failure modes; shared broad loading context is not enough." + "reason": "Broad model-family overlap only; gpt-oss GGUF support and Phi-tiny-MoE dtype mismatch are different issues." }, { "left": "issue:44155", "right": "issue:45584", "accept": false, - "reason": "AudioFlamingo3 batch-leak bug and Whisper empty-transcription failure are unrelated code paths." + "reason": "Different models and failure modes; no evidence they share the same underlying bug." }, { "left": "issue:24643", "right": "issue:30064", "accept": false, - "reason": "Completely different subsystems: DeepSpeed training weight shape error vs image processor void segmentation handling." + "reason": "Completely separate domains: DeepSpeed training weight-shape error vs image processor void segmentation map failure." }, { "left": "issue:43366", "right": "issue:43531", "accept": false, - "reason": "Both mention Qwen3-MoE/GGUF-ish areas, but one is architecture support and the other is sliding_window behavior; not the same bug." + "reason": "Same ecosystem, but support for GGUF gpt-oss and a Qwen3-MoE sliding_window bug are not the same change." }, { "left": "issue:43550", "right": "issue:44910", "accept": false, - "reason": "Different models and distinct failures: Bamba torch.compile/SDPA vs Qwen3.5 flash-attention position_ids crash." + "reason": "Different concrete failures: Bamba torch.compile/SDPA vs Qwen3.5 flash-attention position_ids handling." }, { "left": "issue:43645", "right": "issue:43646", - "accept": false, - "reason": "Both are custom-model-init regressions, but one is Jupyter-specific and the other is generic initialization; not enough to merge as one bug." + "accept": true, + "reason": "These describe the same custom-model initialization regression; one is a notebook-specific manifestation of the broader failure." }, { "left": "issue:43950", "right": "issue:45325", "accept": false, - "reason": "Non-persistent-buffer corruption and Qwen2.5-VL rope scaling are unrelated issues." + "reason": "Both are Transformers regressions, but one is non-persistent-buffer corruption and the other is Qwen2.5-VL temporal position_id scaling." }, { "left": "issue:43366", "right": "issue:44589", "accept": false, - "reason": "GGUF gpt-oss support request is unrelated to Float8 storage lookup failure." + "reason": "No shared code-path evidence; GGUF architecture support and Float8 storage lookup are distinct bugs." }, { "left": "issue:44560", "right": "issue:44918", "accept": false, - "reason": "Both involve Qwen3.x video/text workflows, but the reported failures are different and not clearly the same code-path." + "reason": "Both mention Qwen video/embedding behavior, but the reported failures are different enough that they should not be merged on title similarity alone." }, { "left": "issue:43824", "right": "issue:43957", "accept": false, - "reason": "ImportError for a missing Qwen2.5-VL class is distinct from meta-device loading regressions." + "reason": "Separate bugs: import/export naming for Qwen2.5-VL vs meta-device loading regression." }, { "left": "issue:42915", "right": "issue:43957", "accept": false, - "reason": "Qwen3-MoE FP8 config failure and meta-device loading issues are different bugs." + "reason": "Both involve model loading, but the concrete failures are different and not the same bug." }, { "left": "issue:42617", "right": "issue:43366", "accept": false, - "reason": "3d_parallel execution failure is unrelated to GGUF model-architecture support." + "reason": "A script execution failure and GGUF model support are unrelated." }, { "left": "issue:38175", "right": "issue:43994", - "accept": true, - "reason": "Same SigLIP2 inference corruption on google/siglip2-base-patch16-224; both describe broken outputs for the same model family and likely same fix." + "accept": false, + "reason": "Shared SigLIP2 model family only; the packet does not show enough shared bug-body evidence to call them duplicates." }, { "left": "issue:43646", "right": "issue:43950", "accept": false, - "reason": "Custom model initialization regressions and non-persistent-buffer corruption are different issues, even though both involve Transformers 5.x loading behavior." + "reason": "Custom-model initialization and non-persistent-buffer corruption are different regressions." }, { "left": "issue:44560", "right": "issue:44805", "accept": false, - "reason": "Different bugs: a Qwen3-vl-embedding video StopIteration vs a mask/tensor shape mismatch IndexError." + "reason": "Different failure classes: video StopIteration vs mask/tensor shape mismatch." }, { "left": "issue:36010", "right": "issue:42915", "accept": false, - "reason": "GenerationMixin import error and Qwen3-MoE FineGrainedFP8Config failure are unrelated." + "reason": "ImportError for GenerationMixin is unrelated to Qwen3Moe FP8 loading." }, { "left": "issue:43638", "right": "issue:44661", "accept": false, - "reason": "Deepspeed ZeRO-3 Bert index error is not the same as add-new-model-like/tokenizer mapping validation." + "reason": "Both are loading/registration issues, but they affect different code paths and are not the same underlying bug." }, { "left": "issue:44877", "right": "issue:45030", "accept": false, - "reason": "Strict config blocking granite_speech and tiny-random glm4v config validation are different loading issues." + "reason": "Similar validation theme, but different models and different concrete config failures." }, { "left": "issue:43596", "right": "issue:44661", "accept": false, - "reason": "ZeRO-3 empty-index error with BertModel is unrelated to tokenizer mapping-name validation." + "reason": "DeepSpeed ZeRO-3 Bert init failure and TOKENIZER_MAPPING_NAMES restriction are unrelated." }, { "left": "issue:43644", "right": "issue:44534", "accept": true, - "reason": "Same non-persistent-buffer regression in Transformers v5; the titles describe the same bug almost verbatim." + "reason": "Near-verbatim duplicate: both report Transformers v5 filling non-persistent buffers with junk." }, { "left": "issue:43317", "right": "issue:44756", "accept": false, - "reason": "device_map/offload loading failure and Strix Halo mmap OOM are different problems." + "reason": "Different issues: dequantized-model offload failure vs mmap/OOM behavior on Strix Halo." }, { "left": "issue:45313", "right": "issue:45357", "accept": false, - "reason": "Qwen3.5 ZeRO-3 weight loading failure and save_pretrained visual-encoder key regression are different bugs." + "reason": "Same model family, but ZeRO-3 weight loading and save_pretrained visual-key serialization are different bugs." }, { "left": "issue:38175", "right": "issue:39692", "accept": false, - "reason": "Related SigLIP2 context, but 39692 is a docs-example/model-processor mismatch plus quantization failure, not the same concrete runtime bug as zero probabilities." + "reason": "SigLIP2 overlap is too broad; the issues describe different symptoms and one is an example/documentation problem." }, { "left": "issue:42491", "right": "issue:43720", "accept": false, - "reason": "LoRA/qwen3_moe hf4.x vs hf5.x compatibility is unrelated to BitNet packed-weight loading." + "reason": "Different code paths: Qwen3-MoE LoRA compatibility vs BitNet packed-weight loading." }, { "left": "issue:44960", "right": "issue:44977", "accept": false, - "reason": "GLM5 and Qwen3.5 flash-attention generation failures are unrelated model-specific issues." + "reason": "Both are Qwen/GLM-family generation reports, but they do not appear to be the same concrete bug." }, { "left": "issue:44451", "right": "issue:44488", "accept": false, - "reason": "Both are model-loading complaints for different checkpoints, but the available evidence is not enough to conclude the same underlying bug." + "reason": "Similar BERT-loading symptom, but the packet does not provide enough evidence that they are the same underlying checkpoint issue." } ] }, "evaluator_result": { - "accept": true, - "feedback": "The summary is grounded in the packet, and the accepted soft pairs are defensible: 41093/44805 and 43644/44534 are near-verbatim duplicate regressions, and 38175/43994 is a reasonable conservative match on the same SigLIP2 output-corruption symptom. The rejects are appropriately cautious and do not overclaim." + "accept": false, + "feedback": "Several soft-edge accepts are not conservative enough given the packet. In particular, 41093\u219444805, 44479\u219444560, 38175\u219443994, and 44451\u219444488 are all supported only by broad title similarity/model family overlap; the packet provides no shared issue-body evidence, so treating them as duplicates is too strong. The summary also overstates the number of \u2018clear duplicates\u2019 and is a bit inconsistent with the verdict list. The strong near-exact duplicate 43644\u219444534 looks grounded, but the overall clustering should be more cautious." }, "evaluator_used": true, - "retried": false, + "retried": true, "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null @@ -1641,8 +1635,8 @@ "issue:39692|issue:44960", "issue:44387|issue:44756", "issue:43425|issue:43957", - "issue:44292|issue:44912", "issue:43854|issue:45216", + "issue:44292|issue:44912", "issue:42947|issue:43381", "issue:44291|issue:44756", "issue:45276|issue:45292", @@ -1654,15 +1648,15 @@ "issue:45292|issue:45335", "issue:44479|issue:44877", "issue:44402|issue:44451", - "issue:43257|issue:44863", - "issue:43645|issue:44164" + "issue:43645|issue:44164", + "issue:43257|issue:44863" ], - "prepared_review_unit_hash": "7f2d7383e6fde3637a5a1bd896236fae67c50eb0c3f1332e8c4eebaac87827fb", + "prepared_review_unit_hash": "e07bfa254ad8ee572a4915fbba0937250bc39573183d250a41d3950f3325f6e0", "component_budget": { "node_count": 367, "item_count": 367, "soft_pair_count": 1370, - "serialized_chars": 706987, + "serialized_chars": 706986, "estimated_input_tokens": 176747, "estimated_eval_tokens": 353750 }, @@ -1670,7 +1664,7 @@ "node_count": 48, "item_count": 48, "soft_pair_count": 40, - "serialized_chars": 39470, + "serialized_chars": 39471, "estimated_input_tokens": 9868, "estimated_eval_tokens": 19992 }, @@ -1679,258 +1673,258 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This set is mostly a grab-bag of unrelated issue pairs. The only clear duplicate subgroup is the `resize_token_embeddings` regression, plus one strong pair about untied `lm_head` weights in vision-language generation models.", - "confidence": 0.82, - "canonical_issue_reason": "issue:45276 is the best canonical issue for the `resize_token_embeddings` regression cluster: it is the broadest report and directly covers the downstream embedding/output-head propagation problem that also appears in 45292 and 45335.", + "summary": "Mostly a heterogeneous set of unrelated model/load/regression reports. The only clear duplicate family is the `resize_token_embeddings` regression affecting model-specific embedding/output layers across Gemma4, T5Gemma, and the generic output_embeddings report.", + "confidence": 0.86, + "canonical_issue_reason": "Issue 45292 is the best representative of the duplicate family: it states the core, model-agnostic symptom (`resize_token_embeddings` not affecting `output_embeddings`) that the Gemma4 and T5Gemma reports appear to be instances of.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45276 is the strongest global issue anchor in this set because it represents the clearest multi-report duplicate cluster; the other pairs are mostly unrelated or only loosely similar.", + "best_issue_reason": "Issue 45292 is the strongest global issue candidate because it is the most generic and reusable description of the shared `resize_token_embeddings` bug, and it is open while the near-duplicates are more model-specific.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43708", "right": "issue:44075", "accept": false, - "reason": "Different bugs: checkpoint resumption step calculation vs unused SGD optimizer args." + "reason": "Different bugs: checkpoint resumption step calculation vs optimizer arguments being ignored." }, { "left": "issue:43854", "right": "issue:43931", "accept": false, - "reason": "Both are model-loading failures, but for different models and different root symptoms." + "reason": "Both are loading failures, but for different models and different failure modes." }, { "left": "issue:43856", "right": "issue:43873", "accept": false, - "reason": "Both mention memory/quantization, but one is Qwen3 MoE training memory use and the other is quantization offloading behavior." + "reason": "One is a Qwen3 MoE memory issue; the other is offloading with quantization. Not the same underlying bug." }, { "left": "issue:42491", "right": "issue:45276", "accept": false, - "reason": "Unrelated: LoRA compatibility across versions vs token-embedding resizing behavior." + "reason": "LoRA compatibility on hf4.x/main branch is unrelated to Gemma4 token-embedding resizing." }, { "left": "issue:41720", "right": "issue:42915", "accept": false, - "reason": "Same model family, but different failures: device-map CUDA assert vs FineGrainedFP8Config load failure." + "reason": "Both involve Qwen3-family models, but one is auto device mapping/cuda assert and the other is FineGrainedFP8 config failure." }, { "left": "issue:43257", "right": "issue:43854", "accept": false, - "reason": "Different models and different loading paths; no concrete shared bug." + "reason": "Accelerate+deepspeed weight conversion is unrelated to GLM-4.7-Flash loading in unit tests." }, { "left": "issue:39692", "right": "issue:43994", "accept": false, - "reason": "Same SigLIP2 family, but one is a doc-example/quantization issue and the other is incorrect runtime outputs." + "reason": "Both mention SigLIP2, but one is a broken docs example and the other is bad inference outputs for a specific checkpoint." }, { "left": "issue:43856", "right": "issue:44387", "accept": false, - "reason": "Both concern memory, but one is Qwen3 MoE training efficiency and the other is int4 reserved-memory OOM." + "reason": "Memory inefficiency in Qwen3 MoE training is not the same as increased reserved CUDA memory under int4 quantization." }, { "left": "issue:44488", "right": "issue:44991", "accept": false, - "reason": "Both are tokenizer-loading failures, but the reports are on different models and the underlying cause is not specific enough to merge." + "reason": "Tokenizer loading failures for different models; same broad area, different underlying breakages." }, { "left": "issue:45362", "right": "issue:45406", "accept": false, - "reason": "Different model families and different failures: chat crash vs Gemma4Processor attribute error in serve." + "reason": "Qwen3.5 chat crash and Gemma4Processor missing `_tokenizer` are unrelated code paths." }, { "left": "issue:43644", "right": "issue:43645", "accept": false, - "reason": "Both are Transformers 5.0 regressions, but they describe different breakages." + "reason": "Non-persistent buffer junk and Jupyter custom-model initialization are different regressions." }, { "left": "issue:42915", "right": "issue:44756", "accept": false, - "reason": "Different subsystems: Qwen3 MoE FP8 loading vs mmap-related OOM on Strix Halo." + "reason": "Qwen3Moe FP8 failure is unrelated to the Strix Halo mmap/OOM report." }, { "left": "issue:42491", "right": "issue:43931", "accept": false, - "reason": "Different model/version compatibility problems with no shared concrete code path." + "reason": "LoRA reuse on main branch is not the same as Qwen3-VL weight-shape mismatch." }, { "left": "issue:44661", "right": "issue:44683", "accept": false, - "reason": "Tokenizer mapping logic vs compiled flex attention on newer torch are unrelated." + "reason": "Tokenizer mapping lookup failure and compiled flex_attention on newer torch are unrelated." }, { "left": "issue:42491", "right": "issue:43856", "accept": false, - "reason": "Different problems: LoRA portability across versions vs Qwen3 MoE training memory usage." + "reason": "LoRA compatibility and Qwen3 MoE memory usage are separate issues." }, { "left": "issue:43404", "right": "issue:43454", - "accept": true, - "reason": "Same underlying bug pattern: missing `lm_head` weight tying causes bad generation in related vision-language conditional generation models." + "accept": false, + "reason": "Both mention untied lm_head weights, but they are different model implementations and symptom scopes." }, { "left": "issue:45276", "right": "issue:45335", "accept": true, - "reason": "Same regression class: `resize_token_embeddings` does not propagate updated embeddings/output heads to model-specific components." + "reason": "Same regression family: `resize_token_embeddings` fails to propagate to model-specific embedding/output layers." }, { "left": "issue:36010", "right": "issue:39404", "accept": false, - "reason": "Completely different issues: import error for `GenerationMixin` vs Whisper pipeline `return_language` behavior." + "reason": "ImportError for GenerationMixin and Whisper return_language pipeline regression are unrelated." }, { "left": "issue:44164", "right": "issue:44291", "accept": false, - "reason": "Different serialization/init failures: `extra_state` handling vs `init_empty_weights` unexpected argument." + "reason": "Save/from_pretrained extra_state handling is different from init_empty_weights/_is_hf_initialized argument handling." }, { "left": "issue:41720", "right": "issue:44155", "accept": false, - "reason": "Different bugs: Qwen3 device-mapping assert vs AudioFlamingo3 track-to-track leakage." + "reason": "Qwen3 device-mapping CUDA assert and AudioFlamingo3 embedding/token leak are unrelated." }, { "left": "issue:44402", "right": "issue:44488", "accept": false, - "reason": "Both are tokenizer/model loading complaints, but not the same concrete defect." + "reason": "Tokenizer vocab mismatch for one model is not the same as loading a different model failing entirely." }, { "left": "issue:43873", "right": "issue:44387", "accept": false, - "reason": "Related to quantization and memory, but the failures are distinct." + "reason": "Quantization offloading behavior and int4 reserved-memory OOM are different failure mechanisms." }, { "left": "issue:39692", "right": "issue:44960", "accept": false, - "reason": "Different model families and unrelated symptoms." + "reason": "SigLIP2 docs/load issues are unrelated to GLM5." }, { "left": "issue:44387", "right": "issue:44756", "accept": false, - "reason": "Both are OOM reports, but one is int4 quantization memory growth and the other is mmap on specific hardware." + "reason": "Int4 memory OOM and disabling mmap on Strix Halo are different problems." }, { "left": "issue:43425", "right": "issue:43957", "accept": false, - "reason": "Both involve loading/runtime compatibility, but they are different torch/version-specific failures." + "reason": "Torch version incompatibility and meta-device model loading failures are not the same bug." }, { - "left": "issue:44291", - "right": "issue:44912", + "left": "issue:43854", + "right": "issue:45216", "accept": false, - "reason": "Unexpected `_is_hf_initialized` argument vs MXFP4 fallback to bf16 are unrelated." + "reason": "Unit-test loading issues for GLM-4.7-Flash and Qwen3.5 save_pretrained regression are unrelated." }, { - "left": "issue:43854", - "right": "issue:45216", + "left": "issue:44292", + "right": "issue:44912", "accept": false, - "reason": "Different model-loading and save-pretrained regressions." + "reason": "Different model/version loading problems with no shared concrete code path." }, { "left": "issue:42947", "right": "issue:43381", "accept": false, - "reason": "Both mention gradient checkpointing, but one is ineffectiveness with PEFT LoRA and the other is an eval-mode restriction." + "reason": "Gradient checkpointing inefficacy in LoRA and inability to use checkpointing in eval mode are distinct behaviors." }, { "left": "issue:44291", "right": "issue:44756", "accept": false, - "reason": "Unrelated: init_empty_weights argument handling vs mmap OOM." + "reason": "init_empty_weights constructor mismatch is unrelated to Strix Halo mmap/OOM." }, { "left": "issue:45276", "right": "issue:45292", "accept": true, - "reason": "Same `resize_token_embeddings` bug: downstream output embeddings are not updated correctly." + "reason": "Both report the same `resize_token_embeddings` family of failures affecting output/embedding layers." }, { "left": "issue:44743", "right": "issue:44960", "accept": false, - "reason": "Unrelated model-specific recurrence/cache bug vs GLM5 issue." + "reason": "Qwen3.5 recurrent-state reset and GLM5 are unrelated." }, { "left": "issue:43381", "right": "issue:44928", "accept": false, - "reason": "Different training/runtime failures: checkpointing in eval mode vs SDPA fallback causing gradient explosion." + "reason": "Eval-mode checkpointing limitation is unrelated to the Qwen3.5 gradient explosion from 3D position_ids." }, { "left": "issue:43957", "right": "issue:44756", "accept": false, - "reason": "Different causes and code paths: meta-device loading vs mmap-induced OOM." + "reason": "Meta-device loading regressions and Strix Halo mmap OOM are not the same underlying issue." }, { "left": "issue:44164", "right": "issue:44756", "accept": false, - "reason": "Different issues: save/load extra_state vs mmap on Strix Halo." + "reason": "extra_state save/load bug is unrelated to mmap/OOM behavior." }, { "left": "issue:43646", "right": "issue:43708", "accept": false, - "reason": "Custom model initialization regression vs checkpoint resume step counting are unrelated." + "reason": "Custom model initialization regression and trainer checkpoint step calculation are unrelated." }, { "left": "issue:45292", "right": "issue:45335", "accept": true, - "reason": "Same bug family: `resize_token_embeddings` fails to update model-specific embedding/output modules." + "reason": "Same underlying bug: `resize_token_embeddings` does not update model-specific embedding/output modules." }, { "left": "issue:44479", "right": "issue:44877", "accept": false, - "reason": "Different model-loading regressions: video-input path vs strict config loading." + "reason": "Video-input regression across Qwen variants and strict config preventing granite_speech load are different issues." }, { "left": "issue:44402", "right": "issue:44451", "accept": false, - "reason": "Different tokenizer/model loading failures without evidence of the same root cause." + "reason": "Different model/tokenizer load failures, not the same defect." }, { - "left": "issue:43257", - "right": "issue:44863", + "left": "issue:43645", + "right": "issue:44164", "accept": false, - "reason": "Different models and different loading failures." + "reason": "Jupyter custom-model initialization and extra_state serialization are unrelated." }, { - "left": "issue:43645", - "right": "issue:44164", + "left": "issue:43257", + "right": "issue:44863", "accept": false, - "reason": "Jupyter custom-model initialization breakage is unrelated to `extra_state` save/load handling." + "reason": "Qwen3 MOE weight conversion with accelerate/deepspeed is unrelated to NemotronH checkpoint loading." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is grounded in the packet and stays conservative on the soft-edge pairs. The three accepted duplicate links around `resize_token_embeddings` are well supported, and the `lm_head`-tying pair is also plausibly the same bug class. No pair is overstated into a merge where the packet evidence only shows loose similarity." + "feedback": "The summary is grounded in the packet and stays conservative about the soft-edge pairs. The only clearly supported duplicate family is the `resize_token_embeddings` regression spanning 45276/45292/45335, and the chosen canonical issue is a reasonable generic representative." }, "evaluator_used": true, "retried": false, @@ -2037,12 +2031,12 @@ "issue:42947|issue:44756", "issue:44206|issue:44220" ], - "prepared_review_unit_hash": "642228e3e48f6478b3226f547d49c164a42ee086f976a1217408c7ba44c01f4d", + "prepared_review_unit_hash": "a5db50251de14db2f31c420599a517837bd7aceafaa26103fd8f65fb29ba80fe", "component_budget": { "node_count": 367, "item_count": 367, "soft_pair_count": 1370, - "serialized_chars": 706987, + "serialized_chars": 706986, "estimated_input_tokens": 176747, "estimated_eval_tokens": 353750 }, @@ -2050,7 +2044,7 @@ "node_count": 48, "item_count": 48, "soft_pair_count": 35, - "serialized_chars": 38178, + "serialized_chars": 38179, "estimated_input_tokens": 9545, "estimated_eval_tokens": 19346 }, @@ -2059,222 +2053,222 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "No strong duplicate cluster emerges here: the soft pairs are mostly broad-topic lookalikes (v5 regressions, Whisper issues, tokenizer loading, model loading) but they point to different concrete bugs/code paths. I\u2019d treat them as separate issues.", - "confidence": 0.89, - "canonical_issue_reason": "If one issue must stand in for the set, issue #43611 is the broadest core regression: it targets a central Transformers v5 model-loading breakage and is more representative than the narrower, model-specific reports.", + "summary": "This cluster is mostly a loose collection of unrelated Transformers v5 regressions, tokenizer/loading bugs, Whisper/audio issues, and docs complaints rather than a true duplicate set. The soft links are largely broad-similarity false positives.", + "confidence": 0.77, + "canonical_issue_reason": "43611 is the broadest representative of the v5 loading/init regression family and the closest umbrella issue among these mostly separate reports.", "canonical_pr_reason": null, - "best_issue_reason": "Issue #43611 is the best single representative candidate because it describes a core loading regression in v5 and sits closest to the common theme among the similar-looking reports.", + "best_issue_reason": "43611 is the most general issue here and best represents the cluster\u2019s dominant theme of Transformers v5 model-loading regressions.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43425", "right": "issue:43828", "accept": false, - "reason": "One is a Torch 2.10 compatibility claim; the other is a Phi-tiny-MoE autocast dtype mismatch. Different failures." + "reason": "Different bugs: torch version incompatibility vs an autocast dtype-mismatch in Phi-tiny-MoE." }, { "left": "issue:43828", "right": "issue:43957", "accept": false, - "reason": "Autocast dtype mismatch for one model vs meta-device loading breakage across models; not the same bug." + "reason": "Both are model/runtime failures, but one is autocast dtype mismatch and the other is meta-device loading; not the same code path." }, { "left": "issue:44589", "right": "issue:44756", "accept": false, - "reason": "Float8 storage serialization error vs Strix Halo mmap/OOM behavior; unrelated code paths." + "reason": "Unrelated failures: Float8 storage lookup vs mmap/OOM behavior on Strix Halo." }, { "left": "issue:43906", "right": "issue:44530", "accept": false, - "reason": "A reproduction of an older issue vs a Qwen3.5 PagedAttentionCache group-type crash; different problems." + "reason": "The first is a reproduction of an earlier issue; the second is a Qwen3.5 cache/group-type crash. Different root causes." }, { "left": "issue:43931", "right": "issue:44863", "accept": false, - "reason": "Both are model-loading failures, but for different models and likely different root causes/code paths." + "reason": "Both involve loading models, but they are different model families with different checkpoint-shape/implementation failures." }, { "left": "issue:43611", "right": "issue:43645", "accept": false, - "reason": "Base_model_prefix loading regression vs custom model initialization in Jupyter notebooks; distinct regressions." + "reason": "Both are v5 regressions, but one is base_model_prefix loading and the other is custom model init in notebooks; not the same bug." }, { "left": "issue:39404", "right": "issue:44912", "accept": false, - "reason": "Whisper pipeline return_language regression vs MXFP4 quantization fallback on git-oss-20b; unrelated." + "reason": "Whisper return_language pipeline behavior and GPT-OSS MXFP4 quantization are unrelated." }, { "left": "issue:43425", "right": "issue:43716", "accept": false, - "reason": "Generic Torch compatibility complaint vs Mistral-3 image preprocessor/model dtype mismatch; different issues." + "reason": "Torch compatibility is not the same as a Mistral image-preprocessor dtype mismatch." }, { "left": "issue:44530", "right": "issue:45084", "accept": false, - "reason": "Qwen3.5 cache/group-type crash vs a template compilation error; unrelated." + "reason": "Different failures: PagedAttention cache group-type crash vs template-node compilation error." }, { "left": "issue:43901", "right": "issue:44509", "accept": false, - "reason": "Both are docs-related, but they cover different pipeline-task deprecations and behavior changes." + "reason": "Both are docs complaints, but they refer to different removed/changed API surfaces and not one concrete issue." }, { "left": "issue:43299", "right": "issue:43931", "accept": false, - "reason": "Different model families and different load-time failures; too broad to merge." + "reason": "Same general model family, but one is a Qwen3VL MoE loading regression and the other is a checkpoint shape mismatch; not clearly the same bug." }, { "left": "issue:44265", "right": "issue:44898", "accept": false, - "reason": "torch.export failure with torch_compilable_check vs Perceiver resize/interpolation failure; separate code paths." + "reason": "torch.export with torch_compilable_check and Perceiver interpolation are separate model/export issues." }, { "left": "issue:44479", "right": "issue:44805", "accept": false, - "reason": "Qwen VL video regression vs a mask/tensor shape IndexError; not the same underlying bug." + "reason": "Both are runtime errors, but one is a video-input regression and the other is a mask/tensor shape mismatch." }, { "left": "issue:43994", "right": "issue:44079", "accept": false, - "reason": "Model output quality issue vs a ModelOutput key-assignment bug; different layers of the stack." + "reason": "Model output quality on SigLIP2 is not the same as incorrect ModelOutput key assignment." }, { "left": "issue:44671", "right": "issue:44991", "accept": false, - "reason": "CamemBERT MLM prediction regression vs tokenizer loading failure for est-roberta; unrelated." + "reason": "CamemBERT prediction errors and a tokenizer load failure for est-roberta are different regressions." }, { "left": "issue:43856", "right": "issue:44756", "accept": false, - "reason": "Qwen3 MoE memory usage vs Strix Halo mmap/OOM workaround; different problems." + "reason": "One is Qwen3 MoE training memory usage; the other is an mmap/OOM workaround on Strix Halo." }, { "left": "issue:39404", "right": "issue:44671", "accept": false, - "reason": "Whisper pipeline bug vs CamemBERT MLM predictions; unrelated." + "reason": "Whisper return_language and CamemBERT masked-LM predictions are unrelated bugs." }, { "left": "issue:40990", "right": "issue:42947", "accept": false, - "reason": "High perplexity on gpt-oss-20b vs gradient checkpointing/LoRA behavior; not the same issue." + "reason": "Perplexity on gpt-oss and gradient checkpointing with LoRA are different problems." }, { "left": "issue:44589", "right": "issue:45084", "accept": false, - "reason": "Float8 storage lookup error vs template-node compilation failure; different failures." + "reason": "Float8 storage deserialization and template compilation failures are unrelated." }, { "left": "issue:43596", "right": "issue:44683", "accept": false, - "reason": "DeepSpeed zero3/BERT index error vs flex_attention compile failure on newer Torch; unrelated." + "reason": "DeepSpeed zero3/BertModel initialization and compiled flex_attention on newer torch versions are separate issues." }, { "left": "issue:43425", "right": "issue:44918", "accept": false, - "reason": "Torch compatibility complaint vs Qwen3.5 embedding unpacking with TRL; different concrete bugs." + "reason": "Torch compatibility is unrelated to Qwen3.5 input-embedding unpacking in TRL SFT trainer." }, { "left": "issue:36032", "right": "issue:43452", "accept": false, - "reason": "T5 tokenizer method-name conflict vs gguf_file loading breakage; distinct tokenizer/model-loading paths." + "reason": "T5 tokenizer method-name conflict and gguf_file loading breakage are different tokenizer/model-loading bugs." }, { "left": "issue:32090", "right": "issue:39404", "accept": false, - "reason": "Trainer GPU broadcast TypeError vs Whisper return_language regression; unrelated." + "reason": "Trainer GPU broadcast NoneType error is unrelated to Whisper pipeline return_language behavior." }, { "left": "issue:43582", "right": "issue:43632", "accept": false, - "reason": "Apple Silicon warmup TypeError vs _is_hf_initialized flag regression; separate issues." + "reason": "Apple Silicon allocator warmup TypeError and _is_hf_initialized flag breakage do not share a concrete code path." }, { "left": "issue:44991", "right": "issue:45081", "accept": false, - "reason": "Different tokenizer-loading bugs for different models and different mechanisms." + "reason": "Both are tokenizer-loading problems, but they affect different tokenizers and different failure mechanisms." }, { "left": "issue:43957", "right": "issue:44164", "accept": false, - "reason": "Meta-device loading regression vs save/from_pretrained extra_state handling; different lifecycle stages." + "reason": "Meta-device loading regressions and save/from_pretrained extra_state handling are distinct issues." }, { "left": "issue:40990", "right": "issue:43421", "accept": false, - "reason": "Perplexity/model-quality issue vs special-token post-processor update bug; unrelated." + "reason": "A perplexity regression is not the same as special-token post-processor updates at runtime." }, { "left": "issue:44164", "right": "issue:44479", "accept": false, - "reason": "Save/load extra_state bug vs Qwen VL video regression; different subsystems." + "reason": "save/from_pretrained extra_state handling and video-input regressions are unrelated." }, { "left": "issue:44488", "right": "issue:45081", "accept": false, - "reason": "Loading sleng-bert fails vs Mistral regex patch crash; not the same bug." + "reason": "Both are tokenizer load failures, but they target different models and different tokenizer code paths." }, { "left": "issue:43883", "right": "issue:44488", "accept": false, - "reason": "Missing all_tied_weights_keys on MolmoForCausalLM vs model-loading failure for a different checkpoint; unrelated." + "reason": "Missing tied-weights metadata in MolmoForCausalLM is unrelated to loading sleng-bert." }, { "left": "issue:43425", "right": "issue:45161", "accept": false, - "reason": "Torch 2.10 incompatibility vs GPT-OSS MoE tensor-parallel issue; different root causes." + "reason": "Torch 2.10 compatibility is a dependency issue; GPT-OSS tensor-parallel failure is a model-parallelism bug." }, { "left": "issue:43611", "right": "issue:43644", "accept": false, - "reason": "Both are v5 regressions, but one is loading models via base_model_prefix and the other is non-persistent buffer initialization; not the same fix." + "reason": "Both are v5 regressions, but one is model-loading with base_model_prefix and the other is non-persistent buffer initialization." }, { "left": "issue:44811", "right": "issue:45584", "accept": false, - "reason": "Both are Whisper-related, but one is batch_decode skipping special tokens and the other is empty-transcription generation after align_special_tokens." + "reason": "Both involve Whisper token handling, but batch_decode skip_special_tokens and empty-transcription generation after align_special_tokens are different bugs." }, { "left": "issue:42947", "right": "issue:44756", "accept": false, - "reason": "LoRA gradient checkpointing effectiveness vs Strix Halo mmap/OOM; unrelated." + "reason": "LoRA gradient-checkpointing effectiveness and Strix Halo mmap OOM are unrelated." }, { "left": "issue:44206", "right": "issue:44220", "accept": false, - "reason": "Related audio feature-extraction area, but one is an unsupported-arg crash and the other is a generic _torch_extract_fbank_features issue; not clearly the same underlying bug." + "reason": "The titles suggest related audio-feature code, but they describe different failures without enough evidence of one underlying bug." } ] }, @@ -2381,15 +2375,15 @@ "issue:44220|issue:44683", "issue:43404|issue:45127", "issue:42915|issue:45084", - "issue:44479|issue:45325", - "issue:44987|issue:45081" + "issue:44987|issue:45081", + "issue:44479|issue:45325" ], - "prepared_review_unit_hash": "75e22df5a2acf91198120031ec5ebe5c94f7d18256a29b543447e5c1311baa6f", + "prepared_review_unit_hash": "04da4f47d7694292770f2b927958323dd20fff88492dbb9be6b7a593ea0bb7ee", "component_budget": { "node_count": 367, "item_count": 367, "soft_pair_count": 1370, - "serialized_chars": 706987, + "serialized_chars": 706986, "estimated_input_tokens": 176747, "estimated_eval_tokens": 353750 }, @@ -2397,7 +2391,7 @@ "node_count": 48, "item_count": 48, "soft_pair_count": 35, - "serialized_chars": 38187, + "serialized_chars": 38188, "estimated_input_tokens": 9547, "estimated_eval_tokens": 19350 }, @@ -2406,228 +2400,228 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Mostly a heterogeneous set of issue reports with only a couple of clear duplicate pairs: GGUF loading regressions and Qwen3.5 save_pretrained regressions. The rest are different bugs that only share broad subsystem themes.", + "summary": "Mostly a heterogeneous cluster of unrelated Transformers regressions; only a couple of pairs look like true duplicates, centered on Qwen3.5 save_pretrained serialization and GGUF loading.", "confidence": 0.37, - "canonical_issue_reason": "issue:43452 is the broadest clean duplicate candidate: it describes a generic GGUF/from_pretrained loading regression affecting both tokenizer and causal LM code paths, making it a better anchor than the more model-specific follow-up reports.", + "canonical_issue_reason": "The cluster is broadly mixed, but the clearest duplicate family is the Qwen3.5 `save_pretrained` regression; issue 45216 is the broader earlier report, with 45357 as a later, more specific follow-up.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43452 is the most reusable representative issue in the cluster because it names the shared API/code path and likely covers the duplicate GGUF-loading reports with the least model-specific noise.", + "best_issue_reason": "45216 is the best representative issue because it states the broader checkpoint-serialization failure and predates the narrower variant.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44488", "right": "issue:44987", "accept": false, - "reason": "Both are load failures, but they target different models and different failure modes." + "reason": "Both are model-loading failures, but they affect different models and appear to be different breakages rather than one shared bug." }, { "left": "issue:34567", "right": "issue:40444", "accept": false, - "reason": "Trainer token-count bookkeeping is unrelated to the iterable-dataset multimodal loading failure." + "reason": "Trainer token-count tracking and Qwen2.5-VL iterable-dataset image handling are unrelated code paths." }, { "left": "issue:43716", "right": "issue:45084", "accept": false, - "reason": "Image dtype mismatch and template compilation errors are unrelated bugs." + "reason": "Mistral-3 dtype mismatch and a template-node compilation TypeError are different failures with no clear shared root cause." }, { "left": "issue:43883", "right": "issue:44991", "accept": false, - "reason": "Different models and different failure paths; only a broad tokenizer-loading theme is shared." + "reason": "Molmo tied-weights attribute error and tokenizer loading for EMBEDDIA/est-roberta are unrelated." }, { "left": "issue:43644", "right": "issue:43646", "accept": false, - "reason": "One is about non-persistent buffer initialization, the other about custom model initialization; not the same bug." + "reason": "Both mention v5 regressions, but one is about non-persistent buffers and the other about custom model initialization; not the same concrete bug." }, { "left": "issue:43643", "right": "issue:44530", "accept": false, - "reason": "AutoConfig remote-code field loss is unrelated to PagedAttention cache group-type handling." + "reason": "Missing fields from `trust_remote_code=True` config loading is unrelated to the Qwen3.5 linear-attention cache crash." }, { "left": "issue:43452", "right": "issue:43482", "accept": true, - "reason": "Both point to the same GGUF/from_pretrained loading regression in v5; one is the generic API report, the other a model-specific manifestation." + "reason": "Both report the same GGUF-loading regression in v5; 43482 looks like the Qwen2.5-specific manifestation of the generic `gguf_file` breakage in 43452." }, { "left": "issue:43317", "right": "issue:43482", "accept": false, - "reason": "GPU+CPU offload/dequantized-model loading is a different code path from the GGUF loading failure." + "reason": "Device-map/offload loading failure is a different problem from GGUF model loading." }, { "left": "issue:44451", "right": "issue:44987", "accept": false, - "reason": "These are separate model-specific tokenizer/load failures with no clear shared bug." + "reason": "Two separate model-loading failures with different repos and likely different causes." }, { "left": "issue:36032", "right": "issue:42915", "accept": false, - "reason": "T5 tokenizer special-token conflict and Qwen3Moe FP8 config failure are unrelated." + "reason": "Tokenizer method-name conflict in T5 and Qwen3MoE FP8 loading are unrelated." }, { "left": "issue:43541", "right": "issue:45084", "accept": false, - "reason": "Torch Dynamo grouped_mm tracing failure is unrelated to the template compilation error." + "reason": "Different compile/tracing errors in different models; not the same code-path issue." }, { "left": "issue:39404", "right": "issue:43825", "accept": false, - "reason": "Whisper return_language regression and pipeline error-message wording are different issues." + "reason": "Whisper `return_language` behavior and a pipeline error-message regression are not the same bug." }, { "left": "issue:42915", "right": "issue:44530", "accept": false, - "reason": "Different Qwen3-family failures with different triggers and code paths." + "reason": "Qwen3MoE FP8 loading and Qwen3.5 PagedAttentionCache linear-attention crash are unrelated." }, { "left": "issue:44849", "right": "issue:45325", "accept": false, - "reason": "Hidden-states output bug and RoPE temporal-position scaling bug are unrelated." + "reason": "Both are Qwen3.5/Qwen2.5-VL regressions, but one is about hidden states and the other about rope/position-id scaling; too different to merge." }, { "left": "issue:43262", "right": "issue:43377", "accept": false, - "reason": "Audio chat-template sampling-rate default and MIMI padding-mask batching bug are different problems." + "reason": "Audio chat-template sampling-rate defaulting and MIMI batch-vs-single padding-mask mismatch are separate issues." }, { "left": "issue:43404", "right": "issue:44466", "accept": false, - "reason": "Both involve tied weights, but one is a missing tie and the other is serialization inconsistency; not the same concrete bug." + "reason": "Both involve tied weights, but one is a missing tie in model init and the other is device-dependent serialization; different symptoms and likely different fixes." }, { "left": "issue:36032", "right": "issue:43645", "accept": false, - "reason": "Tokenizer special-token conflict is unrelated to custom model initialization in notebooks." + "reason": "T5 tokenizer loading conflict and custom-model initialization in notebooks are unrelated." }, { "left": "issue:43646", "right": "issue:44534", "accept": false, - "reason": "Custom model initialization and non-persistent buffer junk are separate regressions." + "reason": "Both are v5 initialization regressions, but the titles point to different concrete failures; not enough to treat as the same bug." }, { "left": "issue:44987", "right": "issue:44991", "accept": false, - "reason": "Different model loading failures; no shared underlying bug is evident." + "reason": "Two distinct tokenizer/model-loading failures for different model families." }, { "left": "issue:45084", "right": "issue:45507", "accept": false, - "reason": "Template compilation failure and GraniteMoE invalid method call are unrelated." + "reason": "Template compilation error and GraniteMoEHybrid invalid-method call are different issues." }, { "left": "issue:43994", "right": "issue:44610", "accept": false, - "reason": "Model output quality issue and processor/model size mismatch are different bugs." + "reason": "Vision encoder nonsense outputs and OmDet-Turbo processor size mismatch are unrelated." }, { "left": "issue:43645", "right": "issue:44534", "accept": false, - "reason": "Notebook custom-model init breakage is not the same as the buffer-initialization regression." + "reason": "Notebook custom-model initialization and non-persistent buffer junk are related in theme but not clearly the same underlying bug." }, { "left": "issue:45216", "right": "issue:45357", "accept": true, - "reason": "Both describe the same Qwen3.5 save_pretrained regression; 45357 is a later, more specific report about incorrect visual-encoder keys." + "reason": "Both describe the same Qwen3.5 `save_pretrained` serialization regression; 45357 is a later, narrower report about the incorrect visual-encoder keys." }, { "left": "issue:43322", "right": "issue:43540", "accept": false, - "reason": "Llava Next segfault and Qwen3OmniMoe video ValueError are distinct failures." + "reason": "Llava Next segfault and Qwen3OmniMoe video-processing ValueError are different multimodal loading bugs." }, { "left": "issue:43377", "right": "issue:43819", "accept": false, - "reason": "MIMI batched-vs-single padding-mask mismatch is unrelated to DAC from_latents STE behavior." + "reason": "MIMI padding-mask batching bug and DAC latent reconstruction/STE bug are unrelated." }, { "left": "issue:44403", "right": "issue:45081", "accept": false, - "reason": "Loading-noise complaint and Mistral regex patch crash are not the same bug." + "reason": "Generic loading noise and a Mistral regex patch crash are different concerns." }, { "left": "issue:44451", "right": "issue:45081", "accept": false, - "reason": "Different model-loading failure versus tokenizer regex patch crash." + "reason": "Different model-loading failures with no shared concrete code path." }, { "left": "issue:44265", "right": "issue:44610", "accept": false, - "reason": "torch.export / torch_compilable_check failure is unrelated to the OmDet processor size mismatch." + "reason": "torch.export failure with `torch_compilable_check` and an OmDet-Turbo processor size bug are unrelated." }, { "left": "issue:41553", "right": "issue:43582", "accept": false, - "reason": "Bad error message for tokenizer loading and Apple Silicon warmup TypeError are unrelated." + "reason": "Tokenizer error message quality and Apple Silicon allocator warmup TypeError are unrelated." }, { "left": "issue:43632", "right": "issue:44756", "accept": false, - "reason": "_is_hf_initialized flag regression and mmap OOM workaround are unrelated." + "reason": "HF initialization flag regression and Strix Halo mmap OOM mitigation are different issues." }, { "left": "issue:44220", "right": "issue:44683", "accept": false, - "reason": "fbank feature extraction issue and compiled flex_attention failure are separate code paths." + "reason": "Fbank feature extraction and compiled flex_attention failing on torch>=2.9 are unrelated." }, { "left": "issue:43404", "right": "issue:45127", "accept": false, - "reason": "Tied-weight serialization and LoRA-merge model collapse are different bugs." + "reason": "Both mention tied weights, but one is missing weight tying in Mistral3 and the other is LoRA merge collapse with extended vocab; not the same bug." }, { "left": "issue:42915", "right": "issue:45084", "accept": false, - "reason": "FP8 config failure and template compilation failure are unrelated." + "reason": "Qwen3MoE FP8 loading and a template compilation error are unrelated." }, { - "left": "issue:44479", - "right": "issue:45325", + "left": "issue:44987", + "right": "issue:45081", "accept": false, - "reason": "Video-input regression and still-image RoPE scaling bug are not the same issue." + "reason": "Model-loading failure for physical-intelligence/fast and Mistral regex crash are different problems." }, { - "left": "issue:44987", - "right": "issue:45081", + "left": "issue:44479", + "right": "issue:45325", "accept": false, - "reason": "Model loading failure and Mistral regex patch crash do not share a concrete underlying bug." + "reason": "Both are Qwen VL regressions, but one is a broad video-input regression and the other is a specific still-image position-id scaling bug; not clearly the same fix." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is grounded in the packet and stays conservative. The two accepted soft-edge pairs are reasonably supported by the titles, and the remaining rejections avoid overclaiming duplicate status from broad subsystem similarity alone. The canonical/representative choice for the GGUF loading cluster is also defensible from the issue titles." + "feedback": "Grounded overall: the cluster summary stays conservative, and the two accepted soft pairs are the only ones that are clearly supported by the titles. The representative/canonical issue choice for 45216 vs 45357 is reasonable, and the rest of the false verdicts do not overclaim shared root cause." }, "evaluator_used": true, "retried": false, @@ -2700,9 +2694,9 @@ "soft_pairs": [ "issue:43296|issue:43531", "issue:42175|issue:45542", - "issue:40990|issue:44811", "issue:44062|issue:44987", "issue:42915|issue:44291", + "issue:40990|issue:44811", "issue:43611|issue:44534", "issue:43317|issue:43856", "issue:36032|issue:43742", @@ -2714,8 +2708,8 @@ "issue:43421|issue:44568", "issue:36032|issue:41720", "issue:43482|issue:45538", - "issue:44829|issue:45479", "issue:44683|issue:45020", + "issue:44829|issue:45479", "issue:33666|issue:40444", "issue:43425|issue:44403", "issue:44403|issue:44991", @@ -2728,19 +2722,19 @@ "issue:43749|issue:44079", "issue:44079|issue:45072", "issue:43644|issue:44079", - "issue:44265|issue:45072", "issue:43632|issue:45072", + "issue:44265|issue:45072", "issue:40444|issue:45381", "issue:45325|issue:45381", "issue:43531|issue:44291", "issue:43611|issue:43646" ], - "prepared_review_unit_hash": "642079e48ef2694d0212721d6ab6665ae611659d4f0b6461c369e98362eef172", + "prepared_review_unit_hash": "e88ff0dea395ae7723cb1e1d31972aaef7dbc0cfb68788970213e168b38d9062", "component_budget": { "node_count": 367, "item_count": 367, "soft_pair_count": 1370, - "serialized_chars": 706987, + "serialized_chars": 706986, "estimated_input_tokens": 176747, "estimated_eval_tokens": 353750 }, @@ -2748,244 +2742,253 @@ "node_count": 48, "item_count": 48, "soft_pair_count": 36, - "serialized_chars": 38396, - "estimated_input_tokens": 9599, - "estimated_eval_tokens": 19454 + "serialized_chars": 38398, + "estimated_input_tokens": 9600, + "estimated_eval_tokens": 19456 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is thematically similar but mostly contains distinct transformers regressions and model-specific failures (loading, tokenizer, config, FP8, attention, multimodal position IDs). I do not see true duplicate pairs; the best representative issue is the broad open loading-regression report.", - "confidence": 0.88, - "canonical_issue_reason": "issue:45020 is the broadest open report and best umbrella for the recent transformers loading/compatibility regressions, while the other issues are narrower and describe different concrete failures.", + "summary": "This is a noisy mixed cluster of mostly unrelated Transformers regression issues. I only see a few genuine duplicate families: the `_is_hf_initialized`/`init_empty_weights` v5 init regression, the non-persistent buffer junk regression, and likely the Qwen2.5-VL rope/position-id bug.", + "confidence": 0.79, + "canonical_issue_reason": "issue:43632 is the clearest representative of a concrete v5 loading/init regression and anchors one of the strongest duplicate pairs in the set.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45020 is the strongest global representative because it is open, broad, and covers the shared theme of recent transformers breakages without being tied to a single model.", + "best_issue_reason": "issue:43632 is the most specific, actionable issue among the duplicate candidates and best represents the shared `_is_hf_initialized` initialization bug.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43296", "right": "issue:43531", "accept": false, - "reason": "Different bugs: PaddleOCR-VL load failure vs Qwen3-MoE sliding_window behavior." + "reason": "Different models and failures: PaddleOCR-VL load failure vs Qwen3-MoE sliding-window behavior." }, { "left": "issue:42175", "right": "issue:45542", "accept": false, - "reason": "Both touch TF packaging, but one is missing TensorFlow in the torch extra and the other is an undefined tf backend from tensorboard-only install." - }, - { - "left": "issue:40990", - "right": "issue:44811", - "accept": false, - "reason": "Unrelated: GPT-OSS perplexity regression vs Whisper processor.batch_decode skipping special tokens." + "reason": "Both mention TF installation/backend, but one is missing TensorFlow in extras while the other is a runtime backend selection error." }, { "left": "issue:44062", "right": "issue:44987", "accept": false, - "reason": "Tokenizer AddedToken kwarg conflict vs model loading failure for a specific remote_code model; different code paths." + "reason": "Tokenizer AddedToken special-arg conflict and a model load failure for a different repo are not the same bug." }, { "left": "issue:42915", "right": "issue:44291", "accept": false, - "reason": "Both are transformers-v5 loading regressions, but one is FineGrainedFP8Config and the other is init_empty_weights/_is_hf_initialized." + "reason": "Qwen3Moe FP8 config failure and `_is_hf_initialized` loading TypeError are unrelated regressions." + }, + { + "left": "issue:40990", + "right": "issue:44811", + "accept": false, + "reason": "Perplexity regression on GPT-OSS and Whisper skip_special_tokens decoding are different code paths." }, { "left": "issue:43611", "right": "issue:44534", "accept": false, - "reason": "Distinct model-init issues: base_model_prefix loading regression vs non-persistent buffers being filled with junk." + "reason": "Both are Transformers v5 regressions, but one is model loading with base_model_prefix and the other is buffer initialization junk; no clear same bug." }, { "left": "issue:43317", "right": "issue:43856", "accept": false, - "reason": "Different problems: device_map/offload loading failure vs Qwen3 MoE training memory inefficiency." + "reason": "Device-map/offload loading failure and Qwen3 MoE training memory usage are different issues." }, { "left": "issue:36032", "right": "issue:43742", "accept": false, - "reason": "Tokenizer method-name conflict vs a model key error during load; not the same bug." + "reason": "Tokenizer add_special_tokens conflict vs MobileLLM key error are distinct loading failures." }, { "left": "issue:40444", "right": "issue:45325", "accept": false, - "reason": "Both are Qwen2.5-VL, but one fails on iterable multi-image finetuning and the other is a still-image temporal position-id scaling bug." + "reason": "Both involve Qwen2.5-VL, but one is iterable-dataset multi-image finetuning and the other is rope/temporal position scaling." }, { "left": "issue:43632", "right": "issue:44534", "accept": false, - "reason": "Both are v5 regressions, but they concern different internals: _is_hf_initialized vs non-persistent buffers." + "reason": "These are different v5 regressions: `_is_hf_initialized` loading behavior vs non-persistent buffer contents." }, { "left": "issue:43531", "right": "issue:44877", "accept": false, - "reason": "Qwen3-MoE sliding_window issue vs granite_speech strict-config loading; unrelated failures." + "reason": "Qwen3-MoE sliding-window behavior and granite_speech strict config loading are unrelated." }, { "left": "issue:42915", "right": "issue:44589", "accept": false, - "reason": "Different FP8-related failures: FineGrainedFP8Config vs missing Float8 storage object." + "reason": "FP8 config failure and missing Float8 storage are different failure modes." }, { "left": "issue:43818", "right": "issue:45357", "accept": false, - "reason": "Video-LLaVA visual-tower issue vs Qwen3.5 save_pretrained key mapping regression; different models and code paths." + "reason": "Video-LLaVA missing temporal attention/shared weights is unrelated to Qwen3.5 save_pretrained key regression." }, { "left": "issue:43421", "right": "issue:44568", "accept": false, - "reason": "Runtime post-processor update bug vs add_special_tokens not adding BOS/EOS for one tokenizer." + "reason": "Runtime post-processor update for special tokens and BOS/EOS omission on add_special_tokens are related in topic but not the same concrete bug." }, { "left": "issue:36032", "right": "issue:41720", "accept": false, - "reason": "Tokenizer loading conflict vs Qwen3 auto device mapping cudaErrorAssert; not the same underlying issue." + "reason": "Tokenizer loading error and Qwen3 auto device-map cuda assert are unrelated." }, { "left": "issue:43482", "right": "issue:45538", "accept": false, - "reason": "GGUF loading breakage in transformers v5 vs CLIPTokenizer max_length being too large; unrelated." + "reason": "GGUF loading with transformers v5 and CLIPTokenizer max_length are different issues." }, { - "left": "issue:44829", - "right": "issue:45479", + "left": "issue:44683", + "right": "issue:45020", "accept": false, - "reason": "Both affect training behavior, but one is flash_attention_3 degeneration and the other is the single-label num_labels=1 zero-loss bug." + "reason": "Torch compiled flex_attention failure and remote_code model loading regressions do not share a code path." }, { - "left": "issue:44683", - "right": "issue:45020", + "left": "issue:44829", + "right": "issue:45479", "accept": false, - "reason": "Different layers: compiled flex_attention on torch>=2.9 vs transformers remote_code loading regressions." + "reason": "Flash-Attention-3 training degeneration and single-label-zero-loss are different sequence-classification problems." }, { "left": "issue:33666", "right": "issue:40444", "accept": false, - "reason": "Both are vision-language training issues, but Qwen2 multi-GPU training and Qwen2.5-VL multi-image iterable loading are different problems." + "reason": "Both are Qwen VL training issues, but one is multi-GPU training and the other is multi-image iterable dataset handling." }, { "left": "issue:43425", "right": "issue:44403", "accept": false, - "reason": "Torch compatibility issue vs loading noise/logging issue; not a duplicate." + "reason": "Torch 2.10 incompatibility and noisy transformer loading are unrelated." }, { "left": "issue:44403", "right": "issue:44991", "accept": false, - "reason": "Unnecessary loading noise vs tokenizer load failure for a specific model; different symptoms and fixes." + "reason": "Noisy loading vs tokenizer load failure for a specific model are different symptoms and code paths." }, { "left": "issue:44079", "right": "issue:44534", "accept": false, - "reason": "ModelOutput key assignment bug vs non-persistent buffer initialization bug; unrelated." + "reason": "ModelOutput key assignment and non-persistent buffer junk are unrelated." }, { "left": "issue:35707", "right": "issue:40444", "accept": false, - "reason": "Progressive generation with inputs_embeds/past_key_values is unrelated to multi-image iterable dataset finetuning." + "reason": "Progressive generation with inputs_embeds/past_key_values is unrelated to Qwen2.5-VL multi-image finetuning." }, { "left": "issue:43377", "right": "issue:45491", "accept": false, - "reason": "Both mention padding-related behavior, but MIMI encoder batch-vs-single mismatch and Gemma3 NaNs from all-padding sliding windows are different bugs." + "reason": "MIMI encoder padding-mask mismatch and Gemma3 sliding-window NaNs are different bugs." }, { "left": "issue:43792", "right": "issue:44589", "accept": false, - "reason": "Whisper runtime failure vs missing Float8 storage type; separate issues." + "reason": "Whisper runtime failure and Float8 storage lookup error are not the same problem." }, { "left": "issue:43482", "right": "issue:43525", "accept": false, - "reason": "Qwen2.5-GGUF loading regression vs missing pad_token_id on Llama4Config; distinct model/config bugs." + "reason": "GGUF loading failure and missing pad_token_id on Llama4Config are unrelated." }, { "left": "issue:44291", "right": "issue:44589", "accept": false, - "reason": "Init-empty-weights argument mismatch vs Float8 storage lookup failure; different failure modes." + "reason": "`init_empty_weights`/_is_hf_initialized and Float8 storage errors are different regressions." }, { "left": "issue:43749", "right": "issue:44079", "accept": false, - "reason": "FSDP CPU RAM efficient loading regression vs ModelOutput key assignment bug; unrelated." + "reason": "FSDP CPU RAM efficient loading and ModelOutput key assignment are unrelated." }, { "left": "issue:44079", "right": "issue:45072", "accept": false, - "reason": "ModelOutput key handling vs dtype mismatch in bfloat16 inference; different concerns." + "reason": "ModelOutput key handling and bfloat16 dtype mismatch are different issues." }, { "left": "issue:43644", "right": "issue:44079", "accept": false, - "reason": "Non-persistent buffer junk vs ModelOutput key assignment; no shared concrete bug." + "reason": "Non-persistent buffer junk and ModelOutput key assignment are unrelated." }, { - "left": "issue:44265", + "left": "issue:43632", "right": "issue:45072", "accept": false, - "reason": "torch.export/exportability issue vs bfloat16 dtype mismatch in inference; different code paths." + "reason": "Different Transformers v5 regressions: initialization flag handling vs dtype mismatch in inference." }, { - "left": "issue:43632", + "left": "issue:44265", "right": "issue:45072", "accept": false, - "reason": "_is_hf_initialized regression is unrelated to bfloat16 dtype mismatches." + "reason": "torch.export/torch_compilable_check failure and bfloat16 dtype mismatch are unrelated." }, { "left": "issue:40444", "right": "issue:45381", "accept": false, - "reason": "Both are Qwen2.5-VL, but one is multi-image finetuning and the other is video position-id logic." + "reason": "Both are Qwen2.5-VL, but one is multi-image finetuning failure and the other is video position-id corruption." }, { "left": "issue:45325", "right": "issue:45381", - "accept": false, - "reason": "Related area, but still-image temporal position scaling and video vision_position_ids are not the same concrete bug." + "accept": true, + "reason": "Both describe the same Qwen2.5-VL rope/position-id computation bug in transformers 5.3.0, just exposed by still-image vs video inputs." }, { "left": "issue:43531", "right": "issue:44291", "accept": false, - "reason": "Qwen3-MoE sliding_window issue vs init_empty_weights argument error; unrelated." + "reason": "Qwen3-MoE sliding_window behavior and `_is_hf_initialized` loading TypeError are unrelated." }, { "left": "issue:43611", "right": "issue:43646", "accept": false, - "reason": "Both are transformers-v5 loading regressions, but one is base_model_prefix handling and the other is custom initialization; too different to merge." + "reason": "Both are v5 loading/init regressions, but the titles point to different concrete failures and the same code path is not clear enough to merge." + }, + { + "left": "issue:44291", + "right": "issue:43632", + "accept": true, + "reason": "These are the same `_is_hf_initialized` / init-empty-weights Transformers v5 loading regression expressed from two angles." } ] }, - "evaluator_result": null, - "evaluator_used": false, + "evaluator_result": { + "accept": true, + "feedback": "The summary is conservative and grounded in the packet. The rejected pairs mostly align with distinct titles/failure modes, and the accepted pairs are plausibly the same bug family and mergeable into one fix. No clear overstatement beyond the evidence shown." + }, + "evaluator_used": true, "retried": false, - "accepted_nontrivial_soft_edge": false, + "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null }, @@ -3052,8 +3055,8 @@ "issue:45584" ], "soft_pairs": [ - "issue:43881|issue:44488", "issue:41720|issue:45081", + "issue:43881|issue:44488", "issue:42175|issue:43296", "issue:43454|issue:44466", "issue:43994|issue:44898", @@ -3070,27 +3073,27 @@ "issue:43881|issue:44991", "issue:43606|issue:44265", "issue:34634|issue:35707", - "issue:44186|issue:44265", "issue:44291|issue:44877", + "issue:44186|issue:44265", "issue:43606|issue:43819", "issue:43377|issue:44898", "issue:35707|issue:41720", - "issue:44610|issue:45325", "issue:43454|issue:45072", "issue:43454|issue:43550", + "issue:44610|issue:45325", "issue:43377|issue:44610", - "issue:39692|issue:40444", "issue:43749|issue:44534", + "issue:39692|issue:40444", "issue:44466|issue:45357", "issue:43296|issue:44991", "issue:43296|issue:44877" ], - "prepared_review_unit_hash": "3d656a95d7d743cd48d6a28579bd50fa695c36af79cefcdeb6019ad8d256285e", + "prepared_review_unit_hash": "90356b718f479d8f0ccfa85bebc20aec1d0225a4f4c99732d3939d6bc9c2a438", "component_budget": { "node_count": 367, "item_count": 367, "soft_pair_count": 1370, - "serialized_chars": 706987, + "serialized_chars": 706986, "estimated_input_tokens": 176747, "estimated_eval_tokens": 353750 }, @@ -3098,7 +3101,7 @@ "node_count": 48, "item_count": 48, "soft_pair_count": 32, - "serialized_chars": 37331, + "serialized_chars": 37330, "estimated_input_tokens": 9333, "estimated_eval_tokens": 18922 }, @@ -3107,210 +3110,204 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is largely heterogeneous and mostly consists of unrelated single-issue bug reports. The only plausible duplicate family is the offloading/quantization loading failures around issue 43873 and 43317; the rest look distinct by model, subsystem, or failure mode.", - "confidence": 0.62, - "canonical_issue_reason": "Issue 43873 is the broadest and most reusable anchor in the only overlapping family here: offloading/quantization failures. It can subsume 43317 as a specific device_map=auto reproduction.", + "summary": "This cluster is mostly a set of false-positive similarities around model loading/offloading, quantization, and a few unrelated vision/audio/tokenizer regressions. Only one pair looks like a plausible duplicate-level match: offloading with quantization/device placement.", + "confidence": 0.46, + "canonical_issue_reason": "issue:43873 is the broadest and most central issue in the only real overlap area (offloading + quantization/device placement), and it can naturally absorb the narrower repro in issue:43317.", "canonical_pr_reason": null, - "best_issue_reason": "43873 is the best global issue candidate because it is broad enough to represent the shared offloading/quantization bug family and is already framed as an umbrella problem.", + "best_issue_reason": "issue:43873 is the best representative issue overall: it is open, broadly scoped, and already has inbound references, while the other items are mostly narrower one-off bugs or unrelated regressions.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:43881", - "right": "issue:44488", + "left": "issue:41720", + "right": "issue:45081", "accept": false, - "reason": "Different models and different loading failures; no sign of the same underlying bug." + "reason": "Different bugs: device-map CUDA assert vs tokenizer regex patch crash." }, { - "left": "issue:41720", - "right": "issue:45081", + "left": "issue:43881", + "right": "issue:44488", "accept": false, - "reason": "One is a CUDA assert during Qwen3 auto device mapping, the other is a Mistral tokenizer patch crash; unrelated code paths." + "reason": "Both are load failures, but for different models and likely different root causes." }, { "left": "issue:42175", "right": "issue:43296", "accept": false, - "reason": "Backend packaging mismatch vs PaddleOCR-VL/vLLM model loading failure; clearly separate problems." + "reason": "One is backend/extras packaging, the other is a model loading failure in vLLM." }, { "left": "issue:43454", "right": "issue:44466", "accept": false, - "reason": "Both involve lm_head/tied weights, but one is a failed tie causing bad generation and the other is inconsistent serialization; different concrete bugs." + "reason": "Both mention lm_head/tied weights, but one is a missing tie causing bad generation and the other is device-dependent serialization; not the same concrete bug." }, { "left": "issue:43994", "right": "issue:44898", "accept": false, - "reason": "Different models and different failure modes; both are vision/model-output issues but not the same bug." + "reason": "Different vision-model regressions with different symptoms and code paths." }, { "left": "issue:44387", "right": "issue:44617", "accept": false, - "reason": "Both mention OOM, but one is int4 quantization memory regression and the other is a model-specific CUDA OOM; too broad to merge." + "reason": "Both are OOMs, but one is an int4 quantization memory regression and the other is a model-specific Sam3Video OOM." }, { "left": "issue:43828", "right": "issue:44683", "accept": false, - "reason": "Autocast dtype mismatch in Phi-tiny-MoE is unrelated to compiled flex_attention failures on newer torch." + "reason": "Autocast dtype mismatch and compiled flex_attention failure are unrelated." }, { "left": "issue:42915", "right": "issue:43482", "accept": false, - "reason": "Different models and different loading/configuration paths; no shared underlying defect." + "reason": "Different model families and different failure modes (FP8 config vs GGUF/v5 loading)." }, { "left": "issue:43526", "right": "issue:43883", "accept": false, - "reason": "Image processor label reduction bug vs missing all_tied_weights_keys attribute; unrelated." + "reason": "A label-reduction bug is not the same as a missing attribute during model loading." }, { "left": "issue:43701", "right": "issue:45325", "accept": false, - "reason": "Checkpoint key mismatch is unrelated to Qwen2.5-VL rope position scaling." + "reason": "Checkpoint resume key mismatch vs Qwen2.5-VL rope-index scaling are unrelated." }, { "left": "issue:43792", "right": "issue:43957", "accept": false, - "reason": "Both are model loading/runtime failures, but the titles point to different causes and different affected models." + "reason": "Whisper runtime failure and meta-device loading regression are different issues." }, { "left": "issue:43262", "right": "issue:43746", "accept": false, - "reason": "Audio processor sampling-rate defaulting and GraniteSpeech PEFT checkpoint loading are separate issues." + "reason": "Audio chat-template sample-rate default bug vs PEFT checkpoint loading bug." }, { "left": "issue:43632", "right": "issue:45563", "accept": false, - "reason": "Transformers v5 internal flag regression vs a stale warning in paged generate; not the same bug." + "reason": "_is_hf_initialized regression and stale paged-generate warning are unrelated." }, { "left": "issue:40990", "right": "issue:45584", "accept": false, - "reason": "Perplexity regression on GPT-OSS and Whisper empty-transcription generation failure are unrelated." + "reason": "Perplexity regression and empty-transcription generation failure are different." }, { "left": "issue:43317", "right": "issue:43873", "accept": true, - "reason": "Both describe offloading/quantization loading failures in the same general code path; 43873 is the broader umbrella and 43317 is a specific device_map=auto reproduction." + "reason": "Both describe offloading/device_map loading failures in the quantization/dequantization path and could plausibly be fixed together." }, { "left": "issue:43881", "right": "issue:44991", "accept": false, - "reason": "Different model load failures for different repositories; no shared root cause visible." + "reason": "Different model/tokenizer loading failures with no clear shared root cause." }, { "left": "issue:43606", "right": "issue:44265", "accept": false, - "reason": "CPU-offload device mismatch in Bark-small is unrelated to torch.export issues with torch_compilable_check." + "reason": "CPU offload device mismatch is unrelated to torch.export failures around torch_compilable_check." }, { "left": "issue:34634", "right": "issue:35707", "accept": false, - "reason": "BarkProcessor voice_preset bug and inputs_embeds/past_key_values generation bug are different areas and symptoms." + "reason": "Bark voice preset handling and progressive generation with inputs_embeds/past_key_values are unrelated." }, { - "left": "issue:44186", - "right": "issue:44265", + "left": "issue:44291", + "right": "issue:44877", "accept": false, - "reason": "Tokenizer padding/truncation crash vs export failure from torch_compilable_check; different failure surfaces." + "reason": "Unexpected _is_hf_initialized argument and strict granite_speech config loading are different bugs." }, { - "left": "issue:44291", - "right": "issue:44877", + "left": "issue:44186", + "right": "issue:44265", "accept": false, - "reason": "Both are loading/config regressions in v5, but one is init_empty_weights argument handling and the other is strict config validation; not the same bug." + "reason": "Tokenizer padding/NER crash is unrelated to torch.export compatibility." }, { "left": "issue:43606", "right": "issue:43819", "accept": false, - "reason": "Device mismatch in Bark CPU offload and DAC forward/from_latents mismatch are distinct audio-model bugs." + "reason": "CPU offload device mismatch and DAC.from_latents forward mismatch are different code paths." }, { "left": "issue:43377", "right": "issue:44898", "accept": false, - "reason": "MIMI padding-mask batch inconsistency is unrelated to Perceiver image classification resolution issues." + "reason": "Padding-mask batching bug and Perceiver image-size issue are unrelated." }, { "left": "issue:35707", "right": "issue:41720", "accept": false, - "reason": "Generation with inputs_embeds/past_key_values is unrelated to Qwen3 device-map CUDA asserts." - }, - { - "left": "issue:44610", - "right": "issue:45325", - "accept": false, - "reason": "Processor output size mismatch and Qwen2.5-VL temporal position scaling are different bugs." + "reason": "Progressive generation bug vs device-mapping CUDA assert are unrelated." }, { "left": "issue:43454", "right": "issue:45072", "accept": false, - "reason": "lm_head tying/serialization issues are unrelated to bfloat16 dtype mismatches in SwitchTransformers/TimmWrapperModel." + "reason": "Tied-weights serialization/generation bug vs bf16 dtype mismatch are different." }, { "left": "issue:43454", "right": "issue:43550", "accept": false, - "reason": "Tied-weights generation/serialization bug is unrelated to Bamba torch.compile SDPA failure." + "reason": "Tied-weights bug and torch.compile SDPA failure do not share a concrete root cause." }, { - "left": "issue:43377", - "right": "issue:44610", + "left": "issue:44610", + "right": "issue:45325", "accept": false, - "reason": "MIMI encoder padding-mask bug and OmDet-Turbo processor/model size mismatch are unrelated." + "reason": "Processor resolution mismatch and temporal rope-index scaling are different bugs." }, { - "left": "issue:39692", - "right": "issue:40444", + "left": "issue:43377", + "right": "issue:44610", "accept": false, - "reason": "SigLIP2 doc example errors and Qwen2.5-VL finetuning failures are different model-specific issues." + "reason": "Batched padding-mask behavior and processor/model input-size mismatch are unrelated." }, { "left": "issue:43749", "right": "issue:44534", "accept": false, - "reason": "FSDP CPU RAM efficient loading failure and non-persistent buffer junk initialization are separate regressions." + "reason": "FSDP RAM-efficient loading and non-persistent buffer initialization junk are different regressions." }, { - "left": "issue:44466", - "right": "issue:45357", + "left": "issue:39692", + "right": "issue:40444", "accept": false, - "reason": "Both involve serialization/loading in v5, but one is tied-weight serialization and the other is incorrect visual encoder keys; not the same bug." + "reason": "Doc-example model/processor mismatch and multi-image IterableDataset fine-tuning failure are not the same bug." }, { - "left": "issue:43296", - "right": "issue:44991", + "left": "issue:44466", + "right": "issue:45357", "accept": false, - "reason": "PaddleOCR-VL/vLLM loading failure is unrelated to tokenizer loading for EMBEDDIA/est-roberta." + "reason": "lm_head serialization issue and incorrect visual-encoder key saving are different save_pretrained regressions." }, { "left": "issue:43296", - "right": "issue:44877", + "right": "issue:44991", "accept": false, - "reason": "PaddleOCR-VL load failure and strict granite_speech config validation are unrelated." + "reason": "PaddleOCR-VL/vLLM loading failure and tokenizer loading for est-roberta are unrelated." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary stays grounded in the packet and is appropriately cautious. The only accepted duplicate-family link (43317 \u2194 43873) is supported by the shared offloading/quantization theme, and the other pairwise rejections are conservative rather than overstated." + "feedback": "Grounded overall and conservatively framed. The representative issue choice is supported by the packet (open, broader scope, inbound refs), and the one accepted soft-edge pair is plausibly the same offloading/quantization bug. The remaining pair rejections are cautious and do not overclaim shared root cause." }, "evaluator_used": true, "retried": false, @@ -3417,12 +3414,12 @@ "issue:36331|issue:41720", "issue:44514|issue:44545" ], - "prepared_review_unit_hash": "81ae8958578aabc55058b5cd2f3782a7d025d03ca305721a05b60af0f755f7ec", + "prepared_review_unit_hash": "c40d944a7ebd82310614d29f462d0d5ccaa5648806c978e1b27b59e8ad14481a", "component_budget": { "node_count": 367, "item_count": 367, "soft_pair_count": 1370, - "serialized_chars": 706987, + "serialized_chars": 706986, "estimated_input_tokens": 176747, "estimated_eval_tokens": 353750 }, @@ -3430,7 +3427,7 @@ "node_count": 48, "item_count": 48, "soft_pair_count": 35, - "serialized_chars": 38021, + "serialized_chars": 38022, "estimated_input_tokens": 9506, "estimated_eval_tokens": 19268 }, @@ -3439,228 +3436,228 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a grab-bag of unrelated Transformers bug reports across Qwen, multimodal serving, configs, offloading, tokenizers, and trainer behavior. Only one clear duplicate pair stands out: the two Qwen2_5_VLProcessor apply_chat_template batched-padding reports.", - "confidence": 0.98, - "canonical_issue_reason": "issue:44545 is the cleanest representative of the only true duplicate pair in this cluster; it matches issue:44514 on the same batched padding=False crash and has the clearer canonical title.", + "summary": "This cluster is mostly heterogeneous, but one near-exact duplicate pair stands out: `Qwen2_5_VLProcessor.apply_chat_template` crashes on batched input when `padding=False` (issues 44514 and 44545). The rest are related at most by broad themes like loading/offloading, quantization, and model-specific regressions, but not the same underlying bug.", + "confidence": 0.77, + "canonical_issue_reason": "Issue 44545 is the clearest canonical issue: it is an exact duplicate of 44514 with a cleaner title and the same underlying batched `apply_chat_template` padding=False crash.", "canonical_pr_reason": null, - "best_issue_reason": "issue:44545 is the best global issue representative here because it is the clearest canonical example of the only duplicate pair; the rest of the cluster is too heterogeneous to unify.", + "best_issue_reason": "Issue 44545 is the best issue to keep as the representative duplicate target because it most directly names the concrete failing code path and matches the only clear duplicate pair.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43454", "right": "issue:43720", "accept": false, - "reason": "Different bugs: lm_head weight tying vs packed-weight unpacking during Accelerate loading." + "reason": "Both are model-loading/weight-handling bugs, but they involve different models and different failure modes (`lm_head` tying vs packed-weight unpacking)." }, { "left": "issue:41720", "right": "issue:43322", "accept": false, - "reason": "Different failure modes and code paths: Qwen3 auto device mapping CUDA assert vs Llava Next segfault." + "reason": "Both mention severe runtime failures, but one is Qwen3 auto device mapping on A800 and the other is Llava Next loading segfault; no shared concrete bug." }, { "left": "issue:43792", "right": "issue:44683", "accept": false, - "reason": "Whisper runtime failure and torch>=2.9 flex_attention compilation regression are unrelated." + "reason": "Different subsystems and root causes: Whisper loading/run failure vs compiled flex_attention on torch >= 2.9." }, { "left": "issue:41720", "right": "issue:44977", "accept": false, - "reason": "Both involve Qwen models, but one is a device-mapping CUDA assert and the other is a flash-attention generation regression." + "reason": "Both involve Qwen3-family generation/runtime issues, but one is a CUDA assert during auto device mapping and the other is flash-attention generation behavior." }, { "left": "issue:43404", "right": "issue:45357", "accept": false, - "reason": "Both are model-weight issues, but Mistral3 lm_head tying and Qwen3.5 visual-encoder key saving are distinct." + "reason": "Both are weight-tying/serialization problems, but they affect different models and different assets (Mistral3 lm_head vs Qwen3.5 visual encoder keys)." }, { "left": "issue:44423", "right": "issue:45406", "accept": false, - "reason": "Different serve crashes: a multimodal '.to()' path bug vs Gemma4Processor missing _tokenizer." + "reason": "Different crash sites: continuous-batching multimodal AttributeError vs serve crash from missing `_tokenizer` in `Gemma4Processor`." }, { "left": "issue:44186", "right": "issue:45072", "accept": false, - "reason": "Tokenizer padding/NER crash is unrelated to bf16 dtype mismatches in model inference." + "reason": "Both are dtype/input handling bugs, but one is tokenizer NER padding/truncation and the other is inference dtype mismatches in wrappers." }, { "left": "issue:43526", "right": "issue:43881", "accept": false, - "reason": "BeitImageProcessor label reduction bug and glm-4v-9b load failure are different problems." + "reason": "Different components and symptoms: BeitImageProcessorFast label reduction vs glm-4v-9b loading failure." }, { "left": "issue:43873", "right": "issue:44756", "accept": false, - "reason": "Both mention offloading, but one is quantization/offload behavior and the other is mmap-related OOM on Strix Halo." + "reason": "Both discuss offloading/memory behavior, but one is quantization offloading and the other is mmap/OOM on Strix Halo; not the same bug." }, { "left": "issue:43577", "right": "issue:43708", "accept": false, - "reason": "Blip2 dtype loading bug vs Trainer resume/max_steps calculation bug." + "reason": "Different trainer/config issues: BLIP2 dtype propagation vs resume_from_checkpoint max_steps calculation." }, { "left": "issue:42617", "right": "issue:42915", "accept": false, - "reason": "3d_parallel execution failure and Qwen3Moe FineGrainedFP8Config failure are unrelated." + "reason": "Both are runtime failures, but one is running `3d_parallel.py` and the other is Qwen3Moe with FineGrainedFP8Config; no shared code path." }, { "left": "issue:43550", "right": "issue:43720", "accept": false, - "reason": "torch.compile/SDPA regression is not the same as packed-weight unpacking during loading." + "reason": "Different issues in acceleration/compilation vs loading packed weights; too broad to be the same bug." }, { "left": "issue:44610", "right": "issue:44898", "accept": false, - "reason": "Both are vision input-size issues, but they affect different models and different preprocessing expectations." + "reason": "Both concern image-size/preprocessing mismatches, but they affect different models and different expected resolutions." }, { "left": "issue:44877", "right": "issue:44991", "accept": false, - "reason": "Strict config loading and tokenizer loading regression are distinct issues." + "reason": "Both are config/tokenizer loading regressions, but one is strict config for granite_speech and the other is tokenizer loading for a specific model." }, { "left": "issue:44479", "right": "issue:44918", "accept": false, - "reason": "Video-input regression in Qwen VL models is different from TRL SFT embedding unpacking." + "reason": "Both mention Qwen3.5-family regressions, but one is video-input regression and the other is SFT unpacking of input embeddings." }, { "left": "issue:44186", "right": "issue:44610", "accept": false, - "reason": "LayoutLMv2 tokenizer batching crash is unrelated to OmDet-Turbo image-size mismatch." + "reason": "Different model families and bugs: LayoutLMv2Tokenizer crashes vs OmDet-Turbo processor output size mismatch." }, { "left": "issue:36331", "right": "issue:42915", "accept": false, - "reason": "Trainer API break and Qwen3Moe FP8 config failure do not share a code path." + "reason": "Trainer custom loss signature mismatch vs Qwen3Moe FP8 failure; unrelated failures." }, { "left": "issue:36331", "right": "issue:42617", "accept": false, - "reason": "CustomTrainer.compute_loss signature change is unrelated to 3d_parallel.py failing." + "reason": "Unexpected `num_items_in_batch` argument is a trainer API mismatch, not related to 3d_parallel execution failure." }, { "left": "issue:44671", "right": "issue:44743", "accept": false, - "reason": "CamemBERT masked-LM predictions and Qwen3.5 recurrent-state reset are different bugs." + "reason": "Different Qwen3.5 areas: CamemBERT masked LM predictions vs recurrent states resetting in modular_qwen3_5.py." }, { "left": "issue:43316", "right": "issue:44913", "accept": false, - "reason": "Config API discrepancy and GPTNeoX rotary_pct reload bug are not the same underlying issue." + "reason": "Both are config reload discrepancies, but they involve different fields and model families (`Gemma3TextConfig` vs GPTNeoX rotary_pct)." }, { "left": "issue:43550", "right": "issue:43606", "accept": false, - "reason": "Different problems: torch.compile SDPA regression vs CPU offload device mismatch." + "reason": "Both involve device/compile/offload behavior, but one is SDPA with torch.compile and the other is CPU offload device mismatch for bark-small." }, { "left": "issue:42915", "right": "issue:43317", "accept": false, - "reason": "Qwen3Moe FineGrainedFP8Config failure and dequantized-model offload loading are distinct offload/config bugs." + "reason": "Both are loading/offload-related, but one is Qwen3Moe FP8 failure and the other is device_map=auto dequantized model offload failure." }, { "left": "issue:41720", "right": "issue:42617", "accept": false, - "reason": "Qwen3 auto device mapping assert and 3d_parallel runner failure are unrelated." + "reason": "Both are execution failures, but one is CUDA assert on auto device mapping and the other is inability to run `3d_parallel.py`; not the same bug." }, { "left": "issue:44292", "right": "issue:44479", "accept": false, - "reason": "Qwen-3 NVFP4 runtime issue and Qwen VL video-input regression are separate bugs." + "reason": "Different Qwen variants and different failure modes: NVFP4 run error vs video-input regression." }, { "left": "issue:41720", "right": "issue:45161", "accept": false, - "reason": "Auto device mapping CUDA assert is not the same as GPT-OSS MoE tensor-parallel failure." + "reason": "Both touch parallelism/device placement, but one is a Qwen3 auto device-map crash and the other is only TP not working in GPT-OSS MoE." }, { "left": "issue:43531", "right": "issue:43572", "accept": false, - "reason": "Qwen3-MoE sliding-window behavior and StableLmConfig pad_token_idx regression are unrelated." + "reason": "Different model/config issues: Qwen3-MoE sliding window vs StableLmConfig missing pad_token_idx." }, { "left": "issue:44186", "right": "issue:44361", "accept": false, - "reason": "LayoutLMv2 tokenizer crash and MLuke tokenizer AttributeError are different tokenizer bugs." + "reason": "Both are tokenizer/input-processing bugs, but they affect different tokenizers and different crash conditions." }, { "left": "issue:43388", "right": "issue:45059", "accept": false, - "reason": "Metric label dropping and SAM3 text/bbox odd behavior do not share a code path." + "reason": "Different multimodal/data-prep issues: metric gathering drops labels vs SAM3 text/bounding-box behavior." }, { "left": "issue:43540", "right": "issue:43782", "accept": false, - "reason": "Video processing ValueError and Qwen3VL weight_only loading error are different failures." + "reason": "Different Qwen3 Omni/MoE input-processing and from_pretrained weight_only error; not the same underlying bug." }, { "left": "issue:41720", "right": "issue:43873", "accept": false, - "reason": "Device-mapping CUDA assert is not the same as the quantization/offloading bug in 43873." + "reason": "Both involve Qwen3 and device/offload concerns, but one is a CUDA assert on A800 and the other is quantization offloading behavior; not mergeable as one fix." }, { "left": "issue:42915", "right": "issue:43322", "accept": false, - "reason": "Qwen3Moe FP8 failure and Llava Next segmentation fault are unrelated." + "reason": "Different model paths and failure types: FineGrainedFP8Config failure vs segmentation fault when loading Llava Next." }, { "left": "issue:36010", "right": "issue:42617", "accept": false, - "reason": "ImportError for GenerationMixin and 3d_parallel.py failure are unrelated." + "reason": "Completely different problems: import path regression vs inability to run `3d_parallel.py`." }, { "left": "issue:45406", "right": "issue:45464", "accept": false, - "reason": "Gemma4Processor serve crash and Qwen3.5 streaming inference failure are different server bugs." + "reason": "Both affect serving, but one is a missing `_tokenizer` on `Gemma4Processor` and the other is streaming chat/completions on Qwen3.5-0.8B." }, { "left": "issue:36331", "right": "issue:41720", "accept": false, - "reason": "Trainer compute_loss API change and Qwen3 auto device mapping assert are unrelated." + "reason": "Trainer loss-signature API mismatch is unrelated to Qwen3 auto device-map CUDA failure." }, { "left": "issue:44514", "right": "issue:44545", "accept": true, - "reason": "Same title, same Qwen2_5_VLProcessor apply_chat_template batched padding=False crash; clear duplicate pair." + "reason": "Exact same crash and API path; the titles/body describe the same batched `Qwen2_5_VLProcessor.apply_chat_template` padding=False bug, with only minor wording differences." } ] }, "evaluator_result": { "accept": true, - "feedback": "Grounded overall. The only accepted duplicate pair in the packet is indeed issue:44514 / issue:44545, and the rejection reasons for the soft-edge pairs stay conservative. The summary does not overclaim a larger unification beyond that one clear duplicate pair." + "feedback": "The summary is grounded in the packet and stays conservative. The only accepted duplicate pair, 44514/44545, is well supported by the near-identical titles and the shared `Qwen2_5_VLProcessor.apply_chat_template` padding=False crash. The remaining soft-edge verdicts are appropriately cautious and do not overstate similarity." }, "evaluator_used": true, "retried": false, @@ -3751,23 +3748,23 @@ "issue:44451|issue:44991", "issue:44062|issue:44451", "issue:44987|issue:45042", - "issue:44918|issue:44936", "issue:43531|issue:44589", + "issue:44918|issue:44936", "issue:43994|issue:44671", "issue:39692|issue:44610", "issue:43421|issue:44186", "issue:45081|issue:45356", "issue:43525|issue:43531", - "issue:45183|issue:45479", "issue:39692|issue:43873", + "issue:45183|issue:45479", "issue:43746|issue:45381" ], - "prepared_review_unit_hash": "b8c1265a2d0d30a979591bf8bd19b3c8ffceb792714de41f48859e489eb2fd5c", + "prepared_review_unit_hash": "27bbba0e0dac76012521bdd0bbd6a1b16f2717e0fd4bd2e4bd4b22c7114c4e41", "component_budget": { "node_count": 367, "item_count": 367, "soft_pair_count": 1370, - "serialized_chars": 706987, + "serialized_chars": 706986, "estimated_input_tokens": 176747, "estimated_eval_tokens": 353750 }, @@ -3775,214 +3772,217 @@ "node_count": 47, "item_count": 47, "soft_pair_count": 31, - "serialized_chars": 36687, - "estimated_input_tokens": 9172, - "estimated_eval_tokens": 18600 + "serialized_chars": 36694, + "estimated_input_tokens": 9174, + "estimated_eval_tokens": 18604 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These items are overwhelmingly distinct bugs/features across different models and code paths; none of the soft pairs look like true duplicates. No PRs are present.", - "confidence": 0.93, - "canonical_issue_reason": "Issue 44843 is the most concrete, actionable open bug with a clear code-path description (_patch_mistral_regex / offline mode) and a strong reproduction-oriented report.", + "summary": "This cluster is mostly heterogeneous: tokenizer/load regressions, model-specific inference bugs, processor mismatches, and training issues. Only one soft edge looks like the same underlying bug: the offline tokenizer/cache failure tied to _patch_mistral_regex.", + "confidence": 0.67, + "canonical_issue_reason": "issue:44843 is the clearest root-cause report; it names the concrete failing code path (_patch_mistral_regex calling model_info() unconditionally), which directly explains the offline-cache symptom in issue:42886.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 44843 is the best representative issue here because it is specific, current, and directly points to an implementation fault rather than a vague symptom.", + "best_issue_reason": "issue:44843 is the most actionable representative of the only real duplicate subcluster, because it identifies the exact offending call site and the HF_HUB_OFFLINE breakage.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:33357", "right": "issue:35707", "accept": false, - "reason": "MacOS CLIP bus error and progressive generation with inputs_embeds/past_key_values are unrelated failures." + "reason": "Different failures in unrelated paths: MacOS bus error on a CLIP model vs. progressive generation with inputs_embeds/past_key_values." }, { "left": "issue:43606", "right": "issue:43720", "accept": false, - "reason": "Both involve accelerate/offload, but Bark CPU offload device mismatch and BitNet packed-weight unpacking are different code paths." + "reason": "Both involve loading/offload context, but they are different bugs with different models and failure modes (CPU offload device mismatch vs. packed weights not unpacked)." }, { "left": "issue:41720", "right": "issue:43746", "accept": false, - "reason": "Qwen3 auto-device-map cuda assert and GraniteSpeech PEFT local-checkpoint loading are separate model/load bugs." + "reason": "Different model families and root causes: Qwen3 device-mapping cuda assert vs. GraniteSpeech PEFT local checkpoint loading." }, { "left": "issue:30064", "right": "issue:34634", "accept": false, - "reason": "ImageProcessor void segmentation maps and BarkProcessor voice_preset are unrelated processor issues." + "reason": "Unrelated processor bugs: void segmentation maps vs. BarkProcessor voice_preset handling." }, { "left": "issue:43550", "right": "issue:43653", "accept": false, - "reason": "Bamba torch.compile/SDPA failure and BigBirdTokenizer special-token registration are different subsystems and bugs." + "reason": "Different subsystems entirely: torch.compile/SDPA on Bamba vs. BigBirdTokenizer special-token registration." }, { "left": "issue:44492", "right": "issue:44509", "accept": false, - "reason": "A typo in cache strategies is not the same as docs still mentioning removed pipeline tasks." + "reason": "A typo in cache strategy docs is not the same issue as stale pipeline-task documentation." }, { "left": "issue:44368", "right": "issue:45357", "accept": false, - "reason": "Qwen3.5 tie_word_embeddings warning and incorrect visual encoder keys are different save/load and config issues." + "reason": "One is a warning about tie_word_embeddings during LoRA fine-tuning; the other is a save_pretrained regression for Qwen3.5 visual encoder keys." }, { "left": "issue:44625", "right": "issue:45479", "accept": false, - "reason": "num_labels propagation and zero-loss single-label classification are related only loosely; the bugs are distinct." + "reason": "Both mention classification config, but one is Qwen3.5 num_labels propagation and the other is a generic zero-loss bug for num_labels=1." }, { "left": "issue:43606", "right": "issue:43653", "accept": false, - "reason": "Bark CPU offload device mismatch and BigBirdTokenizer mask-token decoding are unrelated bugs." + "reason": "Different bugs: offloading/device mismatch vs. tokenizer special-token decoding." }, { "left": "issue:43421", "right": "issue:43927", "accept": false, - "reason": "Runtime post-processor updates and save/load of custom token IDs are different tokenizer lifecycle problems." + "reason": "Both are tokenizer/config related, but one is runtime post-processor updates and the other is save/load losing custom token IDs." }, { "left": "issue:44898", "right": "issue:45072", "accept": false, - "reason": "Perceiver image-size handling and bfloat16 dtype mismatches are separate inference failures." + "reason": "Different model/runtime problems: Perceiver image-size handling vs. dtype mismatch in bfloat16 inference." }, { "left": "issue:43782", "right": "issue:45381", "accept": false, - "reason": "Qwen3VL weight_only load error and Qwen2.5-VL video position IDs are different VL bugs." + "reason": "Both are Qwen/VL-related, but one is a from_pretrained weight_only error and the other is a video vision_position_ids regression." }, { "left": "issue:34567", "right": "issue:35707", "accept": false, - "reason": "TrainerState token-count tracking and progressive generation with past_key_values are unrelated." + "reason": "TrainerState token counting is unrelated to progressive generation with embeddings/past key values." }, { "left": "issue:34634", "right": "issue:41720", "accept": false, - "reason": "BarkProcessor voice_preset and Qwen3 auto device mapping are different model/runtime issues." + "reason": "BarkProcessor voice_preset and Qwen3 device-mapping are unrelated bugs." }, { "left": "issue:43381", "right": "issue:43844", "accept": false, - "reason": "Eval-mode checkpointing restriction and ZeRO-3 gradient explosion are not the same underlying bug." + "reason": "One is about checkpointing in eval mode; the other is a training-gradient anomaly with ZeRO-3." }, { "left": "issue:42886", "right": "issue:44843", - "accept": false, - "reason": "Both mention offline/tokenizer loading, but 44843 is a specific mistral-regex network call while 42886 is a broader cache-load failure." + "accept": true, + "reason": "Same underlying offline tokenizer/cache regression: 42886 reports the symptom, while 44843 pinpoints the root cause in _patch_mistral_regex/model_info()." }, { "left": "issue:43421", "right": "issue:43653", "accept": false, - "reason": "Special-token post-processor updates and BigBird mask-token registration are different tokenizer issues." + "reason": "Different tokenizer issues: runtime special-token updates vs. mask token registration/decoding." }, { "left": "issue:44077", "right": "issue:44625", "accept": false, - "reason": "patchtsmixer post_init policy and Qwen3.5 num_labels propagation are unrelated." + "reason": "Different problems: patchtsmixer post_init policy vs. Qwen3.5 num_labels propagation." }, { "left": "issue:44451", "right": "issue:44991", "accept": false, - "reason": "These are tokenizer/model-load failures for different models and likely different regressions." + "reason": "Both are tokenizer loading failures, but they affect different models and don\u2019t show the same root cause." }, { "left": "issue:44062", "right": "issue:44451", "accept": false, - "reason": "AddedToken special-arg duplication is not the same as a model/tokenizer load failure for ScandiBERT." + "reason": "Different failures: AddedToken special-argument conflict vs. model loading failure for ScandiBERT." }, { "left": "issue:44987", "right": "issue:45042", "accept": false, - "reason": "A model-loading regression and a PIL processor torchvision dependency bug are different classes of issue." + "reason": "Unrelated: loading a specific model vs. PIL image processors wrongly requiring torchvision." }, { - "left": "issue:44918", - "right": "issue:44936", + "left": "issue:43531", + "right": "issue:44589", "accept": false, - "reason": "SFT embedding unpacking and train-then-evaluate failure do not share a concrete code path." + "reason": "Different issues: sliding_window behavior in Qwen3-MoE vs. missing Float8 storage type." }, { - "left": "issue:43531", - "right": "issue:44589", + "left": "issue:44918", + "right": "issue:44936", "accept": false, - "reason": "Qwen3-MoE sliding_window behavior and Float8 storage lookup are unrelated." + "reason": "Different trainer failures: unpacking Qwen3.5 embeddings in TRL SFT vs. evaluate() after train()." }, { "left": "issue:43994", "right": "issue:44671", "accept": false, - "reason": "SigLIP2 nonsensical outputs and CamemBERT MLM regressions affect different models and paths." + "reason": "Both are wrong-model-output complaints, but they are model-specific and not the same bug." }, { "left": "issue:39692", "right": "issue:44610", "accept": false, - "reason": "SigLIP2 docs/example mismatch and OmDet-Turbo 640px vs 224px processor mismatch are separate issues." + "reason": "Both mention processor/model mismatch, but one is a documentation example with multiple errors and the other is a specific OmDet-Turbo image-size mismatch." }, { "left": "issue:43421", "right": "issue:44186", "accept": false, - "reason": "Tokenizer post-processor updates and LayoutLMv2Tokenizer crashes are different bugs." + "reason": "Different tokenizer problems: post-processor updates vs. LayoutLMv2 NER/padding crashes." }, { "left": "issue:45081", "right": "issue:45356", "accept": false, - "reason": "Both touch fix_mistral_regex, but one crashes on backend_tokenizer access and the other is a Kimi-K2.5 codec regression; not the same bug." + "reason": "Both mention fix_mistral_regex/tokenizer regressions, but the symptoms and affected models differ enough that they do not clearly look like the same underlying bug." }, { "left": "issue:43525", "right": "issue:43531", "accept": false, - "reason": "Llama4Config pad_token_id and Qwen3-MoE sliding_window are unrelated config/model problems." + "reason": "Llama4Config pad_token_id and Qwen3-MoE sliding_window are unrelated configuration bugs." }, { - "left": "issue:45183", - "right": "issue:45479", + "left": "issue:39692", + "right": "issue:43873", "accept": false, - "reason": "Input validation error messaging and single-label zero-loss behavior are different problems." + "reason": "Documentation/model-processor example errors are unrelated to offloading with quantization." }, { - "left": "issue:39692", - "right": "issue:43873", + "left": "issue:45183", + "right": "issue:45479", "accept": false, - "reason": "SigLIP2 example issues and quantization/offloading behavior are unrelated." + "reason": "Both are broad validation/config bugs, but one is about input_ids/inputs_embeds error messaging and the other is a degenerate loss issue for num_labels=1." }, { "left": "issue:43746", "right": "issue:45381", "accept": false, - "reason": "GraniteSpeech PEFT checkpoint loading and Qwen2.5-VL video position IDs are different code paths." + "reason": "Different Qwen/GraniteSpeech failures with no shared concrete code path." } ] }, - "evaluator_result": null, - "evaluator_used": false, + "evaluator_result": { + "accept": true, + "feedback": "Grounded in the packet and conservative overall. The only accepted soft edge, issue:42886 \u2194 issue:44843, is well supported by the titles and the described offline/cache regression. The rejected edges are phrased cautiously and do not overclaim duplicate status." + }, + "evaluator_used": true, "retried": false, - "accepted_nontrivial_soft_edge": false, + "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null }, @@ -4051,8 +4051,8 @@ "soft_pairs": [ "issue:43596|issue:44220", "issue:43746|issue:43873", - "issue:43618|issue:43697", "issue:43540|issue:44008", + "issue:43618|issue:43697", "issue:40444|issue:43746", "issue:44220|issue:44661", "issue:43716|issue:44008", @@ -4065,28 +4065,28 @@ "issue:30064|issue:36010", "issue:43550|issue:45072", "issue:43577|issue:43818", - "issue:43550|issue:43927", "issue:44062|issue:44488", + "issue:43550|issue:43927", "issue:43540|issue:45381", "issue:43643|issue:44877", "issue:43577|issue:43582", "issue:43531|issue:43643", "issue:43782|issue:44977", "issue:43644|issue:43749", - "issue:43653|issue:43819", "issue:44448|issue:44671", + "issue:43653|issue:43819", "issue:44561|issue:45020", "issue:43825|issue:43994", "issue:43550|issue:43819", "issue:44186|issue:44442", "issue:43825|issue:43827" ], - "prepared_review_unit_hash": "0dd42a3d8cf6a65b165b5816b2047787a6f4e4fa6c8e239931f4d09c9adbc36d", + "prepared_review_unit_hash": "62c85041d11196b4d996beacf2ab6b3fb56dd5e40c61c64dd408d7752077b034", "component_budget": { "node_count": 367, "item_count": 367, "soft_pair_count": 1370, - "serialized_chars": 706987, + "serialized_chars": 706986, "estimated_input_tokens": 176747, "estimated_eval_tokens": 353750 }, @@ -4094,7 +4094,7 @@ "node_count": 48, "item_count": 48, "soft_pair_count": 31, - "serialized_chars": 37218, + "serialized_chars": 37219, "estimated_input_tokens": 9305, "estimated_eval_tokens": 18866 }, @@ -4103,198 +4103,198 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly heterogeneous; the soft-similarity pairs share broad subsystems or recurring transformers themes, but they do not look like the same underlying bug/change. I would not merge any of the proposed soft edges.", - "confidence": 0.62, - "canonical_issue_reason": "issue:45020 is the broadest current issue in the set and is the best umbrella for a family of recent-version / remote_code regressions, even though the overall cluster is too diverse to treat as one duplicate group.", + "summary": "This cluster is not a true duplicate set; it\u2019s a heterogeneous bucket of unrelated GitHub issues across tokenizers, multimodal models, loading/config paths, and v5 regressions. I rejected all soft edges because each pair only shares broad keywords or subsystem overlap, not the same concrete bug.", + "confidence": 0.97, + "canonical_issue_reason": "No single canonical issue stands out: the items span many unrelated failures, so there is no stable duplicate root to anchor the cluster.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45020 is the strongest representative issue: open, broad in scope, and closest to the recurring theme of recent transformers regressions affecting remote_code models.", + "best_issue_reason": "None of the issues is a good global representative; the similarity links are mostly superficial rather than evidence of one underlying defect.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43596", "right": "issue:44220", "accept": false, - "reason": "Different subsystems and failures: Deepspeed zero3/BertModel indexing vs audio fbank feature extraction." + "reason": "Both mention crashes, but one is DeepSpeed/BertModel init and the other is audio fbank feature extraction; different code paths and bugs." }, { "left": "issue:43746", "right": "issue:43873", "accept": false, - "reason": "One is a local checkpoint load failure for GraniteSpeech PEFT adapters; the other is a quantization/offloading behavior report." + "reason": "One is local checkpoint loading with PEFT adapters, the other is offloading under quantization; same broad loading area but not the same failure." }, { - "left": "issue:43618", - "right": "issue:43697", + "left": "issue:43540", + "right": "issue:44008", "accept": false, - "reason": "CLIP attentions regression vs RTDetrV2 output drift in v5; different models and code paths." + "reason": "Different multimodal models and different bugs: video input ValueError vs a Gemma3n variable-name collision in forward()." }, { - "left": "issue:43540", - "right": "issue:44008", + "left": "issue:43618", + "right": "issue:43697", "accept": false, - "reason": "Qwen3OmniMoe video-input ValueError vs Gemma3n variable-name collision in forward(); unrelated bugs." + "reason": "CLIP attentions assignment regression and RTDetrV2 output drift are unrelated model-specific issues." }, { "left": "issue:40444", "right": "issue:43746", "accept": false, - "reason": "Qwen2.5-VL multi-image finetuning failure is unrelated to GraniteSpeech PEFT checkpoint loading." + "reason": "IterableDataset multi-image finetuning failure is unrelated to GraniteSpeech local checkpoint loading." }, { "left": "issue:44220", "right": "issue:44661", "accept": false, - "reason": "Audio feature extraction bug vs tokenizer-mapping model-addition failure; no common underlying defect." + "reason": "Audio feature extraction failure and tokenizer-mapping/model addition failure are different subsystems." }, { "left": "issue:43716", "right": "issue:44008", "accept": false, - "reason": "Mistral image-preprocessor dtype mismatch vs Gemma3n audio tensor attribute collision; different models and symptoms." + "reason": "Mistral-3 image preprocessor dtype mismatch is not the same bug as Gemma3n\u2019s audio_mask attribute collision." }, { "left": "issue:43720", "right": "issue:43819", "accept": false, - "reason": "Packed-weight loading bug in BitNet vs DAC latent/STE mismatch; distinct code paths." + "reason": "BitNet packed-weight loading and DAC.from_latents/STE mismatch are distinct model implementations and defects." }, { "left": "issue:43335", "right": "issue:43756", "accept": false, - "reason": "SwitchTransformers sparse-layer construction bug and Smollm3 RoPE layer dropping are unrelated." + "reason": "SwitchTransformers sparse-layer config bug and Smollm3 RoPE-layer dropping are unrelated configuration/model issues." }, { "left": "issue:43653", "right": "issue:43720", "accept": false, - "reason": "BigBirdTokenizer special-token registration bug vs BitNet packed-weight loading bug; different components." + "reason": "Tokenizer special-token registration and BitNet accelerate loading are unrelated." }, { "left": "issue:44843", "right": "issue:45081", "accept": false, - "reason": "Both involve _patch_mistral_regex, but one is an offline model_info call and the other is a backend_tokenizer AttributeError; same helper, different bugs." + "reason": "Both touch _patch_mistral_regex, but one is an offline model_info call and the other is a backend_tokenizer AttributeError; different concrete failures." }, { "left": "issue:43421", "right": "issue:44811", "accept": false, - "reason": "Runtime post-processor update bug vs Whisper batch_decode skip_special_tokens handling; not the same issue." + "reason": "Runtime special-token/post-processor update behavior is not the same as Whisper batch_decode ignoring skip_special_tokens." }, { "left": "issue:44186", "right": "issue:44898", "accept": false, - "reason": "LayoutLMv2 tokenizer padding/NER crash vs Perceiver non-default-resolution image classification failure." + "reason": "LayoutLMv2 tokenizer padding/NER crash and Perceiver image-classification failure are unrelated." }, { "left": "issue:30064", "right": "issue:36010", "accept": false, - "reason": "Image processor void-segmentation-map handling is unrelated to the GenerationMixin import error." + "reason": "Void segmentation maps in image processing and a GenerationMixin import error have no shared code path." }, { "left": "issue:43550", "right": "issue:45072", "accept": false, - "reason": "torch.compile SDPA failure in Bamba-9B-v2 vs dtype mismatch in SwitchTransformers/TimmWrapperModel; different bugs." + "reason": "Bamba torch.compile with SDPA and SwitchTransformers/TimmWrapperModel dtype mismatches are different bugs." }, { "left": "issue:43577", "right": "issue:43818", "accept": false, - "reason": "Blip2 dtype propagation issue vs Video-LLaVA temporal-attention/weight-sharing issue; unrelated." + "reason": "BLIP2 dtype propagation and Video-LLaVA temporal-attention/weight-sharing issues are unrelated." }, { - "left": "issue:43550", - "right": "issue:43927", + "left": "issue:44062", + "right": "issue:44488", "accept": false, - "reason": "Compile/SDPA failure vs DiaConfig custom token ID loss after save/load; different failure modes." + "reason": "AddedToken duplicate special kwarg and failing to load a specific model are not the same defect." }, { - "left": "issue:44062", - "right": "issue:44488", + "left": "issue:43550", + "right": "issue:43927", "accept": false, - "reason": "AddedToken constructor argument clash vs model-loading failure for cjvt/sleng-bert; not the same defect." + "reason": "Bamba compile/SDPA failure and DiaConfig save/load token-id loss are unrelated." }, { "left": "issue:43540", "right": "issue:45381", "accept": false, - "reason": "Different Qwen multimodal/video bugs affecting different models and code paths." + "reason": "Both are video-related Qwen issues, but they affect different model versions and different bugs (input processing vs vision_position_ids)." }, { "left": "issue:43643", "right": "issue:44877", "accept": false, - "reason": "trust_remote_code config field loss vs strict config rejection for granite_speech; related area but distinct issues." + "reason": "AutoConfig trust_remote_code field loss and strict config blocking granite_speech loading are separate config-loading problems." }, { "left": "issue:43577", "right": "issue:43582", "accept": false, - "reason": "BLIP2 dtype mismatch vs Apple Silicon caching_allocator_warmup TypeError; unrelated." + "reason": "BLIP2 dtype mismatch and Apple Silicon caching_allocator_warmup TypeError are unrelated." }, { "left": "issue:43531", "right": "issue:43643", "accept": false, - "reason": "Qwen3-MoE sliding_window bug and remote_code config field loss do not share an underlying code change." + "reason": "Qwen3-MoE sliding_window behavior and trust_remote_code AutoConfig field loss are different issues." }, { "left": "issue:43782", "right": "issue:44977", "accept": false, - "reason": "Qwen3VL from_pretrained weight_only error vs Qwen3.5 flash-attention generation bug; different model paths." + "reason": "Qwen3VL weight_only loading error and Qwen3.5 flash-attention generation failure are different model/path problems." }, { "left": "issue:43644", "right": "issue:43749", "accept": false, - "reason": "Non-persistent buffer corruption in Transformers 5.0 vs FSDP CPU RAM efficient loading breakage; not duplicates." + "reason": "Non-persistent buffer junk in Transformers 5.0 and FSDP CPU RAM efficient loading are distinct loading regressions." }, { - "left": "issue:43653", - "right": "issue:43819", + "left": "issue:44448", + "right": "issue:44671", "accept": false, - "reason": "Tokenizer special-token registration bug vs DAC latent reconstruction mismatch; unrelated." + "reason": "Pegasus v4/v5 output drift and CamemBERT masked-LM prediction regressions are unrelated." }, { - "left": "issue:44448", - "right": "issue:44671", + "left": "issue:43653", + "right": "issue:43819", "accept": false, - "reason": "Pegasus v4/v5 output drift and CamemBERT masked-LM regressions are separate model-specific issues." + "reason": "Tokenizer special-token registration and DAC STE mismatch are unrelated bugs." }, { "left": "issue:44561", "right": "issue:45020", "accept": false, - "reason": "Both mention remote_code regressions, but 44561 is specifically about is_torch_fx_available removal while 45020 is a broader catch-all; not the same concrete bug." + "reason": "Both mention remote_code breakage, but one is specifically is_torch_fx_available removal and the other is a broad recent-version regression; not the same concrete bug." }, { "left": "issue:43825", "right": "issue:43994", "accept": false, - "reason": "Pipeline error-message wording vs SigLIP2 nonsensical outputs; different problems." + "reason": "Pipeline error-message wording and Siglip2 nonsensical results are different problems." }, { "left": "issue:43550", "right": "issue:43819", "accept": false, - "reason": "Bamba torch.compile/SDPA failure and DAC from_latents/STE mismatch are unrelated." + "reason": "Bamba SDPA/torch.compile failure and DAC forward/from_latents mismatch are unrelated." }, { "left": "issue:44186", "right": "issue:44442", "accept": false, - "reason": "LayoutLMv2 tokenizer crash vs AutoTokenizer loading FastSpeech2ConformerTokenizer; different tokenizer paths." + "reason": "LayoutLMv2 tokenizer crash and AutoTokenizer loading FastSpeech2ConformerTokenizer are different tokenizer issues." }, { "left": "issue:43825", "right": "issue:43827", "accept": false, - "reason": "One is a runtime error-message bug, the other is stale docs referencing removed pipeline APIs." + "reason": "Both involve pipeline v5 messaging/docs, but one is a runtime error message and the other is stale documentation; not the same bug/change." } ] }, @@ -4368,8 +4368,8 @@ "issue:45538" ], "soft_pairs": [ - "issue:43650|issue:43906", "issue:44821|issue:45084", + "issue:43650|issue:43906", "issue:41720|issue:44991", "issue:43606|issue:45072", "issue:44805|issue:44936", @@ -4398,12 +4398,12 @@ "issue:43638|issue:44936", "issue:43854|issue:44977" ], - "prepared_review_unit_hash": "071141a6d5efbb2c88b5063cfffbdaae7320c861c981d26b471d407eed38b103", + "prepared_review_unit_hash": "cafd1b8c75caa3610108444bef45fd54e7d117783b54cbc849ff18bf5d74ba44", "component_budget": { "node_count": 367, "item_count": 367, "soft_pair_count": 1370, - "serialized_chars": 706987, + "serialized_chars": 706986, "estimated_input_tokens": 176747, "estimated_eval_tokens": 353750 }, @@ -4411,195 +4411,195 @@ "node_count": 48, "item_count": 48, "soft_pair_count": 29, - "serialized_chars": 36846, - "estimated_input_tokens": 9212, - "estimated_eval_tokens": 18680 + "serialized_chars": 36843, + "estimated_input_tokens": 9211, + "estimated_eval_tokens": 18678 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is a loose collection of mostly unrelated Transformers bug reports spanning tokenizers, processors, loading, dtype, and training paths. No soft duplicate pair is strong enough to merge, so the cluster should not be collapsed aggressively.", - "confidence": 0.93, - "canonical_issue_reason": "issue:45081 is the most actionable representative: it is an open, concrete crash report with multiple inbound references and a clearly described code-path failure.", + "summary": "This cluster is mostly a heterogeneous grab-bag of unrelated Transformers issues. None of the soft pairs look like the same underlying bug or change, so I would not merge any of them as duplicates. The best anchor issue is the bfloat16/device-mismatch report because it is the most central to the few loosely related loading/inference items, but it is still not a true duplicate of the others.", + "confidence": 0.43, + "canonical_issue_reason": "issue:45072 is the most cluster-central item because several other soft links loosely point at dtype/device loading/inference failures, but the cluster is too heterogeneous to justify a stronger duplicate canonical.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45081 is the best global issue candidate because it is specific, reproducible, and currently active, making it the most suitable anchor for the cluster.", + "best_issue_reason": "issue:45072 has the broadest relevance to the few related runtime-mismatch reports (device/dtype/inference), making it the best representative anchor even though the edges are not true duplicates.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:43650", - "right": "issue:43906", + "left": "issue:44821", + "right": "issue:45084", "accept": false, - "reason": "Completely different problems; one is a vague data request, the other is a reproduction of an unrelated issue." + "reason": "AutoImageProcessor URL loading and a tokenizer template compilation error are unrelated failure modes." }, { - "left": "issue:44821", - "right": "issue:45084", + "left": "issue:43650", + "right": "issue:43906", "accept": false, - "reason": "Different subsystems and failures: image processor URL loading vs template compilation error." + "reason": "One is a vague data request; the other is a reproduction of a specific upstream bug. Not the same issue." }, { "left": "issue:41720", "right": "issue:44991", "accept": false, - "reason": "Different model/load paths; one is CUDA auto device mapping, the other is tokenizer loading for a specific model." + "reason": "Different subsystems and different failures: CUDA assert during Qwen3 device mapping vs tokenizer loading for est-roberta." }, { "left": "issue:43606", "right": "issue:45072", "accept": false, - "reason": "Both are inference mismatches, but they affect different models and have different root causes." + "reason": "Both mention mismatches, but they are different models and different concrete code paths; not a mergeable duplicate." }, { "left": "issue:44805", "right": "issue:44936", "accept": false, - "reason": "A tensor shape/mask IndexError is not the same as trainer.evaluate() failing after training." + "reason": "A mask/indexing shape error is not the same as trainer.evaluate() failing after train()." }, { "left": "issue:30064", "right": "issue:36331", "accept": false, - "reason": "Different components entirely: image processor segmentation handling vs CustomTrainer signature mismatch." + "reason": "Image processor segmentation-map handling and a custom trainer keyword-argument regression are unrelated." }, { "left": "issue:44485", "right": "issue:44509", "accept": false, - "reason": "Unrelated topics: GLM RoPE implementation vs docs still mentioning removed pipeline tasks." + "reason": "RoPE implementation discussion and stale docs about removed pipelines are unrelated." }, { "left": "issue:42175", "right": "issue:45081", "accept": false, - "reason": "Backend packaging issue is unrelated to a tokenizer regex crash." + "reason": "Packaging/backends availability and a Mistral regex patch crash are different problems." }, { "left": "issue:44610", "right": "issue:45072", "accept": false, - "reason": "Processor input-size mismatch is a different bug from dtype mismatches." + "reason": "Processor input-size mismatch and dtype mismatch in inference are different concrete bugs." }, { "left": "issue:43782", "right": "issue:45325", "accept": false, - "reason": "Different Qwen/VL failures with different loading and position-index code paths." + "reason": "One is a from_pretrained weight_only load error; the other is a RoPE position-id scaling bug in a different Qwen VL model." }, { "left": "issue:34634", "right": "issue:36010", "accept": false, - "reason": "BarkProcessor voice_preset and GenerationMixin import error are unrelated." + "reason": "Bark voice_preset behavior and a GenerationMixin import error are unrelated." }, { "left": "issue:43720", "right": "issue:45072", "accept": false, - "reason": "Packed-weight loading for BitNet is a different issue than bfloat16 dtype mismatches." + "reason": "Packed-weight unpacking during accelerate loading is a different failure from dtype mismatch in bfloat16 inference." }, { "left": "issue:44060", "right": "issue:44368", "accept": false, - "reason": "Both mention tied-weights warnings, but the affected models and warning mechanisms differ." + "reason": "Both mention tied-weights warnings, but one is an incorrect tying bug and the other is a config warning in fine-tuning." }, { "left": "issue:45081", "right": "issue:45406", "accept": false, - "reason": "Mistral regex patch crash and Gemma4Processor missing _tokenizer are different loader bugs." + "reason": "Mistral regex patching and Gemma4Processor missing _tokenizer are unrelated tokenizer/serve failures." }, { "left": "issue:43749", "right": "issue:44291", "accept": false, - "reason": "FSDP CPU RAM-efficient loading failure is not the same as init_empty_weights rejecting an extra argument." + "reason": "Both concern low-memory loading, but one is broken FSDP CPU RAM efficient loading and the other is an unexpected constructor arg regression." }, { "left": "issue:43262", "right": "issue:43819", "accept": false, - "reason": "Audio chat-template sample-rate defaulting is unrelated to DAC latent reconstruction." + "reason": "Audio sampling-rate defaults and DAC latent reconstruction mismatch are unrelated." }, { "left": "issue:43582", "right": "issue:44857", "accept": false, - "reason": "Apple Silicon warmup TypeError and CUDA AMP loss crash are different platform-specific bugs." + "reason": "Apple Silicon warmup TypeError and CUDA AMP loss crash are different platform/code-path issues." }, { "left": "issue:43653", "right": "issue:44671", "accept": false, - "reason": "Tokenizer special-token registration and CamemBERT masked-LM output regressions are not the same defect." + "reason": "Different tokenizer/model bugs: BigBird special-token registration vs CamemBERT masked LM predictions." }, { "left": "issue:43526", "right": "issue:43697", "accept": false, - "reason": "BEiT label reduction bug and RTDetrV2 output differences are separate image-model issues." + "reason": "Label reduction in BeitImageProcessorFast is not the same as RTDetrV2 output drift in v5." }, { "left": "issue:43526", "right": "issue:44554", "accept": false, - "reason": "Image label reduction has no meaningful overlap with the MPS attention correctness report." + "reason": "Image processor label handling and an MPS attention correctness issue are unrelated." }, { "left": "issue:43653", "right": "issue:44912", "accept": false, - "reason": "BigBird token decode failure and BitNet packed-weight loading are unrelated." + "reason": "Tokenizer special-token registration and MXFP4 quantization fallback are different bugs." }, { "left": "issue:45084", "right": "issue:45538", "accept": false, - "reason": "Template compilation error is unrelated to CLIPTokenizer model_max_length behavior." + "reason": "Template-compilation failure and CLIPTokenizer max-length misconfiguration are not the same problem." }, { "left": "issue:43454", "right": "issue:43577", "accept": false, - "reason": "Different multimodal model-loading bugs with different symptoms and root causes." + "reason": "AyaVision weight tying and Blip2 dtype casting are different multimodal loading bugs." }, { "left": "issue:34634", "right": "issue:36331", "accept": false, - "reason": "Processor preset behavior and trainer loss-signature mismatch are unrelated." + "reason": "BarkProcessor voice preset and CustomTrainer compute_loss signature are unrelated." }, { "left": "issue:44568", "right": "issue:45488", "accept": false, - "reason": "Both are tokenizer regressions, but they affect different tokenizer classes and different behaviors." + "reason": "Different tokenizer regressions affecting different models and behaviors; not the same code path." }, { "left": "issue:43257", "right": "issue:43329", "accept": false, - "reason": "MOE weight conversion during loading and multimodal token counting are unrelated." + "reason": "Qwen3 MOE loading with accelerate/deepspeed and multimodal token counting/video-branch undefined names are unrelated." }, { "left": "issue:43452", "right": "issue:44991", "accept": false, - "reason": "GGUF loading support and a specific tokenizer loading regression are different problems." + "reason": "gguf_file loading support and est-roberta tokenizer loading failure are different issues." }, { "left": "issue:43638", "right": "issue:44936", "accept": false, - "reason": "Zero3 index error during training is not the same as evaluate() failing after train()." + "reason": "Zero3/Bert training index error and post-train evaluation failure are unrelated." }, { "left": "issue:43854", "right": "issue:44977", "accept": false, - "reason": "GLM-4.7-Flash test-loading issue and Qwen3.5 flash-attention generation bug are different model-specific regressions." + "reason": "GLM-4.7 unit-test loading and Qwen3.5 flash-attention generation issues are different model-specific bugs." } ] }, @@ -4674,16 +4674,16 @@ ], "soft_pairs": [ "issue:43746|issue:44811", - "issue:38175|issue:44610", "issue:44589|issue:45042", + "issue:38175|issue:44610", "issue:39692|issue:44912", "issue:43257|issue:43696", "issue:43819|issue:43927", "issue:43720|issue:44610", "issue:43550|issue:44912", "issue:36010|issue:43645", - "issue:43262|issue:44811", "issue:44568|issue:45356", + "issue:43262|issue:44811", "issue:43749|issue:44756", "issue:43329|issue:44479", "issue:43531|issue:45042", @@ -4700,17 +4700,17 @@ "issue:44479|issue:45381", "issue:43927|issue:44912", "issue:44987|issue:45538", - "issue:43317|issue:45538", "issue:42915|issue:43742", "issue:44291|issue:45042", + "issue:43317|issue:45538", "issue:33357|issue:34567" ], - "prepared_review_unit_hash": "3ba5a4ae1ef1c371ec82c037cd705cd01cedcf735aece03c74682daedc372a89", + "prepared_review_unit_hash": "1177635d0dbb40e6d3b10aca178c2aade811f200e3da111666a61ee846954453", "component_budget": { "node_count": 367, "item_count": 367, "soft_pair_count": 1370, - "serialized_chars": 706987, + "serialized_chars": 706986, "estimated_input_tokens": 176747, "estimated_eval_tokens": 353750 }, @@ -4727,198 +4727,198 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is not a true duplicate set: the items span many unrelated bugs across audio, vision, tokenization, loading, quantization, and training. No single canonical issue or PR stands out, and all soft-edge pairs should be rejected.", - "confidence": 0.96, - "canonical_issue_reason": "No canonical issue: the cluster is heterogeneous and the items do not share one underlying bug or code path.", + "summary": "These are mostly unrelated issues grouped by broad Transformers subsystems. I do not see a single true duplicate cluster, and none of the soft pairs are strong enough to merge as the same underlying bug/change.", + "confidence": 0.88, + "canonical_issue_reason": "No canonical issue: the set spans many distinct failures (tokenizers, multimodal/video processors, loading/quantization, hardware-specific crashes, and trainer behavior) rather than one shared root cause.", "canonical_pr_reason": null, - "best_issue_reason": "No single issue is representative enough to serve as a canonical duplicate target; the set spans unrelated subsystems and failure modes.", + "best_issue_reason": "No global best issue: there is no one issue that represents the whole set, and the pairwise candidates are not close enough to be treated as duplicates.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43746", "right": "issue:44811", "accept": false, - "reason": "Both are audio-related, but one is GraniteSpeech PEFT checkpoint loading and the other is Whisper batch_decode special-token handling; different bugs." + "reason": "Different surfaces and root causes: local checkpoint loading for GraniteSpeech adapters vs Whisper processor decoding behavior." }, { - "left": "issue:38175", - "right": "issue:44610", + "left": "issue:44589", + "right": "issue:45042", "accept": false, - "reason": "Different models and failures: siglip2 zero probabilities vs OmDet-Turbo processor shape/size mismatch." + "reason": "Unrelated bugs: Float8 storage lookup failure vs PIL image processors incorrectly requiring torchvision." }, { - "left": "issue:44589", - "right": "issue:45042", + "left": "issue:38175", + "right": "issue:44610", "accept": false, - "reason": "Unrelated issues: Float8 storage lookup error vs image processors unnecessarily requiring torchvision." + "reason": "Both are model/processor complaints, but they concern different models and different failures (zero probabilities vs input-size mismatch)." }, { "left": "issue:39692", "right": "issue:44912", "accept": false, - "reason": "Docs/example errors for SigLIP2 are not the same bug as GPT-OSS MXFP4 loading falling back to bf16." + "reason": "Different models and different bugs: SigLIP2 doc example errors vs GPT-OSS MXFP4 quantization fallback/loading." }, { "left": "issue:43257", "right": "issue:43696", "accept": false, - "reason": "Different problems: Qwen3 MoE weight conversion/loading under accelerate+deepspeed vs GPT-oss OOM." + "reason": "Accelerate/DeepSpeed MoE weight conversion is not the same problem as a CUDA OOM on GPT-oss-20b." }, { "left": "issue:43819", "right": "issue:43927", "accept": false, - "reason": "Different code paths: DAC latent reconstruction mismatch vs DiaConfig losing custom token IDs on save/load." + "reason": "Different code paths: DAC latent reconstruction/STE mismatch vs DiaConfig losing custom token IDs on save/load." }, { "left": "issue:43720", "right": "issue:44610", "accept": false, - "reason": "BitNet packed-weight loading under accelerate is unrelated to OmDet-Turbo processor output shape." + "reason": "Packed-weight loading in BitNet is unrelated to OmDet-Turbo processor image-size output mismatch." }, { "left": "issue:43550", "right": "issue:44912", "accept": false, - "reason": "torch.compile SDPA failure for Bamba is unrelated to GPT-OSS quantization fallback." + "reason": "torch.compile + SDPA failure in Bamba-9B-v2 is a different bug from MXFP4 quantization loading/fallback." }, { "left": "issue:36010", "right": "issue:43645", "accept": false, - "reason": "Both involve Transformers changes, but one is GenerationMixin import failure and the other is custom model definition in notebooks." + "reason": "Import path breakage for GenerationMixin is not the same as Jupyter custom-model initialization regressions." }, { - "left": "issue:43262", - "right": "issue:44811", + "left": "issue:44568", + "right": "issue:45356", "accept": false, - "reason": "Audio processor chat-template sampling-rate defaulting is a different bug from Whisper batch_decode skipping special tokens." + "reason": "Both are tokenizer regressions, but they affect different models and different behaviors, so they are not the same bug." }, { - "left": "issue:44568", - "right": "issue:45356", + "left": "issue:43262", + "right": "issue:44811", "accept": false, - "reason": "Tokenizer special-token insertion regression is unrelated to Kimi-K2.5 codec/regex warning behavior." + "reason": "Audio chat-template sampling-rate defaulting is unrelated to Whisper batch_decode skip_special_tokens handling." }, { "left": "issue:43749", "right": "issue:44756", "accept": false, - "reason": "Both mention loading/memory, but FSDP RAM-efficient loading is not the same as disabling mmap on Strix Halo to avoid OOM." + "reason": "FSDP CPU-RAM-efficient loading and Strix Halo mmap OOM are different loading/memory issues with no shared concrete root cause." }, { "left": "issue:43329", "right": "issue:44479", "accept": false, - "reason": "Both are video-related, but one is undefined variables in _get_num_multimodal_tokens and the other is a broader video-input regression across Qwen VLMs." + "reason": "Both involve video VLMs, but one is undefined names in a helper and the other is a broader regression in video input handling." }, { "left": "issue:43531", "right": "issue:45042", "accept": false, - "reason": "Qwen3-MoE sliding_window behavior is unrelated to image processors requiring torchvision." + "reason": "Qwen3-MoE sliding_window behavior is unrelated to PIL backend image processors requiring torchvision." }, { "left": "issue:43257", "right": "issue:43931", "accept": false, - "reason": "Different loading failures: Qwen3 MoE weight conversion under accelerate/deepspeed vs Qwen3-VL weight shape mismatch." + "reason": "MoE weight conversion during accelerate/deepspeed loading is not the same as Qwen3-VL weight-shape mismatch on model load." }, { "left": "issue:36010", "right": "issue:36331", "accept": false, - "reason": "Importing GenerationMixin and CustomTrainer.compute_loss signature changes are distinct API breakages." + "reason": "GenerationMixin import failure and CustomTrainer.compute_loss signature mismatch are separate API breakages." }, { "left": "issue:43257", "right": "issue:43756", "accept": false, - "reason": "Qwen3 MoE weight-loading issues are unrelated to Smollm3 RoPE-layer mismatch." + "reason": "Qwen3 MoE weight conversion and Smollm3 RoPE-layer count regression are unrelated model-specific issues." }, { "left": "issue:43329", "right": "issue:43450", "accept": false, - "reason": "Video token counting undefined variables and batched video processor shape errors are different bugs." + "reason": "Undefined video-token helper names are not the same as batched video processor shape errors." }, { "left": "issue:34567", "right": "issue:40990", "accept": false, - "reason": "TrainerState token-count tracking is unrelated to GPT-oss perplexity being too high." + "reason": "TrainerState token-count tracking and GPT-oss perplexity are unrelated training/evaluation issues." }, { "left": "issue:43577", "right": "issue:45072", "accept": false, - "reason": "Different models and symptoms: Blip2 dtype remaining float32 vs dtype mismatches in SwitchTransformers/TimmWrapperModel." + "reason": "BLIP2 dtype propagation and SwitchTransformers/TimmWrapper bfloat16 mismatches are different model-loading dtype bugs." }, { "left": "issue:43697", "right": "issue:45357", "accept": false, - "reason": "RTDetrV2 output drift in v5 is unrelated to Qwen3.5 save_pretrained visual encoder keys." + "reason": "RTDetrV2 output drift in v5 is unrelated to Qwen3.5 save_pretrained visual-encoder key regression." }, { "left": "issue:42175", "right": "issue:44938", "accept": false, - "reason": "Missing TensorFlow backend in pip extras is unrelated to Python 3.14 import/load failure." + "reason": "Missing TensorFlow in torch extra install and Python 3.14 import failure are different packaging/runtime problems." }, { "left": "issue:43452", "right": "issue:43582", "accept": false, - "reason": "gguf_file loading issues are unrelated to the AppleSilicon caching_allocator_warmup TypeError." + "reason": "gguf_file tokenizer/model loading breakage is unrelated to Apple Silicon caching_allocator_warmup TypeError." }, { "left": "issue:41720", "right": "issue:43540", "accept": false, - "reason": "Qwen3 auto device-mapping cudaErrorAssert and Qwen3OmniMoe video ValueError are different failures." + "reason": "Qwen3 auto device-map CUDA assert and Qwen3OmniMoe video ValueError are different failures and likely different code paths." }, { "left": "issue:44479", "right": "issue:45381", "accept": false, - "reason": "Both involve Qwen video input regressions, but one is a broad multi-model regression and the other is a specific vision_position_ids bug; not enough to treat as the same concrete issue." + "reason": "Both are video-input regressions, but they mention different models and different symptoms, so they are not clearly the same bug." }, { "left": "issue:43927", "right": "issue:44912", "accept": false, - "reason": "DiaConfig token-ID persistence and GPT-oss quantization fallback are unrelated." + "reason": "Config token-ID persistence and MXFP4 loading fallback are unrelated issues." }, { "left": "issue:44987", "right": "issue:45538", "accept": false, - "reason": "Loading physical-intelligence/fast under v5.1.0 is unrelated to CLIPTokenizer model_max_length behavior." - }, - { - "left": "issue:43317", - "right": "issue:45538", - "accept": false, - "reason": "Dequantized model offload loading and CLIPTokenizer max_length are different subsystems and bugs." + "reason": "Loading physical-intelligence/fast on v5.1.0+ is not the same as CLIPTokenizer model_max_length behavior." }, { "left": "issue:42915", "right": "issue:43742", "accept": false, - "reason": "Qwen3Moe FineGrainedFP8Config failure is unrelated to MobileLLM-125M key error." + "reason": "FineGrainedFP8Config failure for Qwen3Moe and a key error loading MobileLLM-125M are different model-load bugs." }, { "left": "issue:44291", "right": "issue:45042", "accept": false, - "reason": "init_empty_weights/_is_hf_initialized argument handling is unrelated to torchvision being required by PIL image processors." + "reason": "init_empty_weights/_is_hf_initialized TypeError is unrelated to the torchvision requirement regression." + }, + { + "left": "issue:43317", + "right": "issue:45538", + "accept": false, + "reason": "GPU+CPU offload loading failure is unrelated to CLIPTokenizer max-length semantics." }, { "left": "issue:33357", "right": "issue:34567", "accept": false, - "reason": "MacOS bus error with CLIP model loading is unrelated to TrainerState token counting." + "reason": "MacOS bus error on CLIP model load and TrainerState token counting are completely different issues." } ] }, @@ -5007,243 +5007,236 @@ "issue:43646|issue:43749", "issue:43582|issue:43782", "issue:43643|issue:45375", + "issue:43742|issue:44291", "issue:43756|issue:45325", "issue:43493|issue:43994", "issue:28282|issue:29127", - "issue:43742|issue:44291", "issue:36010|issue:42175", "issue:44484|issue:44704", "issue:34567|issue:36331", "issue:44291|issue:44821", - "issue:44062|issue:45081", "issue:43452|issue:44488", + "issue:44062|issue:45081", "issue:44792|issue:44964", "issue:43688|issue:43749", "issue:44514|issue:45290", - "issue:43643|issue:44561", "issue:43452|issue:43577", - "issue:44514|issue:45381" + "issue:43643|issue:44561" ], - "prepared_review_unit_hash": "d24c3b13f82400cc74d846f5c582681715b84f56f28da2d4219048215223f264", + "prepared_review_unit_hash": "42e8fc472bfa84066a4901b9b63b663c1290e95776102c46d2c94ee706323b48", "component_budget": { "node_count": 367, "item_count": 367, "soft_pair_count": 1370, - "serialized_chars": 706987, + "serialized_chars": 706986, "estimated_input_tokens": 176747, "estimated_eval_tokens": 353750 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 31, - "serialized_chars": 37180, - "estimated_input_tokens": 9295, - "estimated_eval_tokens": 18846 + "soft_pair_count": 30, + "serialized_chars": 36890, + "estimated_input_tokens": 9223, + "estimated_eval_tokens": 18702 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Mostly a heterogeneous issue set. I\u2019d merge only two clear duplicate-style pairs: the Qwen chat-template crash on assistant tool-call messages, and the SigLIP2 regression/symptom report pair. The remaining soft pairs are related by subsystem or model family but look like distinct bugs.", - "confidence": 0.78, - "canonical_issue_reason": "issue:44514 is the clearest, most actionable duplicate anchor: it states the exact crash condition in `apply_chat_template(tokenize=True)` with assistant tool calls and no content, and it is directly duplicated by a later report.", + "summary": "Mostly unrelated issue reports; the only clear duplicate pair is the `apply_chat_template` crash with assistant tool calls and empty content. The rest are at best broad topical similarities, not the same concrete bug.", + "confidence": 0.58, + "canonical_issue_reason": "`issue:45290` is the cleanest canonical report: it states the core `apply_chat_template(tokenize=True)` crash with assistant tool calls and no content, while `issue:44514` looks like a narrower variant of the same failure.", "canonical_pr_reason": null, - "best_issue_reason": "issue:44514 is the best global issue anchor because it has a precise repro, a narrowly defined failure mode, and a near-exact duplicate in the set.", + "best_issue_reason": "`issue:45290` is the best representative of the only strong duplicate subgroup in this set because it describes the underlying code-path bug most directly and broadly.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43540", "right": "issue:44560", "accept": false, - "reason": "Both are video-input bugs, but they hit different Qwen model families and report different failures (`ValueError` vs `StopIteration`)." + "reason": "Same broad Qwen video area, but different models and error shapes; not enough to treat as the same concrete bug." }, { "left": "issue:43906", "right": "issue:45381", "accept": false, - "reason": "`43906` is a reproduction of a different issue; `45381` is a Qwen2.5-VL video-position regression. No clear shared bug." + "reason": "Different problems: isolated reproduction of an older issue versus a Qwen2.5-VL video position-id regression." }, { "left": "issue:43540", "right": "issue:44479", "accept": false, - "reason": "Related to video handling, but one is Qwen3OmniMoe-specific and the other is a broader Qwen2.5/Qwen3 video regression. Too broad to merge." + "reason": "Both mention video inputs, but one is Qwen3OmniMoe-specific and the other is a broader Qwen video regression; not clearly the same fixable bug." }, { "left": "issue:45137", "right": "issue:45405", "accept": false, - "reason": "DeepSpeed ZeRO3 deque error and a PEFT version bump are unrelated." + "reason": "Completely different topics: DeepSpeed ZeRO3 deque error vs PEFT version pin bump." }, { "left": "issue:42915", "right": "issue:45538", "accept": false, - "reason": "Different components and failures: FP8 config loading vs CLIP tokenizer max-length behavior." + "reason": "Unrelated: Qwen3Moe FP8 loading failure vs CLIPTokenizer max length behavior." }, { "left": "issue:42175", "right": "issue:43824", "accept": false, - "reason": "Package backend installation issue vs missing exported model class; no shared code-path bug." + "reason": "Both are import/backend-related, but one is TensorFlow backend packaging and the other is a missing Qwen2_5_VL export." }, { "left": "issue:42915", "right": "issue:44821", "accept": false, - "reason": "Qwen3Moe FP8/config issue and AutoImageProcessor URL loading are unrelated." + "reason": "Different failures in different components; no shared code-path evidence." }, { "left": "issue:43526", "right": "issue:45479", "accept": false, - "reason": "Different sequence/image-processing bugs with different symptoms and APIs." + "reason": "Different model families and bugs: BeitImageProcessorFast label reduction vs sequence-classification zero-loss behavior." }, { "left": "issue:42175", "right": "issue:44488", "accept": false, - "reason": "Dependency/backend install bug vs a specific model-loading failure; not the same defect." + "reason": "Both are loading-related, but the specific failures and targets are unrelated." }, { "left": "issue:44062", "right": "issue:44991", "accept": false, - "reason": "AddedToken keyword collision is a tokenizer constructor bug; `est-roberta` loading is a different tokenizer/model issue." + "reason": "Both involve tokenizer loading, but one is a generic AddedToken TypeError and the other is a model-specific tokenizer load failure." }, { "left": "issue:38175", "right": "issue:44898", "accept": false, - "reason": "Both concern vision models, but they involve different architectures and distinct failures." + "reason": "Different model behaviors: SigLIP2 zero probabilities vs Perceiver non-default resolution failure." }, { "left": "issue:28282", "right": "issue:34689", "accept": false, - "reason": "Missing PyTorch dependency error and Llama 3.2 Vision loading regression are not the same bug." + "reason": "Both are import/load complaints, but about different models and different underlying causes." }, { "left": "issue:43646", "right": "issue:43749", "accept": false, - "reason": "Custom model initialization breakage and FSDP CPU RAM-efficient loading are separate code paths." + "reason": "Different regression classes: custom model init breakage vs FSDP CPU RAM-efficient loading." }, { "left": "issue:43582", "right": "issue:43782", "accept": false, - "reason": "Apple Silicon allocator warmup `TypeError` and Qwen3VL `weight_only=True` loading error are unrelated." + "reason": "Apple Silicon allocator TypeError and Qwen3VL weight_only loading error are unrelated." }, { "left": "issue:43643", "right": "issue:45375", "accept": false, - "reason": "Both involve config/loading behavior, but one is generic `trust_remote_code` field loss and the other is a specific model config field being dropped." + "reason": "Both touch config handling, but one is generic remote-code field loss and the other is a specific missing config field." + }, + { + "left": "issue:43742", + "right": "issue:44291", + "accept": false, + "reason": "Different loading bugs: MobileLLM key error vs init_empty_weights/_is_hf_initialized TypeError." }, { "left": "issue:43756", "right": "issue:45325", "accept": false, - "reason": "Smollm3 RoPE-layer mismatch and Qwen2.5-VL temporal position scaling are different model-specific bugs." + "reason": "Both mention RoPE/position handling, but they concern different models and different bugs." }, { "left": "issue:43493", "right": "issue:43994", - "accept": true, - "reason": "Same SigLIP2 model family and same underlying regression: HF output disagrees with the original JAX behavior and yields nonsensical predictions." + "accept": false, + "reason": "Same model family, but one is an implementation discrepancy and the other is a bad AutoModel/pipeline result; too broad to merge." }, { "left": "issue:28282", "right": "issue:29127", "accept": false, - "reason": "Different import/error-message problems with no shared failure mode." - }, - { - "left": "issue:43742", - "right": "issue:44291", - "accept": false, - "reason": "MobileLLM key error and `init_empty_weights` unexpected-argument failure are distinct loading issues." + "reason": "Both are error-message issues, but one is missing PyTorch import dependency and the other is LayoutLMv3 box-content validation." }, { "left": "issue:36010", "right": "issue:42175", "accept": false, - "reason": "Different import/backend problems with no shared root cause." + "reason": "Both are import/version issues, but they concern different APIs and failure modes." }, { "left": "issue:44484", "right": "issue:44704", "accept": false, - "reason": "`max_shard_size` question and missing kwargs forwarding to `cached_file` are unrelated." + "reason": "Different API behaviors: save_pretrained shard sizing vs AutoProcessor kwargs passthrough." }, { "left": "issue:34567", "right": "issue:36331", "accept": false, - "reason": "Trainer state token counting and custom `compute_loss` signature breakage are different trainer bugs." + "reason": "TrainerState token counting and custom trainer compute_loss signature are unrelated trainer issues." }, { "left": "issue:44291", "right": "issue:44821", "accept": false, - "reason": "`init_empty_weights` constructor issue and remote `AutoImageProcessor` loading are unrelated." + "reason": "Different subsystems and failures: empty-weights init vs loading AutoImageProcessor from URL." }, { - "left": "issue:44062", - "right": "issue:45081", + "left": "issue:43452", + "right": "issue:44488", "accept": false, - "reason": "Tokenizer `AddedToken` argument collision vs Mistral regex patch/backend-tokenizer attribute error are different code paths." + "reason": "Both are model-loading complaints, but one is a gguf_file/from_pretrained regression and the other is a specific model repo load failure." }, { - "left": "issue:43452", - "right": "issue:44488", + "left": "issue:44062", + "right": "issue:45081", "accept": false, - "reason": "GGUF loading failure and a specific model failing to load are not the same defect." + "reason": "Tokenizer errors in both, but the root causes and affected paths differ." }, { "left": "issue:44792", "right": "issue:44964", "accept": false, - "reason": "Janus test failure and Phi-4 multimodal loading regression are different multimodal issues." + "reason": "Different multimodal model failures; one is a failing janus test, the other is a Phi-4 multimodal load issue." }, { "left": "issue:43688", "right": "issue:43749", "accept": false, - "reason": "Auxiliary-loss normalization and FSDP loading are unrelated." + "reason": "Aux-loss normalization bug and FSDP loading bug are unrelated." }, { "left": "issue:44514", "right": "issue:45290", "accept": true, - "reason": "Near-exact duplicate: both report `apply_chat_template` crashing for assistant messages with tool calls and no content." - }, - { - "left": "issue:43643", - "right": "issue:44561", - "accept": false, - "reason": "Generic `trust_remote_code` field handling and removal of `is_torch_fx_available` affecting custom models are different regressions." + "reason": "Same underlying `apply_chat_template` crash with assistant tool calls and empty content; one is just a batched/padding variant." }, { "left": "issue:43452", "right": "issue:43577", "accept": false, - "reason": "GGUF tokenizer/model loading and BLIP2 dtype propagation are unrelated." + "reason": "Both involve model loading, but the specific models and failure causes differ." }, { - "left": "issue:44514", - "right": "issue:45381", + "left": "issue:43643", + "right": "issue:44561", "accept": false, - "reason": "Chat-template crash and Qwen2.5-VL video position-id regression are unrelated bugs." + "reason": "Both are about `trust_remote_code`-adjacent loading, but one is missing fields in AutoConfig and the other is a removed helper breaking remote-code models." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is grounded in the packet and stays conservative. The two accepted duplicate-style pairs are plausible from the titles and issue descriptions, and the rejected soft pairs are mostly distinct enough to justify non-mergeability. No obvious overstatement of evidence." + "feedback": "The summary is conservative and supported by the packet. The one accepted soft pair is the only clear duplicate subgroup indicated, and the other pairwise rejections stay within the evidence available from titles/metadata without overstating sameness." }, "evaluator_used": true, "retried": false, @@ -5267,10 +5260,8 @@ "issue:24643", "issue:28282", "issue:29127", - "issue:30064", "issue:32090", "issue:33357", - "issue:33666", "issue:35141", "issue:38175", "issue:41720", @@ -5294,6 +5285,7 @@ "issue:44464", "issue:44466", "issue:44479", + "issue:44514", "issue:44568", "issue:44589", "issue:44625", @@ -5311,19 +5303,21 @@ "issue:45081", "issue:45216", "issue:45356", + "issue:45381", "issue:45440" ], "soft_pairs": [ "issue:43742|issue:44589", + "issue:44514|issue:45381", "issue:43697|issue:45216", "issue:45042|issue:45081", "issue:43531|issue:44821", "issue:44625|issue:44877", "issue:43688|issue:45440", "issue:43720|issue:44898", - "issue:43653|issue:44568", "issue:43452|issue:44843", "issue:43819|issue:44811", + "issue:43653|issue:44568", "issue:44466|issue:45005", "issue:43295|issue:45356", "issue:44464|issue:44743", @@ -5338,18 +5332,17 @@ "issue:44843|issue:44991", "issue:28282|issue:32090", "issue:28282|issue:35141", - "issue:28282|issue:33357", "issue:44479|issue:44936", + "issue:28282|issue:33357", "issue:43122|issue:43906", - "issue:44977|issue:45356", - "issue:30064|issue:33666" + "issue:44977|issue:45356" ], - "prepared_review_unit_hash": "03ea53bfe8a4ea2a532aed2712a06be0cf706a0a59f271f5504e16f021ac850e", + "prepared_review_unit_hash": "e1d30d83ca0947ed1fae1660f69c0d19a24ac41d938e48d521623eaa92f3df00", "component_budget": { "node_count": 367, "item_count": 367, "soft_pair_count": 1370, - "serialized_chars": 706987, + "serialized_chars": 706986, "estimated_input_tokens": 176747, "estimated_eval_tokens": 353750 }, @@ -5357,202 +5350,205 @@ "node_count": 48, "item_count": 48, "soft_pair_count": 29, - "serialized_chars": 36682, - "estimated_input_tokens": 9171, - "estimated_eval_tokens": 18598 + "serialized_chars": 36769, + "estimated_input_tokens": 9193, + "estimated_eval_tokens": 18642 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is highly heterogeneous: the issues span unrelated tokenizer, training, vision, config, and serialization bugs. The soft-similarity pairs are mostly superficial matches on broad vocabulary rather than the same underlying defect, so none should be merged.", - "confidence": 0.92, - "canonical_issue_reason": "No clear canonical issue; the items span unrelated bug classes and subsystems rather than one shared problem.", + "summary": "This batch is mostly a grab-bag of unrelated issue reports. I only see one plausible duplicate pair around tied-weights serialization in v5; the rest are different model families, different code paths, or different failure modes.", + "confidence": 0.83, + "canonical_issue_reason": "If a canonical issue must be chosen, issue 45005 is the best representative because it describes the broader tied-weights regression that also covers the more device-specific symptom in 44466.", "canonical_pr_reason": null, - "best_issue_reason": "No global best issue stands out because the cluster does not share a single underlying bug or change.", + "best_issue_reason": "Issue 45005 is the strongest global representative in this set: it is the clearest, broadest description of the only likely duplicate cluster (tied weights in v5).", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43742", "right": "issue:44589", "accept": false, - "reason": "Different loading failures: a KeyError for MobileLLM-125M vs a missing Float8 storage object." + "reason": "Different load failures and likely different root causes: MobileLLM key error vs Float8 storage lookup error." + }, + { + "left": "issue:44514", + "right": "issue:45381", + "accept": false, + "reason": "Both involve Qwen2.5-VL, but one is chat-template batching/padding and the other is video vision position IDs." }, { "left": "issue:43697", "right": "issue:45216", "accept": false, - "reason": "Different models and symptoms: RTDetrV2 output drift vs Qwen3.5 checkpoint save_pretrained regression." + "reason": "Unrelated regressions: RTDetrV2 output mismatch vs Qwen3.5 checkpoint save_pretrained corruption." }, { "left": "issue:45042", "right": "issue:45081", "accept": false, - "reason": "Unrelated areas: image processor torchvision dependency vs Mistral regex patch crashing in tokenizer loading." + "reason": "Different subsystems entirely: image processor torchvision dependency vs Mistral regex patch crash." }, { "left": "issue:43531", "right": "issue:44821", "accept": false, - "reason": "Different subsystems: Qwen3-MoE sliding window behavior vs AutoImageProcessor loading from URL." + "reason": "Qwen3-MoE sliding-window bug is unrelated to AutoImageProcessor URL loading." }, { "left": "issue:44625", "right": "issue:44877", "accept": false, - "reason": "Config propagation bug for Qwen3.5 text config vs strict config blocking granite_speech loading." + "reason": "Both are config-related, but they affect different models and different failures." }, { "left": "issue:43688", "right": "issue:45440", "accept": false, - "reason": "Both involve MoE-style models, but one is auxiliary-loss normalization and the other is DeepSeekV3 implementation divergence." + "reason": "Both mention MoE models, but one is auxiliary-loss normalization and the other is DeepSeekV3 implementation divergence." }, { "left": "issue:43720", "right": "issue:44898", "accept": false, - "reason": "Different code paths: packed-weight unpacking during accelerate loading vs Perceiver image classification at non-default resolution." - }, - { - "left": "issue:43653", - "right": "issue:44568", - "accept": false, - "reason": "Both are tokenizer special-token issues, but they affect different tokenizers and different missing tokens." + "reason": "Different bug classes: unpacking packed weights during accelerate load vs Perceiver image classification at non-default resolution." }, { "left": "issue:43452", "right": "issue:44843", "accept": false, - "reason": "Both involve tokenizer/model loading, but one is gguf_file breakage and the other is an offline-mode model_info() call in _patch_mistral_regex." + "reason": "Both are from tokenizer/model loading, but one is gguf_file handling and the other is an offline model_info() call in mistral regex patching." }, { "left": "issue:43819", "right": "issue:44811", "accept": false, - "reason": "Completely different bugs: DAC from_latents mismatch vs Whisper batch_decode ignoring skip_special_tokens." + "reason": "DAC latent reconstruction mismatch is unrelated to Whisper batch_decode skip_special_tokens handling." + }, + { + "left": "issue:43653", + "right": "issue:44568", + "accept": false, + "reason": "Both concern special tokens, but they are different tokenizers and different failures." }, { "left": "issue:44466", "right": "issue:45005", - "accept": false, - "reason": "Both mention tied weights, but one is device-dependent lm_head serialization and the other is a broader translation-model issue; not clearly the same bug." + "accept": true, + "reason": "Same underlying v5 tied-weights regression: 44466 is the device-dependent serialization symptom, while 45005 is the broader tied-weights report." }, { "left": "issue:43295", "right": "issue:45356", "accept": false, - "reason": "Different regressions: processor.tokenizer/image handling breakage vs Kimi-K2.5 codec handling and warning behavior." + "reason": "Different regressions affecting different tokenizer/processor paths and different versions." }, { "left": "issue:44464", "right": "issue:44743", "accept": false, - "reason": "Both touch generation/cache behavior, but the affected models and failure modes differ." + "reason": "Both mention generation/state behavior, but one is compiled-forward chunked generation and the other is recurrent-state reset in Qwen3.5 modular code." }, { "left": "issue:41762", "right": "issue:44936", "accept": false, - "reason": "ZeRO-3 model loading index error vs trainer.evaluate failing after trainer.train are different problems." + "reason": "DeepSpeed ZeRO-3 load error is unrelated to trainer.evaluate() failing after trainer.train()." }, { "left": "issue:24643", "right": "issue:29127", "accept": false, - "reason": "DeepSpeed training weight-shape error vs a LayoutLMv3 error-message clarity issue are unrelated." + "reason": "Different errors and code paths: DeepSpeed training weight shape vs LayoutLMV3 box-validation message clarity." }, { "left": "issue:41720", "right": "issue:43906", "accept": false, - "reason": "No evidence they are the same bug; one is a Qwen3 device-mapping cuda assert, the other is only an isolated reproduction of another issue." + "reason": "No evidence of the same bug; the second is only an isolated reproduction of another issue." }, { "left": "issue:43525", "right": "issue:43643", "accept": false, - "reason": "A missing pad_token_id on Llama4Config is not the same as a general trust_remote_code config field omission." + "reason": "Related thematically, but one is a specific missing pad_token_id on Llama4Config and the other is a broader AutoConfig trust_remote_code field-loss bug." }, { "left": "issue:44361", "right": "issue:44442", "accept": false, - "reason": "Different tokenizer failures on different models, with no shared code path evident." + "reason": "Different tokenizer failures with different models and symptoms." }, { "left": "issue:43577", "right": "issue:43697", "accept": false, - "reason": "Unrelated issues: Blip2 dtype propagation vs RTDetrV2 output differences." + "reason": "Blip2 dtype propagation is unrelated to RTDetrV2 output differences." }, { "left": "issue:38175", "right": "issue:43295", "accept": false, - "reason": "Different model classes and bugs: siglip2 zero probabilities vs tokenizer/processor regression in v4.57.5." + "reason": "Siglip2 zero probabilities are unrelated to the processor.tokenizer regression." }, { "left": "issue:43653", "right": "issue:44361", "accept": false, - "reason": "Both are tokenizer bugs, but they affect different tokenizers and different failure modes." + "reason": "Different tokenizer bugs affecting different tokenizers and failure modes." }, { "left": "issue:44843", "right": "issue:44991", "accept": false, - "reason": "Both are tokenizer loading regressions, but one is offline-mode network access and the other is a specific tokenizer load failure." + "reason": "Both are tokenizer-loading regressions, but they appear to be separate root causes." }, { "left": "issue:28282", "right": "issue:32090", "accept": false, - "reason": "Completely unrelated: missing PyTorch import vs NoneType broadcast error in Trainer." + "reason": "ImportError for missing PyTorch and NoneType broadcast error are unrelated." }, { "left": "issue:28282", "right": "issue:35141", "accept": false, - "reason": "No overlap: import dependency error vs tied-embedding reinitialization during post_init." + "reason": "No shared underlying bug: missing PyTorch vs tied embedding resize reinitialization." }, { "left": "issue:28282", "right": "issue:33357", "accept": false, - "reason": "Different problems entirely: PyTorch import error vs MacOS bus error with a CLIP model." + "reason": "Completely different failures: missing backend dependency vs MacOS bus error on CLIP." }, { "left": "issue:44479", "right": "issue:44936", "accept": false, - "reason": "Video-input regression in Qwen VL models vs trainer lifecycle failure after training are unrelated." + "reason": "Video-input regression in Qwen models is unrelated to trainer.evaluate() post-train failure." }, { "left": "issue:43122", "right": "issue:43906", "accept": false, - "reason": "A version-to-version tokenization change is not the same as an isolated reproduction of another issue." + "reason": "The isolated reproduction reference is too vague to establish the same underlying tokenization bug." }, { "left": "issue:44977", "right": "issue:45356", "accept": false, - "reason": "Different Qwen3.5 regressions: flash-attention generation behavior vs tokenizer codec/fix_mistral_regex handling." - }, - { - "left": "issue:30064", - "right": "issue:33666", - "accept": false, - "reason": "Different domains: image processor void-segmentation handling vs Qwen2-VL multi-GPU training." + "reason": "Qwen3.5 flash-attention generation bug is unrelated to Kimi-K2.5 tokenizer codec handling." } ] }, - "evaluator_result": null, - "evaluator_used": false, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet and stays conservative: it identifies one plausible duplicate pair (44466/45005) based on tightly related tied-weights v5 titles, and the other soft-edge rejections are appropriately cautious given only title-level evidence. No overstatement stands out." + }, + "evaluator_used": true, "retried": false, - "accepted_nontrivial_soft_edge": false, + "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null }, @@ -5569,16 +5565,15 @@ "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-15354-367", "nodes": [ + "issue:30064", "issue:32090", "issue:33666", "issue:34567", "issue:36010", "issue:41720", "issue:43012", - "issue:43257", "issue:43262", "issue:43278", - "issue:43299", "issue:43329", "issue:43334", "issue:43452", @@ -5619,23 +5614,24 @@ "issue:45357" ], "soft_pairs": [ + "issue:30064|issue:33666", "issue:43653|issue:43927", "issue:44464|issue:44977", "issue:43329|issue:45325", - "issue:43577|issue:44554", "issue:43531|issue:43742", - "issue:43550|issue:44361", + "issue:43577|issue:44554", "issue:44479|issue:44625", + "issue:43550|issue:44361", "issue:43012|issue:43577", "issue:43504|issue:43720", "issue:41720|issue:44464", "issue:43504|issue:43550", "issue:32090|issue:33666", "issue:32090|issue:34567", - "issue:43504|issue:44610", + "issue:43452|issue:44451", "issue:43262|issue:45084", "issue:43278|issue:43688", - "issue:43452|issue:44451", + "issue:43504|issue:44610", "issue:45200|issue:45229", "issue:44849|issue:45357", "issue:43334|issue:43572", @@ -5646,232 +5642,224 @@ "issue:43525|issue:43572", "issue:33666|issue:36010", "issue:43720|issue:43927", - "issue:43606|issue:43927", "issue:43688|issue:43873", - "issue:43278|issue:43844", - "issue:43257|issue:43299", - "issue:43720|issue:44361" + "issue:43606|issue:43927", + "issue:43278|issue:43844" ], - "prepared_review_unit_hash": "f7fa68e3938d21a432d232a640da192891936c75fa151571de90e78c8713f0e0", + "prepared_review_unit_hash": "1d3922b93ccb8be819b125c6018785d50e044f1f45cd62514f578bfe41615d57", "component_budget": { "node_count": 367, "item_count": 367, "soft_pair_count": 1370, - "serialized_chars": 706987, + "serialized_chars": 706986, "estimated_input_tokens": 176747, "estimated_eval_tokens": 353750 }, "budget": { - "node_count": 48, - "item_count": 48, - "soft_pair_count": 32, - "serialized_chars": 37293, - "estimated_input_tokens": 9324, - "estimated_eval_tokens": 18904 + "node_count": 47, + "item_count": 47, + "soft_pair_count": 31, + "serialized_chars": 36418, + "estimated_input_tokens": 9105, + "estimated_eval_tokens": 18466 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The cluster is heterogeneous: most pairs are only loosely related by subsystem or model family, not the same underlying bug. No soft-edge pair should be merged as a duplicate.", - "confidence": 0.91, - "canonical_issue_reason": "No single issue is a safe canonical for this set; the items span unrelated regressions in loading, generation, tokenizer/config handling, multimodal/video logic, and training.", + "summary": "These items are largely unrelated issue reports across different models, processors, loading paths, and training/runtime bugs; the soft pairs share only superficial thematic overlap, not the same underlying defect.", + "confidence": 0.96, + "canonical_issue_reason": "No single canonical duplicate stands out: the set is heterogeneous, spanning unrelated bugs in tokenizer/config loading, multimodal processing, training, quantization, and generation.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 43257 is the most concrete and actionable bug report in the set, but it still does not represent the other items closely enough to serve as a cluster-wide duplicate target.", + "best_issue_reason": "If one representative issue must be chosen, issue 43873 is the broadest/high-activity report, but it is not a duplicate hub and should not be treated as the canonical merge target.", "best_pr_reason": null, "soft_edge_verdicts": [ + { + "left": "issue:30064", + "right": "issue:33666", + "accept": false, + "reason": "Different problems: void segmentation map handling vs Qwen2-VL multi-GPU training." + }, { "left": "issue:43653", "right": "issue:43927", "accept": false, - "reason": "Both involve tokenizer/config behavior, but one is about BigBird special-token registration and the other about custom token IDs surviving save/load." + "reason": "Both involve token-related behavior, but one is BigBird special-token decode and the other is DiaConfig save/load token IDs; different bug." }, { "left": "issue:44464", "right": "issue:44977", "accept": false, - "reason": "Both are generation problems, but they affect different models and different code paths: compiled forward inconsistency vs flash-attention generation." + "reason": "Both affect generation, but one is compiled-forward inconsistency and the other is flash-attention behavior in Qwen3.5." }, { "left": "issue:43329", "right": "issue:45325", "accept": false, - "reason": "Both touch Qwen multimodal/video logic, but one is an undefined variable bug in token counting and the other is a rope-index temporal position scaling regression." - }, - { - "left": "issue:43577", - "right": "issue:44554", - "accept": false, - "reason": "These are unrelated: one is a dtype-loading issue in Blip2, the other is an attention correctness problem on MPS." + "reason": "Both touch VLM multimodal internals, but one is an undefined video token-count path and the other is still-image rope indexing." }, { "left": "issue:43531", "right": "issue:43742", "accept": false, - "reason": "Different failures and different models: sliding_window behavior in Qwen3-MoE versus a key error loading MobileLLM-125M." + "reason": "Sliding-window MoE bug vs a model-loading key error; not the same code path." }, { - "left": "issue:43550", - "right": "issue:44361", + "left": "issue:43577", + "right": "issue:44554", "accept": false, - "reason": "No common bug: torch.compile/SDPA failure in Bamba versus an MLuke tokenizer attribute error on tasks." + "reason": "One is dtype propagation on load, the other is an MPS attention correctness issue." }, { "left": "issue:44479", "right": "issue:44625", "accept": false, - "reason": "One is a video-input regression across Qwen VL models; the other is missing num_labels propagation in Qwen3.5." + "reason": "Video-input regression in multiple Qwen VL variants vs missing num_labels propagation in Qwen3.5." + }, + { + "left": "issue:43550", + "right": "issue:44361", + "accept": false, + "reason": "torch.compile + SDPA failure in Bamba is unrelated to MLukeTokenizer task AttributeError." }, { "left": "issue:43012", "right": "issue:43577", "accept": false, - "reason": "Both mention precision/dtype, but one is a PyTorch warning during compile and the other is an actual model/qformer dtype mismatch on load." + "reason": "Both concern precision/dtype, but one is a warning during bfloat16 compilation and the other is incorrect model dtype on load." }, { "left": "issue:43504", "right": "issue:43720", "accept": false, - "reason": "Different model families and bugs: legacy field blocking BEiT semantic segmentation load versus packed weights not unpacked in BitNet accelerate loading." + "reason": "Legacy-field loading bug in BEiT vs packed-weight unpacking bug in BitNet accelerate loading." }, { "left": "issue:41720", "right": "issue:44464", "accept": false, - "reason": "Different code paths and symptoms: Qwen3 auto device-mapping CUDA assert versus chunked generation divergence under compiled forward." + "reason": "CUDA assert under auto device mapping is unrelated to chunked generation divergence with compiled forward." }, { "left": "issue:43504", "right": "issue:43550", "accept": false, - "reason": "These are separate loader/runtime issues affecting different models and mechanisms." + "reason": "Different models and failures: BEiT config loading vs Bamba torch.compile/SDPA." }, { "left": "issue:32090", "right": "issue:33666", "accept": false, - "reason": "Both are training/multi-GPU related, but one is a NoneType broadcast error and the other is a Qwen2-VL multi-GPU training question/bug." + "reason": "Trainer NoneType broadcast error is not the same as Qwen2-VL multi-GPU training." }, { "left": "issue:32090", "right": "issue:34567", "accept": false, - "reason": "Trainer broadcast failure and TrainerState token counter not updating are distinct training bugs." + "reason": "One is a distributed broadcast TypeError; the other is a TrainerState counter not updating." }, { - "left": "issue:43504", - "right": "issue:44610", + "left": "issue:43452", + "right": "issue:44451", "accept": false, - "reason": "Different models and failures: BEiT legacy-field loading versus OmDet-Turbo processor/model input-size mismatch." + "reason": "gguf_file path handling issue vs a generic inability to load a specific model." }, { "left": "issue:43262", "right": "issue:45084", "accept": false, - "reason": "Audio chat-template sampling-rate defaulting and template-node compilation errors are unrelated." + "reason": "Audio sampling-rate default bug is unrelated to template-node compilation errors." }, { "left": "issue:43278", "right": "issue:43688", "accept": false, - "reason": "Both involve numerical behavior, but one is dtype drift in evaluate and the other is auxiliary-loss normalization." + "reason": "Both mention training behavior, but one is dtype drift between train/eval and the other is auxiliary-loss normalization." }, { - "left": "issue:43452", - "right": "issue:44451", + "left": "issue:43504", + "right": "issue:44610", "accept": false, - "reason": "Different loading failures: gguf_file breaks tokenizer/model from_pretrained paths, while ScandiBERT cannot be loaded for another reason." + "reason": "Pretrained preset loading bug vs processor/model image-size mismatch." }, { "left": "issue:45200", "right": "issue:45229", "accept": false, - "reason": "Same model family, but one is a missing mm_token_type_ids default and the other is a multi-GPU inference OOM." + "reason": "mm_token_type_ids defaulting issue is unrelated to multi-GPU inference OOM." }, { "left": "issue:44849", "right": "issue:45357", "accept": false, - "reason": "Both concern Qwen3.5, but one is output_hidden_states behavior and the other is save_pretrained writing incorrect visual-encoder keys." + "reason": "Hidden-states bug in Qwen3.5 vs incorrect visual encoder keys saved on pretrained serialization." }, { "left": "issue:43334", "right": "issue:43572", "accept": false, - "reason": "Similar missing-field shape, but different configs/models and different attributes: pad_token_id versus pad_token_idx." + "reason": "Both are missing-config-field regressions, but for different models, different fields, and different failure modes." }, { "left": "issue:43756", "right": "issue:44913", "accept": false, - "reason": "RoPE-related but not the same bug: Smollm3 layer-dropping versus GPTNeoX rotary_pct reload defaulting." + "reason": "RoPE-layer configuration bug vs rotary_pct not persisting on reload; similar theme, different defect." }, { "left": "issue:43493", "right": "issue:43825", "accept": false, - "reason": "Different scope: SigLIP2 implementation discrepancy versus pipeline error messaging about translation support." + "reason": "SigLIP2 implementation discrepancy is unrelated to pipeline task support messaging." }, { "left": "issue:43540", "right": "issue:43716", "accept": false, - "reason": "Different multimodal bugs: video-processing ValueError in Qwen3OmniMoe versus image preprocessor/model dtype mismatch in Mistral-3." + "reason": "Video-processing ValueError in Qwen3OmniMoe vs image-preprocessor/model dtype mismatch in Mistral-3." }, { "left": "issue:43742", "right": "issue:44821", "accept": false, - "reason": "KeyError on MobileLLM loading and AutoImageProcessor-from-URL loading are unrelated." + "reason": "Model key error on load vs AutoImageProcessor URL loading; different subsystems." }, { "left": "issue:43525", "right": "issue:43572", "accept": false, - "reason": "Both are missing token-id attribute issues after v5 changes, but they affect different config classes and different model families." + "reason": "Both mention missing pad token fields, but they affect different configs and attributes." }, { "left": "issue:33666", "right": "issue:36010", "accept": false, - "reason": "One is a Qwen2-VL multi-GPU training topic; the other is an import failure for GenerationMixin." + "reason": "Qwen2-VL multi-GPU training is unrelated to GenerationMixin import failure." }, { "left": "issue:43720", "right": "issue:43927", "accept": false, - "reason": "Both involve loading/accelerate behavior, but packed-weight unpacking in BitNet is unrelated to DiaConfig custom token IDs being lost." - }, - { - "left": "issue:43606", - "right": "issue:43927", - "accept": false, - "reason": "CPU offload device mismatch in Bark and DiaConfig token-ID loss are different bugs." + "reason": "Packed-weight loading bug vs config token-ID loss on save/load." }, { "left": "issue:43688", "right": "issue:43873", "accept": false, - "reason": "Different optimization issues: auxiliary-loss normalization versus quantization/offloading behavior." - }, - { - "left": "issue:43278", - "right": "issue:43844", - "accept": false, - "reason": "Both mention gradients/dtypes in training contexts, but one is dtype drift and the other is ZeRO-3 gradient inflation." + "reason": "Aux-loss normalization is unrelated to offloading with quantization." }, { - "left": "issue:43257", - "right": "issue:43299", + "left": "issue:43606", + "right": "issue:43927", "accept": false, - "reason": "Both are Qwen3 MoE loading issues, but one is accelerate/deepspeed weight conversion and the other is a dev-version loading regression in Qwen3VL MoE." + "reason": "CPU offload device mismatch is unrelated to DiaConfig token-ID persistence." }, { - "left": "issue:43720", - "right": "issue:44361", + "left": "issue:43278", + "right": "issue:43844", "accept": false, - "reason": "Unrelated bugs: BitNet accelerate loading versus MLuke tokenizer task handling." + "reason": "Embedding dtype drift vs abnormal gradient growth under DeepSpeed ZeRO-3 are different issues." } ] }, @@ -5895,15 +5883,14 @@ "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-15354-367", "nodes": [ - "issue:15354", "issue:33357", "issue:33666", "issue:36331", - "issue:38175", "issue:39692", "issue:43257", "issue:43278", "issue:43295", + "issue:43299", "issue:43329", "issue:43493", "issue:43504", @@ -5915,6 +5902,7 @@ "issue:43653", "issue:43688", "issue:43701", + "issue:43720", "issue:43746", "issue:43756", "issue:43824", @@ -5944,19 +5932,21 @@ "issue:45584" ], "soft_pairs": [ + "issue:43257|issue:43299", + "issue:43720|issue:44361", "issue:39692|issue:43493", "issue:44964|issue:45405", "issue:43540|issue:44977", "issue:43329|issue:43606", "issue:44488|issue:44843", - "issue:43493|issue:43844", "issue:33666|issue:36331", + "issue:43493|issue:43844", "issue:43756|issue:44060", "issue:44589|issue:45520", "issue:44625|issue:45584", "issue:43646|issue:44466", - "issue:44521|issue:45290", "issue:43688|issue:43746", + "issue:44521|issue:45290", "issue:44423|issue:44734", "issue:43824|issue:44589", "issue:33357|issue:33666", @@ -5964,218 +5954,223 @@ "issue:43504|issue:43606", "issue:43994|issue:44493", "issue:44464|issue:45584", - "issue:43504|issue:43653", "issue:43927|issue:44361", "issue:43572|issue:44291", + "issue:43504|issue:43653", "issue:43575|issue:45405", "issue:43278|issue:43701", "issue:44792|issue:45405", "issue:43257|issue:45405", - "issue:43606|issue:44112", - "issue:15354|issue:38175" + "issue:43606|issue:44112" ], - "prepared_review_unit_hash": "db2d4417fb9d0ba7dccf03da1f946f33704a3e5f6b310e87796d8f447966fca8", + "prepared_review_unit_hash": "01bf351f947c48ffb3ef4dfa6fcf840806a16ec36dc1b01d62ade80582030c3b", "component_budget": { "node_count": 367, "item_count": 367, "soft_pair_count": 1370, - "serialized_chars": 706987, + "serialized_chars": 706986, "estimated_input_tokens": 176747, "estimated_eval_tokens": 353750 }, "budget": { "node_count": 47, "item_count": 47, - "soft_pair_count": 29, - "serialized_chars": 36141, - "estimated_input_tokens": 9036, - "estimated_eval_tokens": 18328 + "soft_pair_count": 30, + "serialized_chars": 36407, + "estimated_input_tokens": 9102, + "estimated_eval_tokens": 18460 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is highly heterogeneous: the items span unrelated bugs in multimodal processing, tokenizers, model loading, training, serving, serialization, and docs. None of the soft pairs look like the same underlying defect, so all soft edges are rejected and there is no safe single canonical issue for the cluster.", - "confidence": 0.95, - "canonical_issue_reason": "No safe canonical issue: the cluster is too broad and the apparent overlaps are only superficial (same model family or subsystem, but different code paths and failure modes).", + "summary": "This cluster is highly heterogeneous: most items are unrelated model-loading, multimodal, tokenizer, training, and CI regressions. The soft pairs share broad subsystem keywords at best, but not the same concrete bug/change, so none should be merged as duplicates.", + "confidence": 0.19, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "No issue here is a strong global representative; the closest topics still map to distinct bugs rather than one duplicate thread.", + "best_issue_reason": null, "best_pr_reason": null, "soft_edge_verdicts": [ + { + "left": "issue:43257", + "right": "issue:43299", + "accept": false, + "reason": "Both involve Qwen3 MOE loading, but one is about missing weight conversion under accelerate/deepspeed and the other is a version regression in Qwen3VL model loading; different failure modes and code paths." + }, + { + "left": "issue:43720", + "right": "issue:44361", + "accept": false, + "reason": "One is a packed-weight loading bug in BitNet/accelerate; the other is an AttributeError in MLukeTokenizer task handling. No shared concrete defect." + }, { "left": "issue:39692", "right": "issue:43493", "accept": false, - "reason": "Both involve SigLIP2, but one is a docs/example mismatch and quantization failure while the other is a model-vs-JAX implementation discrepancy." + "reason": "Both mention SigLIP2, but one is a documentation/example mismatch while the other is an implementation-vs-JAX discrepancy in the model itself." }, { "left": "issue:44964", "right": "issue:45405", "accept": false, - "reason": "Different problem types: multimodal model loading regression vs a PEFT version bump / unreleased dependency issue." + "reason": "A model loading regression is not the same as a dependency version bump issue; one is about loading Phi-4 multimodal, the other about an unreleased MIN_PEFT_VERSION." }, { "left": "issue:43540", "right": "issue:44977", "accept": false, - "reason": "Both are Qwen-related, but one is a video-processing ValueError in Qwen3OmniMoe and the other is a flash-attention generation bug in Qwen3.5." + "reason": "Both are Qwen3.5-related, but one is video-input processing in Qwen3OmniMoe and the other is flash-attention generation behavior; distinct bugs." }, { "left": "issue:43329", "right": "issue:43606", "accept": false, - "reason": "Undefined variable in multimodal token counting is unrelated to the bark-small CPU offload device mismatch." + "reason": "Undefined variables in multimodal token counting is unrelated to a CPU-offload device mismatch in bark-small." }, { "left": "issue:44488", "right": "issue:44843", "accept": false, - "reason": "One is a model loading failure for cjvt/sleng-bert; the other is an offline-mode regression in AutoTokenizer's Mistral regex patch." + "reason": "One is a model loading failure for a specific checkpoint, the other is offline-mode breakage from an unconditional hub metadata call; different root causes." }, { - "left": "issue:43493", - "right": "issue:43844", + "left": "issue:33666", + "right": "issue:36331", "accept": false, - "reason": "SigLIP2 implementation mismatch and ZeRO-3 gradient growth are different bugs with different failure surfaces." + "reason": "Multi-GPU training support and a custom trainer compute_loss signature mismatch are separate training issues." }, { - "left": "issue:33666", - "right": "issue:36331", + "left": "issue:43493", + "right": "issue:43844", "accept": false, - "reason": "Multi-GPU training support for Qwen2-VL is unrelated to the CustomTrainer.compute_loss signature regression." + "reason": "Both are correctness bugs, but they affect different models and mechanisms: SigLIP2 implementation mismatch versus gradient inflation under ZeRO-3." }, { "left": "issue:43756", "right": "issue:44060", "accept": false, - "reason": "Smollm3 RoPE-layer omission and Qwen3-Next tied-weights warning are different model/config issues." + "reason": "Smollm3 RoPE-layer truncation and Qwen3-Next tied-weights warnings are unrelated model-specific configuration bugs." }, { "left": "issue:44589", "right": "issue:45520", "accept": false, - "reason": "Float8 storage deserialization failure is unrelated to the Python 3.13 flash_attn import key error." + "reason": "Both mention missing storage/dispatch paths, but one is a Float8 storage lookup failure and the other is a flash_attn import_utils KeyError on Python 3.13." }, { "left": "issue:44625", "right": "issue:45584", "accept": false, - "reason": "num_labels propagation in Qwen3.5 is distinct from Whisper's empty-transcription failure after align_special_tokens." + "reason": "num_labels propagation in Qwen3.5 is unrelated to Whisper empty-transcription generation after align_special_tokens." }, { "left": "issue:43646", "right": "issue:44466", "accept": false, - "reason": "Transformers 5.0 custom model initialization breakage is not the same as device-dependent lm_head serialization inconsistency." + "reason": "Both are v5 regressions, but one is custom model initialization and the other is inconsistent lm_head serialization; different code paths." }, { - "left": "issue:44521", - "right": "issue:45290", + "left": "issue:43688", + "right": "issue:43746", "accept": false, - "reason": "All-zero assistant masks for multimodal inputs and tool-call content crashes are separate chat-template bugs." + "reason": "Auxiliary-loss normalization in MoE models is not the same as GraniteSpeech PEFT adapter loading from local checkpoints." }, { - "left": "issue:43688", - "right": "issue:43746", + "left": "issue:44521", + "right": "issue:45290", "accept": false, - "reason": "Auxiliary loss normalization in MoE models does not match the PEFT/local-checkpoint loading issue in GraniteSpeech." + "reason": "Both concern apply_chat_template, but one is all-zero assistant masks for multimodal inputs and the other is tool-call messages with no content. Different bugs." }, { "left": "issue:44423", "right": "issue:44734", "accept": false, - "reason": "Continuous-batching multimodal crash and KV-cache continuation indexing bug affect different serve paths." + "reason": "Both hit serve/runtime multimodal paths, but one crashes on string-to-.to() handling and the other on KV-cache continuation tensor indexing." }, { "left": "issue:43824", "right": "issue:44589", "accept": false, - "reason": "Importing Qwen2.5-VL and missing Float8 storage objects are unrelated failures." + "reason": "ImportError for a renamed model class is unrelated to a missing Float8 storage object during serialization/loading." }, { "left": "issue:33357", "right": "issue:33666", "accept": false, - "reason": "Mac bus error on CLIP model load is unrelated to the Qwen2-VL multi-GPU training issue." + "reason": "A MacOS bus error loading CLIP is unrelated to Qwen2-VL multi-GPU training." }, { "left": "issue:39692", "right": "issue:43295", "accept": false, - "reason": "Docs/example errors for SigLIP2 are distinct from the processor.tokenizer regression in v4.57.5." + "reason": "A broken SigLIP2 example in docs is not the same as a regression in processor.tokenizer access and image-to-tokenizer passing." }, { "left": "issue:43504", "right": "issue:43606", "accept": false, - "reason": "Legacy-field loading failure in BeitForSemanticSegmentation is not the same as the bark-small CPU offload device mismatch." + "reason": "A legacy field preventing BEiT semantic-segmentation loading is unrelated to a bark-small CPU offload device mismatch." }, { "left": "issue:43994", "right": "issue:44493", "accept": false, - "reason": "SigLIP2 nonsensical outputs are a model behavior issue, while unexpected position-id keys point to a different config/state bug." + "reason": "Both are model-quality/loading complaints, but one is SigLIP2 nonsensical outputs and the other is unexpected position-id keys across many models." }, { "left": "issue:44464", "right": "issue:45584", "accept": false, - "reason": "Compiled-forward chunked generation inconsistency is unrelated to Whisper empty transcription after token alignment." - }, - { - "left": "issue:43504", - "right": "issue:43653", - "accept": false, - "reason": "Beit legacy-field loading and BigBirdTokenizer special-token registration are different issues in different subsystems." + "reason": "Chunked generation inconsistency with compiled forward is unrelated to Whisper empty-transcription failure after token alignment." }, { "left": "issue:43927", "right": "issue:44361", "accept": false, - "reason": "DiaConfig losing custom token IDs and MLukeTokenizer AttributeError are separate configuration/tokenizer bugs." + "reason": "DiaConfig save/load token ID loss and MLukeTokenizer task AttributeError are different layers and failure modes." }, { "left": "issue:43572", "right": "issue:44291", "accept": false, - "reason": "StableLmConfig missing pad_token_idx and init_empty_weights receiving _is_hf_initialized are different compatibility regressions." + "reason": "Both involve initialization/loading changes in v5, but missing pad_token_idx in StableLmConfig is not the same as an unexpected _is_hf_initialized argument in init_empty_weights." + }, + { + "left": "issue:43504", + "right": "issue:43653", + "accept": false, + "reason": "BEiT legacy-field loading and BigBirdTokenizer special-token registration are unrelated bugs." }, { "left": "issue:43575", "right": "issue:45405", "accept": false, - "reason": "Tensor-parallel OOM on Qwen2-57B is unrelated to the unreleased PEFT version bump." + "reason": "TP OOM for a large Qwen2 model is unrelated to a dependency version bump for PEFT." }, { "left": "issue:43278", "right": "issue:43701", "accept": false, - "reason": "Embedding dtype drift during evaluate and resume_from_checkpoint key mismatch are distinct training-state bugs." + "reason": "Embedding dtype drift between train/eval is unrelated to resume_from_checkpoint key mismatches." }, { "left": "issue:44792", "right": "issue:45405", "accept": false, - "reason": "A janus image-generation test failure is not the same as the PEFT release/versioning issue." + "reason": "A failing janus image-generation test does not match a PEFT minimum-version issue." }, { "left": "issue:43257", "right": "issue:45405", "accept": false, - "reason": "Qwen3 MOE weight conversion under accelerate+deepspeed is unrelated to the PEFT minimum-version bump." + "reason": "Qwen3 MOE weight conversion/loading and a PEFT version requirement are unrelated." }, { "left": "issue:43606", "right": "issue:44112", "accept": false, - "reason": "Both are GraniteSpeech CI-related, but one is a device mismatch and the other is a stale test override; different root causes." - }, - { - "left": "issue:15354", - "right": "issue:38175", - "accept": false, - "reason": "Old torch.jit.script GeneratorExp export failure and SigLIP2 zero-probability outputs are completely unrelated." + "reason": "Both touch GraniteSpeech/device handling, but one is bark-small CPU offload mismatch and the other is a stale CI test override. Not the same defect." } ] }, @@ -6199,7 +6194,9 @@ "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-15354-367", "nodes": [ + "issue:15354", "issue:37428", + "issue:38175", "issue:39401", "issue:43232", "issue:43295", @@ -6219,7 +6216,6 @@ "issue:43957", "issue:43975", "issue:44315", - "issue:44442", "issue:44464", "issue:44466", "issue:44514", @@ -6248,9 +6244,10 @@ "issue:45563" ], "soft_pairs": [ + "issue:15354|issue:38175", "issue:44554|issue:45446", - "issue:43582|issue:44977", "issue:45084|issue:45520", + "issue:43582|issue:44977", "issue:43540|issue:45200", "issue:43844|issue:44964", "issue:45072|issue:45198", @@ -6258,234 +6255,215 @@ "issue:39401|issue:44568", "issue:37428|issue:43957", "issue:44315|issue:44849", - "issue:44554|issue:45356", "issue:43334|issue:43525", + "issue:44554|issue:45356", "issue:43295|issue:44610", "issue:43493|issue:43701", "issue:45200|issue:45325", - "issue:44898|issue:45412", "issue:44466|issue:45127", + "issue:44898|issue:45412", "issue:43873|issue:44464", "issue:44792|issue:45563", "issue:43388|issue:43526", - "issue:44514|issue:44521", "issue:43844|issue:44936", + "issue:44514|issue:44521", "issue:37428|issue:43824", "issue:43572|issue:44987", "issue:43295|issue:43450", "issue:43572|issue:44589", "issue:43824|issue:43975", - "issue:44442|issue:45198", "issue:43295|issue:43873" ], - "prepared_review_unit_hash": "fcd39cbce4303384afbc27fd90a43541b3796a625f6d0baf111b0f779ee8d547", + "prepared_review_unit_hash": "0a150c296e824a0b32e894eabaf61d94e42bc53278af1318b794133910c7a4e7", "component_budget": { "node_count": 367, "item_count": 367, "soft_pair_count": 1370, - "serialized_chars": 706987, + "serialized_chars": 706986, "estimated_input_tokens": 176747, "estimated_eval_tokens": 353750 }, "budget": { - "node_count": 47, - "item_count": 47, - "soft_pair_count": 29, - "serialized_chars": 35993, - "estimated_input_tokens": 8999, - "estimated_eval_tokens": 18254 + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36626, + "estimated_input_tokens": 9157, + "estimated_eval_tokens": 18570 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a mix of unrelated Transformers bug reports. The only strong duplicate-like edges are two Qwen2.5-VL chat-template reports and two model-loading regressions about missing pad_token_id on config objects.", - "confidence": 0.64, - "canonical_issue_reason": "issue:44521 is the clearest and broadest statement of the shared Qwen2.5-VL apply_chat_template/multimodal masking bug; issue:44514 looks like a specific batched-padding=False manifestation of the same path.", + "summary": "This cluster is mostly heterogeneous: several unrelated Transformers regressions are mixed together. The only clear duplicate-style pair is the missing `pad_token_id` config AttributeError; most other soft edges are similar only at a broad subsystem level and should stay separate.", + "confidence": 0.38, + "canonical_issue_reason": "#43334 is the clearest representative of the `pad_token_id`-missing config regression and is earlier/more discussed than its sibling.", "canonical_pr_reason": null, - "best_issue_reason": "issue:44521 is the best representative for the duplicated multimodal chat-template subcluster: it is open, clearly scoped, and covers the broader assistant_masks failure mode.", + "best_issue_reason": "#43334 is the best overall issue to anchor the small `pad_token_id` subcluster; it states the failure mode cleanly and matches the only strong duplicate candidate.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:44554", - "right": "issue:45446", + "left": "issue:15354", + "right": "issue:38175", "accept": false, - "reason": "Different subsystems and failures: MPS attention correctness vs a flex_attention import/version check." + "reason": "Different models and failures: TorchScript GeneratorExp export vs SigLIP2 zero-probability output." }, { - "left": "issue:43582", - "right": "issue:44977", + "left": "issue:44554", + "right": "issue:45446", "accept": false, - "reason": "Unrelated bugs: Apple Silicon allocator warmup TypeError vs Qwen3.5 flash-attention generation behavior." + "reason": "Both touch attention/flex attention areas, but one is an MPS correctness bug and the other is a version-check import bug." }, { "left": "issue:45084", "right": "issue:45520", "accept": false, - "reason": "Different import/compile failures in separate code paths; no clear shared bug." + "reason": "Unrelated import/compile failures with no shared concrete code path." + }, + { + "left": "issue:43582", + "right": "issue:44977", + "accept": false, + "reason": "AppleSilicon allocator warmup TypeError and Qwen3.5 flash-attention generation issue are different bugs." }, { "left": "issue:43540", "right": "issue:45200", "accept": false, - "reason": "Both are multimodal-related, but one is video preprocessing in Qwen3OmniMoe and the other is Gemma 4 token-type defaults." + "reason": "Both are multimodal/model-processing issues, but one is video input handling and the other is Gemma 4 token-type defaults." }, { "left": "issue:43844", "right": "issue:44964", "accept": false, - "reason": "Different problems: ZeRO-3 gradient behavior vs loading Phi-4 multimodal." + "reason": "DeepSpeed/ZeRO gradient instability is unrelated to Phi-4 multimodal model loading." }, { "left": "issue:45072", "right": "issue:45198", "accept": false, - "reason": "Different model families and failures: dtype mismatch in inference vs speech tokenizer save/load failure." + "reason": "Different subsystems: dtype mismatch in inference vs Wav2Vec2 save/tokenization failure." }, { "left": "issue:43232", "right": "issue:44849", "accept": false, - "reason": "Different generation bugs with different triggers and code paths." + "reason": "Generation kwargs syncing and Qwen3.5 hidden-state output are separate generation behaviors." }, { "left": "issue:39401", "right": "issue:44568", "accept": false, - "reason": "Both tokenizer issues, but different models and different symptoms (offset mapping vs special token insertion)." + "reason": "Both are tokenizer bugs, but they affect different tokenizers and different behavior (`offset_mapping` vs special-token insertion)." }, { "left": "issue:37428", "right": "issue:43957", "accept": false, - "reason": "Both involve loading/importing, but the missing flash-attention helper and meta-device loading failure are not the same bug." + "reason": "Flash-attention import failure and meta-device model loading breakage are distinct issues." }, { "left": "issue:44315", "right": "issue:44849", "accept": false, - "reason": "Unrelated: Liger Kernel application in model_init vs a Qwen3.5 hidden-states bug." - }, - { - "left": "issue:44554", - "right": "issue:45356", - "accept": false, - "reason": "Different models and fixes; MPS attention correctness is unrelated to Qwen3.5 tokenizer regression." + "reason": "Liger Kernel application in `model_init` is unrelated to Qwen3.5 hidden-state output behavior." }, { "left": "issue:43334", "right": "issue:43525", "accept": true, - "reason": "Both report the same regression pattern: loading fails because a config object is missing pad_token_id." + "reason": "Both report the same `AttributeError` caused by config objects missing `pad_token_id`; likely one fix addresses both." + }, + { + "left": "issue:44554", + "right": "issue:45356", + "accept": false, + "reason": "Different problems: MPS attention correctness vs Kimi-K2.5 tokenizer codec/regex regression." }, { "left": "issue:43295", "right": "issue:44610", "accept": false, - "reason": "Both touch processors, but one is a processor.tokenizer regression and the other is an input-size mismatch for OmDet-Turbo." + "reason": "Processor/tokenizer API regression and model loading with `torch.device(\"meta\")` are not the same bug." }, { "left": "issue:43493", "right": "issue:43701", "accept": false, - "reason": "Different bugs: SigLIP2 implementation mismatch vs checkpoint resume key mismatch." + "reason": "SigLIP2 implementation discrepancy and resume checkpoint key mismatch do not share a concrete failure path." }, { "left": "issue:45200", "right": "issue:45325", "accept": false, - "reason": "Different multimodal bugs in different models: missing mm_token_type_ids default vs Qwen2.5-VL rope index scaling." + "reason": "Both are Qwen2.5-VL issues, but one is token-type defaults and the other is temporal position-id scaling." }, { - "left": "issue:44898", - "right": "issue:45412", + "left": "issue:44466", + "right": "issue:45127", "accept": false, - "reason": "Image resolution/interpolation failure is unrelated to RT-DETR memory release behavior." + "reason": "Serialization of tied weights and LoRA merge collapse are related only broadly; the concrete bugs differ." }, { - "left": "issue:44466", - "right": "issue:45127", + "left": "issue:44898", + "right": "issue:45412", "accept": false, - "reason": "Both involve tied weights/serialization themes, but the concrete failures differ enough that they do not look like the same bug." + "reason": "Perceiver resolution handling and RT-DETR memory release are unrelated." }, { "left": "issue:43873", "right": "issue:44464", "accept": false, - "reason": "Unrelated: quantization/offloading issue vs compiled-forward chunked generation inconsistency." + "reason": "Quantized offloading and compiled-forward chunked generation are different behaviors." }, { "left": "issue:44792", "right": "issue:45563", "accept": false, - "reason": "Different generation/test warning issues with no shared code-path evidence." + "reason": "Janus image-generation test failure and stale paged-generate warning are not the same bug." }, { "left": "issue:43388", "right": "issue:43526", "accept": false, - "reason": "Both mention labels, but one is gather_for_metrics batching and the other is BeitImageProcessorFast label reduction." - }, - { - "left": "issue:44514", - "right": "issue:44521", - "accept": true, - "reason": "Same Qwen2.5-VL apply_chat_template path; one is a crash on batched padding=False input, the other is the same family of multimodal masking failure." + "reason": "Both involve labels, but one is `gather_for_metrics` dropping tuple elements and the other is `BeitImageProcessorFast` reduce_labels truncation." }, { "left": "issue:43844", "right": "issue:44936", "accept": false, - "reason": "Different training/evaluation problems with no shared underlying change." + "reason": "DeepSpeed gradient growth and trainer evaluate-after-train failure are separate training issues." }, { - "left": "issue:37428", - "right": "issue:43824", - "accept": false, - "reason": "Both are import errors, but they concern different symbols and different subsystems." - }, - { - "left": "issue:43572", - "right": "issue:44987", + "left": "issue:44514", + "right": "issue:44521", "accept": false, - "reason": "Different loading failures; no evidence they stem from the same pad_token_id-related regression." + "reason": "Both involve `apply_chat_template`, but one is a batched padding=False crash and the other is incorrect assistant masks for multimodal inputs." }, { - "left": "issue:43295", - "right": "issue:43450", + "left": "issue:37428", + "right": "issue:43824", "accept": false, - "reason": "Both are processor-related, but one is custom tokenizer/image handling and the other is batched video shape handling." + "reason": "Both are import errors, but they concern different missing symbols and different code paths." }, { "left": "issue:43572", - "right": "issue:44589", - "accept": false, - "reason": "Unrelated config default issue vs a PyTorch storage lookup TypeError." - }, - { - "left": "issue:43824", - "right": "issue:43975", - "accept": false, - "reason": "Different tokenizer/model bugs: missing export vs incorrect detokenization." - }, - { - "left": "issue:44442", - "right": "issue:45198", + "right": "issue:44987", "accept": false, - "reason": "Both are speech/tokenizer related, but they are different models and different failure modes." + "reason": "StableLm config `pad_token_idx` regression and loading `physical-intelligence/fast` under v5.1 are not the same underlying bug." }, { "left": "issue:43295", "right": "issue:43873", "accept": false, - "reason": "Unrelated processor regression vs quantization/offloading behavior." + "reason": "Processor/tokenizer regression and quantized offloading failure do not look like one fixable code path." } ] }, "evaluator_result": { "accept": true, - "feedback": "Grounded overall. The summary stays conservative about the cluster being mostly unrelated reports, and the two accepted soft edges are plausible duplicate/same-bug pairs based on the packet: the Qwen2.5-VL chat-template/multimodal masking issues, and the pad_token_id config regressions. I don\u2019t see an overstatement that would require rejection." + "feedback": "The assessment is grounded in the packet and stays conservative. The pad_token_id pair (43334/43525) is a defensible duplicate-style match, and the remaining soft-edge rejections do not overclaim shared evidence. No PR-pair mergeability issue is present here." }, "evaluator_used": true, "retried": false, @@ -6514,11 +6492,9 @@ "issue:43262", "issue:43278", "issue:43295", - "issue:43335", "issue:43425", "issue:43450", "issue:43493", - "issue:43550", "issue:43575", "issue:43606", "issue:43611", @@ -6534,6 +6510,7 @@ "issue:44112", "issue:44206", "issue:44360", + "issue:44442", "issue:44448", "issue:44462", "issue:44561", @@ -6547,6 +6524,7 @@ "issue:45003", "issue:45083", "issue:45127", + "issue:45198", "issue:45245", "issue:45290", "issue:45310", @@ -6556,9 +6534,10 @@ "issue:45584" ], "soft_pairs": [ + "issue:44442|issue:45198", "issue:43278|issue:43856", - "issue:38175|issue:45412", "issue:44360|issue:44561", + "issue:38175|issue:45412", "issue:43295|issue:43994", "issue:44206|issue:45584", "issue:38175|issue:43493", @@ -6566,32 +6545,31 @@ "issue:44849|issue:45003", "issue:42915|issue:43824", "issue:43653|issue:44112", - "issue:44568|issue:44991", "issue:43450|issue:43493", - "issue:45127|issue:45245", + "issue:44568|issue:44991", "issue:45290|issue:45381", + "issue:45127|issue:45245", "issue:43819|issue:45083", "issue:43450|issue:43575", "issue:44462|issue:44561", "issue:44561|issue:44779", "issue:44610|issue:44857", - "issue:43819|issue:44112", "issue:43262|issue:43450", + "issue:43819|issue:44112", "issue:44448|issue:44743", "issue:43606|issue:45083", "issue:43278|issue:43825", "issue:29942|issue:32090", - "issue:45310|issue:45478", "issue:37428|issue:43425", - "issue:43611|issue:43931", - "issue:43335|issue:43550" + "issue:45310|issue:45478", + "issue:43611|issue:43931" ], - "prepared_review_unit_hash": "9095d74277dbca8a22257fac71ee11259dfa04a9ffaf33065c04f18278d1500d", + "prepared_review_unit_hash": "a42d36626f56171845c0af13dba9ccba7ec9265319ea73dabb85ca7ff1916be2", "component_budget": { "node_count": 367, "item_count": 367, "soft_pair_count": 1370, - "serialized_chars": 706987, + "serialized_chars": 706986, "estimated_input_tokens": 176747, "estimated_eval_tokens": 353750 }, @@ -6599,201 +6577,201 @@ "node_count": 48, "item_count": 48, "soft_pair_count": 29, - "serialized_chars": 36342, - "estimated_input_tokens": 9086, - "estimated_eval_tokens": 18428 + "serialized_chars": 36310, + "estimated_input_tokens": 9078, + "estimated_eval_tokens": 18412 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Most pairs are clearly unrelated despite superficial keyword overlap. The only strong duplicate pair is the Qwen3.5 MoE from_pretrained regression, where the later report appears to be the same underlying bug as the earlier one.", - "confidence": 0.89, - "canonical_issue_reason": "issue:45310 is the clearest canonical issue: it states the concrete Qwen3.5 MoE from_pretrained failure, is more detailed, and has a near-exact follow-up duplicate in issue:45478.", + "summary": "This cluster is mostly a grab-bag of unrelated transformer/model issues. The only clear exact duplicate pair is the Qwen3.5 MoE from_pretrained regression, and the SigLIP2 pair is the only other plausible same-bug match.", + "confidence": 0.71, + "canonical_issue_reason": "issue:45310 is the best canonical issue: it is the earliest report of the exact Qwen3.5 MoE from_pretrained regression on transformers>=5.4.0, and issue:45478 reads like a later duplicate/follow-up.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45310 is the best global issue candidate because it is specific, reproducible, and directly matched by a duplicate report.", + "best_issue_reason": "issue:45310 is the strongest global issue anchor because it names a concrete version-triggered regression, has the same failure mode as its duplicate, and is the most representative dedupe target in this set.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:43278", - "right": "issue:43856", + "left": "issue:44442", + "right": "issue:45198", "accept": false, - "reason": "Different bugs: dtype change during evaluation vs MoE training memory usage." + "reason": "Both are tokenizer/processor bugs, but they affect different models and different failure modes; not the same underlying defect." }, { - "left": "issue:38175", - "right": "issue:45412", + "left": "issue:43278", + "right": "issue:43856", "accept": false, - "reason": "Different models and failure modes: SigLIP2 zero probabilities vs RT-DETR memory not released." + "reason": "Training dtype mismatch and MoE memory inefficiency are different bugs with no shared concrete code-path." }, { "left": "issue:44360", "right": "issue:44561", "accept": false, - "reason": "Unrelated: a DSA indexer activation issue vs a v5 compatibility break in trust_remote_code models." + "reason": "DSA indexer/ReLU discussion is unrelated to the v5 trust_remote_code import breakage." + }, + { + "left": "issue:38175", + "right": "issue:45412", + "accept": false, + "reason": "SigLIP2 zero-probability output and RT-DETR memory leak are unrelated model behaviors." }, { "left": "issue:43295", "right": "issue:43994", "accept": false, - "reason": "Both involve processor/pipeline behavior, but the models and symptoms differ too much to be the same bug." + "reason": "Both mention processor/model usage, but they are different regressions on different models and APIs." }, { "left": "issue:44206", "right": "issue:45584", "accept": false, - "reason": "Unrelated audio processor center-arg crash vs Whisper empty-transcription generation failure." + "reason": "Audio feature-extractor crash on unsupported args is not the same bug as Whisper empty-transcription generation failure." }, { "left": "issue:38175", "right": "issue:43493", - "accept": false, - "reason": "Same model family, but one report is specifically zero probabilities while the other is a broad HF-vs-JAX discrepancy; not enough to merge." + "accept": true, + "reason": "Both are about SigLIP2 producing wrong outputs versus expected behavior/reference, so they look like the same underlying model implementation bug." }, { "left": "issue:43827", "right": "issue:43901", "accept": false, - "reason": "Both are docs/API-deprecation complaints, but they target different APIs and behaviors." + "reason": "Both are docs-related, but they refer to different deprecated/changed pipeline behaviors and are not the same concrete issue." }, { "left": "issue:44849", "right": "issue:45003", "accept": false, - "reason": "Different areas: Qwen3.5 hidden-state bug vs unsafe sys.modules access in modeling_utils." + "reason": "Qwen3.5 hidden-states behavior and unsafe sys.modules access are different code paths and symptoms." }, { "left": "issue:42915", "right": "issue:43824", "accept": false, - "reason": "Unrelated: Qwen3Moe FP8 failure vs Qwen2.5-VL import error." + "reason": "Qwen3 MoE FP8 training failure and Qwen2.5-VL import error are unrelated." }, { "left": "issue:43653", "right": "issue:44112", "accept": false, - "reason": "Different issues: tokenizer special-token registration vs a stale CI device-override test." - }, - { - "left": "issue:44568", - "right": "issue:44991", - "accept": false, - "reason": "Both are tokenizer regressions in v5, but they affect different models and distinct code paths." + "reason": "Tokenizer special-token registration and GraniteSpeech CI device-override failure are different bugs." }, { "left": "issue:43450", "right": "issue:43493", "accept": false, - "reason": "Different subsystems: batched video processor shapes vs SigLIP2 model discrepancies." + "reason": "Video batch-shape bug is unrelated to the SigLIP2 output discrepancy." }, { - "left": "issue:45127", - "right": "issue:45245", + "left": "issue:44568", + "right": "issue:44991", "accept": false, - "reason": "Completely unrelated: LoRA merge collapse vs category-count runtime limit." + "reason": "Different tokenizer regressions affecting different models; no evidence they share the same root cause." }, { "left": "issue:45290", "right": "issue:45381", "accept": false, - "reason": "Different bugs in chat templating vs Qwen2.5-VL video position IDs." + "reason": "Chat-template tool-call crash and Qwen2.5-VL video position-id bug are distinct failures." + }, + { + "left": "issue:45127", + "right": "issue:45245", + "accept": false, + "reason": "LoRA/tied-embeddings collapse and category-count overflow are unrelated." }, { "left": "issue:43819", "right": "issue:45083", "accept": false, - "reason": "Different model internals: DAC latent conversion mismatch vs qwen3_omni_moe feature-length helper behavior." + "reason": "DAC latent/forward mismatch and Qwen3 Omni feature-length helper behavior are different defects." }, { "left": "issue:43450", "right": "issue:43575", "accept": false, - "reason": "Unrelated: batched video shape bug vs TP/OOM on Qwen2-57B-A14B-Instruct." + "reason": "Video processor shape handling and tp OOM on Qwen2-57B are unrelated issues." }, { "left": "issue:44462", "right": "issue:44561", "accept": false, - "reason": "Different problems: tokenizer.json loading vs removal of is_torch_fx_available breaking remote-code models." + "reason": "Tokenizer.json being ignored is a tokenizer-loading bug, while is_torch_fx_available removal affects trust_remote_code imports; different root causes." }, { "left": "issue:44561", "right": "issue:44779", "accept": false, - "reason": "Both are v5-related tokenizer/model regressions, but the concrete failures are different." + "reason": "Both involve v5 regressions, but one is an API removal/import break and the other is tokenizer output correctness; not the same bug." }, { "left": "issue:44610", "right": "issue:44857", "accept": false, - "reason": "Different models and failure modes: processor output size mismatch vs AMP/CUDA loss crash." + "reason": "Processor input-size mismatch and AMP/CUDA loss crash are different model/training paths." }, { - "left": "issue:43819", - "right": "issue:44112", + "left": "issue:43262", + "right": "issue:43450", "accept": false, - "reason": "No shared underlying bug; DAC latent mismatch is unrelated to GraniteSpeech CI device override." + "reason": "Audio chat-template sample-rate defaulting and batched video processor shape handling are unrelated modality bugs." }, { - "left": "issue:43262", - "right": "issue:43450", + "left": "issue:43819", + "right": "issue:44112", "accept": false, - "reason": "Different media types and bugs: audio chat-template sampling rate vs video batching shape." + "reason": "DAC latent mismatch and GraniteSpeech CI device override failure are not the same underlying issue." }, { "left": "issue:44448", "right": "issue:44743", "accept": false, - "reason": "Unrelated: Pegasus v4/v5 output differences vs Qwen3.5 recurrent-state reset bug." + "reason": "Pegasus v4/v5 output difference and Qwen3.5 recurrent-state reset bug are different model regressions." }, { "left": "issue:43606", "right": "issue:45083", "accept": false, - "reason": "Device mismatch on CPU offload is unrelated to qwen3_omni_moe output-length helper behavior." + "reason": "CPU offload device mismatch and feature-length helper behavior do not describe the same failure." }, { "left": "issue:43278", "right": "issue:43825", "accept": false, - "reason": "Different issues: dtype regression in evaluation vs pipeline error-message/docs behavior." + "reason": "Training/eval dtype change is unrelated to a pipeline error-message regression." }, { "left": "issue:29942", "right": "issue:32090", "accept": false, - "reason": "Unrelated: Flash Attention 2 test failures vs Trainer _gpu_broadcast_one NoneType error." - }, - { - "left": "issue:45310", - "right": "issue:45478", - "accept": true, - "reason": "Same concrete regression: both report Qwen3.5 MoE from_pretrained failures under transformers>=5.4.0." + "reason": "Flash Attention 2 test failures and _gpu_broadcast_one NoneType errors are separate CI/runtime problems." }, { "left": "issue:37428", "right": "issue:43425", "accept": false, - "reason": "Different compatibility problems: missing flash-support import vs Torch 2.10 incompatibility." + "reason": "Flash-attention import failure and Torch 2.10 compatibility are too broad and not clearly the same concrete bug." }, { - "left": "issue:43611", - "right": "issue:43931", - "accept": false, - "reason": "Unrelated loading failures: base_model_prefix handling vs Qwen3-VL weight-shape mismatch." + "left": "issue:45310", + "right": "issue:45478", + "accept": true, + "reason": "These are the same Qwen3.5 MoE from_pretrained regression on transformers>=5.4.0; 45478 is a near-verbatim duplicate of 45310." }, { - "left": "issue:43335", - "right": "issue:43550", + "left": "issue:43611", + "right": "issue:43931", "accept": false, - "reason": "Different bugs: SwitchTransformers sparse-layer creation vs Bamba torch.compile/SDPA failure." + "reason": "Base_model_prefix loading breakage and Qwen3-VL weight-shape mismatch are different model-loading errors." } ] }, "evaluator_result": { "accept": true, - "feedback": "Grounded and conservative overall. The canonical issue choice is supported by the packet, and the only accepted pair (45310/45478) is an exact-title duplicate with matching failure context. The remaining soft-edge verdicts stay appropriately cautious given the limited evidence." + "feedback": "Grounded overall. The Qwen3.5 MoE pair is clearly the same issue by title and chronology, and the rejected pairs stay conservative without overclaiming shared root cause. The summary appropriately keeps the cluster characterization modest and avoids broad dedupe claims beyond the evidence." }, "evaluator_used": true, "retried": false, @@ -6823,7 +6801,6 @@ "issue:38175", "issue:39692", "issue:40444", - "issue:41720", "issue:43066", "issue:43278", "issue:43329", @@ -6852,6 +6829,7 @@ "issue:44855", "issue:44857", "issue:44977", + "issue:44987", "issue:45072", "issue:45083", "issue:45200", @@ -6864,8 +6842,9 @@ "issue:45563" ], "soft_pairs": [ - "issue:39692|issue:43526", + "issue:43335|issue:43550", "issue:43819|issue:44493", + "issue:39692|issue:43526", "issue:37428|issue:44351", "issue:43450|issue:43526", "issue:45200|issue:45381", @@ -6875,277 +6854,289 @@ "issue:43335|issue:43653", "issue:29127|issue:32090", "issue:44448|issue:44779", - "issue:44977|issue:45229", "issue:38175|issue:44493", "issue:44292|issue:45520", + "issue:44977|issue:45229", "issue:43335|issue:45083", "issue:43335|issue:43504", "issue:43550|issue:44112", "issue:43335|issue:45072", "issue:44568|issue:45491", "issue:43720|issue:44112", - "issue:40444|issue:43526", "issue:44448|issue:44977", - "issue:43723|issue:44554", + "issue:40444|issue:43526", "issue:43335|issue:44112", + "issue:43723|issue:44554", "issue:34689|issue:35141", "issue:44857|issue:45072", "issue:43606|issue:44857", "issue:43066|issue:44462", "issue:45372|issue:45406", + "issue:43873|issue:44493", "issue:44977|issue:45563", "issue:43873|issue:45229", - "issue:43873|issue:44493", "issue:43278|issue:43329", "issue:43701|issue:45083", "issue:29942|issue:35141", "issue:43550|issue:45083", - "issue:41720|issue:45229" + "issue:37428|issue:44987" ], - "prepared_review_unit_hash": "7dd17096215e804082f0387683d06a56c7260aea66aedca8e9cbf1a49b8b7024", + "prepared_review_unit_hash": "75c19cf1a613041fa746f7802f80b2e451fb3f21caad9d67b908d69d09ff0bd8", "component_budget": { "node_count": 367, "item_count": 367, "soft_pair_count": 1370, - "serialized_chars": 706987, + "serialized_chars": 706986, "estimated_input_tokens": 176747, "estimated_eval_tokens": 353750 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 37, - "serialized_chars": 39030, - "estimated_input_tokens": 9758, - "estimated_eval_tokens": 19772 + "soft_pair_count": 38, + "serialized_chars": 39326, + "estimated_input_tokens": 9832, + "estimated_eval_tokens": 19920 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This is a noisy mixed cluster of unrelated Transformers issues; the soft pairs are mostly false positives across different models, code paths, and failure modes. No soft edge looks like the same underlying bug/change, so I would not merge any of these artifacts as duplicates.", - "confidence": 0.95, - "canonical_issue_reason": "If one issue must be chosen as the representative, 43066 is the cleanest broadly phrased v5 tokenizer regression. That said, the set is too heterogeneous to treat as a true duplicate cluster.", + "summary": "All soft-similarity pairs look like false positives: they span different models, different failure modes, and different code paths, so there is no true duplicate cluster here.", + "confidence": 0.97, + "canonical_issue_reason": "No canonical issue: the items are heterogeneous (loading/import errors, tokenizer behavior, vision/video processing, training/dispatch, CI-only failures) and do not point to one underlying bug.", "canonical_pr_reason": null, - "best_issue_reason": "43066 is the best single issue target because it is concise, broadly applicable, and centered on a core v5 tokenizer compatibility problem rather than a narrow model-specific symptom.", + "best_issue_reason": "No global best issue to anchor duplicates; none of the issues is a clear representative of the others because they are not the same defect.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:39692", - "right": "issue:43526", + "left": "issue:43335", + "right": "issue:43550", "accept": false, - "reason": "Docs/example errors in SigLIP2 are unrelated to BEiT fast processor label reduction." + "reason": "Sparse-layer config creation and torch.compile/SDPA model failures are different bugs in different code paths." }, { "left": "issue:43819", "right": "issue:44493", "accept": false, - "reason": "DAC from_latents/STE mismatch is a different bug from the position-id key regression." + "reason": "DAC latent/forward mismatch is unrelated to unexpected position-id key warnings." + }, + { + "left": "issue:39692", + "right": "issue:43526", + "accept": false, + "reason": "Doc example/model-processor mismatch and quantization problems are not the same as BeitImageProcessorFast reduce_labels behavior." }, { "left": "issue:37428", "right": "issue:44351", "accept": false, - "reason": "Different import failures: missing _flash_supports_window_size vs missing HybridCache." + "reason": "Missing flash-attention helper import and missing HybridCache import are separate import/API regressions." }, { "left": "issue:43450", "right": "issue:43526", "accept": false, - "reason": "Batched video shape handling is unrelated to BeitImageProcessorFast reduce_labels behavior." + "reason": "Batched video processor shape handling is unrelated to reduce_labels returning one label in BEiT fast processing." }, { "left": "issue:45200", "right": "issue:45381", "accept": false, - "reason": "Gemma4 mm_token_type_ids defaults and Qwen2.5-VL video position ids are different multimodal bugs." + "reason": "Gemma4 mm_token_type_ids defaults and Qwen2.5-VL video vision_position_ids are different multimodal bugs." }, { "left": "issue:44855", "right": "issue:45083", "accept": false, - "reason": "Python 3.13 indentation parsing failure is unrelated to qwen3_omni_moe helper logic." + "reason": "Python 3.13 torch.jit parsing/import error is unrelated to qwen3_omni_moe feature-length helper behavior." }, { "left": "issue:43404", "right": "issue:45200", "accept": false, - "reason": "Untied lm_head weights in Mistral3 is a separate issue from Gemma4 token-type defaults." + "reason": "lm_head weight tying in Mistral3 is a different issue from Gemma 4 token-type defaults." }, { "left": "issue:39692", "right": "issue:44493", "accept": false, - "reason": "SigLIP2 doc/example failures are not the same as the unexpected key/position-id regression." + "reason": "SigLIP2 example/quantization issues do not match the unexpected position-id-key warning regression." }, { "left": "issue:43335", "right": "issue:43653", "accept": false, - "reason": "SwitchTransformers sparse-layer config and BigBirdTokenizer special-token registration are unrelated." + "reason": "Sparse-layer configuration creation and tokenizer special-token registration are unrelated defects." }, { "left": "issue:29127", "right": "issue:32090", "accept": false, - "reason": "LayoutLMV3 error-message clarity is not the same bug as the Trainer NoneType broadcast error." + "reason": "LayoutLMv3 error-message clarity and Trainer NoneType broadcast failures are not the same bug." }, { "left": "issue:44448", "right": "issue:44779", "accept": false, - "reason": "These are separate tokenizer/generation regressions in different models with different failure modes." - }, - { - "left": "issue:44977", - "right": "issue:45229", - "accept": false, - "reason": "Flash-attention generation failure and multi-GPU OOM are different problems." + "reason": "Pegasus v4/v5 output differences and Deepseek tokenizer regressions affect different models and code paths." }, { "left": "issue:38175", "right": "issue:44493", "accept": false, - "reason": "SigLIP2 zero-probabilities and position-id key regressions are unrelated." + "reason": "SigLIP2 zero-probability outputs are not the same as unexpected position-id key warnings." }, { "left": "issue:44292", "right": "issue:45520", "accept": false, - "reason": "NVFP4 runtime failure is distinct from the Python 3.13 flash_attn import KeyError." + "reason": "Qwen-3-8B-NVFP4 runtime failure and flash_attn import KeyError are separate issues." + }, + { + "left": "issue:44977", + "right": "issue:45229", + "accept": false, + "reason": "Qwen3.5 flash-attention generation problems are unrelated to Gemma4 multi-GPU CUDA OOM." }, { "left": "issue:43335", "right": "issue:45083", "accept": false, - "reason": "SwitchTransformers sparse-layer construction and qwen3_omni_moe length helper issues are unrelated." + "reason": "Sparse-layer config bug and qwen3_omni_moe length-helper behavior do not share the same failure mode." }, { "left": "issue:43335", "right": "issue:43504", "accept": false, - "reason": "SwitchTransformers config bug is not the same as BEiT legacy-field loading failure." + "reason": "Sparse-layer creation and BEiT legacy-field loading failure are unrelated." }, { "left": "issue:43550", "right": "issue:44112", "accept": false, - "reason": "Bamba SDPA/torch.compile failure is unrelated to the GraniteSpeech stale device-override CI test." + "reason": "torch.compile/SDPA model failure and a stale CI device-override test are not the same defect." }, { "left": "issue:43335", "right": "issue:45072", "accept": false, - "reason": "Sparse-layer config creation and dtype-mismatch inference failures are different code paths." + "reason": "Sparse-layer config creation is unrelated to dtype mismatches in bfloat16 inference." }, { "left": "issue:44568", "right": "issue:45491", "accept": false, - "reason": "Tokenizer add_special_tokens/BOS-EOS regression is unrelated to Gemma3 NaN embeddings." + "reason": "Tokenizer add_special_tokens behavior and Gemma3 NaN embeddings arise from different subsystems." }, { "left": "issue:43720", "right": "issue:44112", "accept": false, - "reason": "BitNet packed-weight loading is a different bug from the GraniteSpeech CI test flake." + "reason": "Packed-weight unpacking during accelerate loading is unrelated to a stale CI device override." + }, + { + "left": "issue:44448", + "right": "issue:44977", + "accept": false, + "reason": "Pegasus output drift and Qwen3.5 flash-attention generation bugs are not the same underlying problem." }, { "left": "issue:40444", "right": "issue:43526", "accept": false, - "reason": "Qwen2.5-VL multi-image dataset handling is unrelated to BeitImageProcessorFast label reduction." + "reason": "IterableDataset multi-image finetuning failure is unrelated to BEiT reduce_labels output." }, { - "left": "issue:44448", - "right": "issue:44977", + "left": "issue:43335", + "right": "issue:44112", "accept": false, - "reason": "Both involve output differences, but they are different models and distinct underlying causes." + "reason": "Config-induced sparse-layer creation is not the same as a CI-only stale device override failure." }, { "left": "issue:43723", "right": "issue:44554", "accept": false, - "reason": "Tokenizer loading in v5 is unrelated to the MPS attention correctness issue." + "reason": "AutoTokenizer loading regression and an MPS attention correctness issue are different problems." }, { "left": "issue:34689", "right": "issue:35141", "accept": false, - "reason": "Model loading breakage and output-embedding reinitialization are separate problems." + "reason": "Llama 3.2 model loading break and post_init reinitializing embeddings are different bugs." }, { "left": "issue:44857", "right": "issue:45072", "accept": false, - "reason": "AMP crash in LwDetr and dtype-mismatch inference issues are not the same bug." + "reason": "AMP/CUDA crash in LwDetrImageLoss is unrelated to bfloat16 dtype mismatches in inference." }, { "left": "issue:43606", "right": "issue:44857", "accept": false, - "reason": "CPU offload device mismatch is distinct from the float16 AMP crash." + "reason": "CPU offload device mismatch and AMP/CUDA loss crash are different runtime failures." }, { "left": "issue:43066", "right": "issue:44462", "accept": false, - "reason": "Both are tokenizer-v5 regressions, but one is wrong decoder type and the other is tokenizer.json loading; not the same bug." + "reason": "Wrong tokenizer decoder type and ignoring tokenizer.json are separate tokenizer-loading regressions." }, { "left": "issue:45372", "right": "issue:45406", "accept": false, - "reason": "Gemma4 processor import failure and serve-time missing _tokenizer failure are different issues." + "reason": "Gemma4 processor import dependency failure and missing _tokenizer attribute in serve are not the same issue." }, { - "left": "issue:44977", - "right": "issue:45563", + "left": "issue:43873", + "right": "issue:44493", "accept": false, - "reason": "Flash-attention generation bug is unrelated to the stale num_return_sequences warning." + "reason": "Quantization/offloading behavior is unrelated to unexpected position-id key warnings." }, { - "left": "issue:43873", - "right": "issue:45229", + "left": "issue:44977", + "right": "issue:45563", "accept": false, - "reason": "Quantization/offloading behavior is unrelated to Gemma4 multi-GPU OOM." + "reason": "Flash-attention generation failure and a stale num_return_sequences warning are different concerns." }, { "left": "issue:43873", - "right": "issue:44493", + "right": "issue:45229", "accept": false, - "reason": "Offloading/quantization issues do not match the unexpected key/position-id regression." + "reason": "Quantized offloading issues and Gemma4 multi-GPU OOM are distinct failure modes." }, { "left": "issue:43278", "right": "issue:43329", "accept": false, - "reason": "Training/eval dtype drift is unrelated to the multimodal token counting video-branch bug." + "reason": "Embedding dtype drift during eval and undefined helper names in multimodal token counting are unrelated." }, { "left": "issue:43701", "right": "issue:45083", "accept": false, - "reason": "resume_from_checkpoint key mismatch is not the same as qwen3_omni_moe helper behavior." + "reason": "resume_from_checkpoint key mismatch is unrelated to qwen3_omni_moe helper behavior." }, { "left": "issue:29942", "right": "issue:35141", "accept": false, - "reason": "Flash Attention 2 test failures are unrelated to token-embedding reinitialization in post_init." + "reason": "Flash Attention 2 test failures and embedding reinitialization in post_init do not describe the same bug." }, { "left": "issue:43550", "right": "issue:45083", "accept": false, - "reason": "Bamba torch.compile SDPA issues are unrelated to qwen3_omni_moe feature-length helper behavior." + "reason": "Bamba torch.compile/SDPA failure is unrelated to qwen3_omni_moe feature-length helper issues." }, { - "left": "issue:41720", - "right": "issue:45229", + "left": "issue:37428", + "right": "issue:44987", "accept": false, - "reason": "CUDA assert from auto device mapping is a different failure from multi-GPU OOM." + "reason": "Missing flash-support import and loading physical-intelligence/fast on transformers>=5.1.0 are different regressions." } ] }, @@ -7177,8 +7168,8 @@ "issue:33290", "issue:33357", "issue:34567", - "issue:37428", "issue:39692", + "issue:41720", "issue:42915", "issue:42994", "issue:43065", @@ -7204,7 +7195,6 @@ "issue:44857", "issue:44871", "issue:44977", - "issue:44987", "issue:45072", "issue:45103", "issue:45198", @@ -7219,24 +7209,24 @@ "issue:45563" ], "soft_pairs": [ + "issue:41720|issue:45229", "issue:45245|issue:45563", - "issue:37428|issue:44987", - "issue:43976|issue:45440", "issue:32090|issue:33357", + "issue:43976|issue:45440", "issue:45372|issue:45459", - "issue:44448|issue:44466", "issue:44451|issue:45520", - "issue:43065|issue:43994", - "issue:43065|issue:45412", + "issue:44448|issue:44466", "issue:25251|issue:32090", + "issue:43065|issue:45412", + "issue:43065|issue:43994", "issue:43278|issue:43381", "issue:43653|issue:44743", "issue:44186|issue:45198", - "issue:44265|issue:45198", "issue:43257|issue:43278", + "issue:44265|issue:45198", "issue:44361|issue:45198", - "issue:43994|issue:45412", "issue:43421|issue:45491", + "issue:43994|issue:45412", "issue:42915|issue:42994", "issue:44871|issue:44977", "issue:44206|issue:44479", @@ -7250,214 +7240,214 @@ "issue:29127|issue:34567", "issue:45072|issue:45229" ], - "prepared_review_unit_hash": "4a8628685548a2bc87c9f044b37faf46e4c446d968ffcccc50132699d147dbe3", + "prepared_review_unit_hash": "860a73e3913dbe1609cdafa958411863c33c7bbdabced4ba906e6a6ecc51f374", "component_budget": { "node_count": 367, "item_count": 367, "soft_pair_count": 1370, - "serialized_chars": 706987, + "serialized_chars": 706986, "estimated_input_tokens": 176747, "estimated_eval_tokens": 353750 }, "budget": { - "node_count": 48, - "item_count": 48, + "node_count": 47, + "item_count": 47, "soft_pair_count": 30, - "serialized_chars": 36725, - "estimated_input_tokens": 9182, - "estimated_eval_tokens": 18620 + "serialized_chars": 36127, + "estimated_input_tokens": 9032, + "estimated_eval_tokens": 18320 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is not a duplicate set: the soft pairs are only loosely similar and describe different bugs, features, and failure modes across unrelated subsystems, so all candidate edges should be rejected.", - "confidence": 0.97, - "canonical_issue_reason": null, - "canonical_pr_reason": null, - "best_issue_reason": null, - "best_pr_reason": null, + "summary": "These are mostly unrelated single-issue reports across different models and subsystems. The soft similarity links are all false positives; none of the pairs look like the same underlying bug or change.", + "confidence": 0.95, + "canonical_issue_reason": "No single canonical issue: the items are mostly unrelated bugs rather than one duplicate family.", + "canonical_pr_reason": "No PRs are present in this cluster.", + "best_issue_reason": "No strong global best issue either, since there is no coherent duplicate cluster to anchor on.", + "best_pr_reason": "No PRs are present in this cluster.", "soft_edge_verdicts": [ + { + "left": "issue:41720", + "right": "issue:45229", + "accept": false, + "reason": "Both are inference/runtime problems, but one is Qwen3 auto device-mapping asserts and the other is Gemma4 multi-GPU OOM; different models and failure modes." + }, { "left": "issue:45245", "right": "issue:45563", "accept": false, - "reason": "One is a category-cardinality runtime error; the other is a stale warning in paged generate(). Different code paths and symptoms." + "reason": "Unrelated problems: category-count limit runtime error vs a stale warning about num_return_sequences in paged generate()." }, { - "left": "issue:37428", - "right": "issue:44987", + "left": "issue:32090", + "right": "issue:33357", "accept": false, - "reason": "Both involve loading/imports, but one is a flash-attention helper import error and the other is a model loading failure for a specific repo. Not the same bug." + "reason": "Trainer broadcast NoneType error and MacOS bus error in CLIP are distinct bugs with no shared code path." }, { "left": "issue:43976", "right": "issue:45440", "accept": false, - "reason": "Python version compatibility regression vs model-implementation divergence. No shared concrete failure." - }, - { - "left": "issue:32090", - "right": "issue:33357", - "accept": false, - "reason": "Trainer broadcast TypeError and MacOS CLIP bus error are unrelated failures in different runtime paths." + "reason": "Python version compatibility vs DeepSeekV3MoE divergence; these are not the same defect." }, { "left": "issue:45372", "right": "issue:45459", "accept": false, - "reason": "Gemma processor import failure from a missing symbol vs tokenizer error masking when protobuf is absent. Different problems." + "reason": "One is a missing ReasoningEffort import breaking Gemma 4 loading; the other is tokenizer error masking when protobuf is absent." }, { - "left": "issue:44448", - "right": "issue:44466", + "left": "issue:44451", + "right": "issue:45520", "accept": false, - "reason": "Different-output generation regression vs state serialization of tied weights. Not the same underlying bug." + "reason": "Model-loading failure for ScandiBERT and a Python 3.13 flash_attn import KeyError are unrelated." }, { - "left": "issue:44451", - "right": "issue:45520", + "left": "issue:44448", + "right": "issue:44466", "accept": false, - "reason": "Model load failure for one repo vs flash_attn import lookup on Python 3.13. No concrete overlap." + "reason": "Both mention version-dependent model behavior, but one is Pegasus output regression and the other is lm_head serialization/tied-weights inconsistency." }, { - "left": "issue:43065", - "right": "issue:43994", + "left": "issue:25251", + "right": "issue:32090", "accept": false, - "reason": "Dummy Conv2d in Sam3PixelDecoder is unrelated to SigLIP2 bad outputs." + "reason": "Pipeline top_k nesting bug has nothing to do with Trainer GPU broadcast of NoneType." }, { "left": "issue:43065", "right": "issue:45412", "accept": false, - "reason": "A placeholder layer issue is unrelated to an RT-DETR memory leak." + "reason": "Dummy Conv2d in Sam3PixelDecoder and RT-DETR memory not released are different lifecycle/code-path issues." }, { - "left": "issue:25251", - "right": "issue:32090", + "left": "issue:43065", + "right": "issue:43994", "accept": false, - "reason": "Pipeline top_k nesting bug and trainer GPU broadcast TypeError are unrelated." + "reason": "Sam3PixelDecoder architecture placeholder vs SigLIP2 wrong outputs; different models and symptoms." }, { "left": "issue:43278", "right": "issue:43381", "accept": false, - "reason": "Training/eval dtype drift is not the same as gradient checkpointing being unsupported in eval mode." + "reason": "Embedding dtype drift during eval and gradient checkpointing in eval mode are separate evaluation-time issues." }, { "left": "issue:43653", "right": "issue:44743", "accept": false, - "reason": "Tokenizer special-token registration bug vs recurrent-state reset in a Qwen model. Different components." + "reason": "Tokenizer special-token registration bug is unrelated to Qwen3.5 recurrent state reset with cache." }, { "left": "issue:44186", "right": "issue:45198", "accept": false, - "reason": "LayoutLMv2 tokenizer crashes on NER padding/truncation, while the other issue is Wav2Vec2 save/tokenization failure." - }, - { - "left": "issue:44265", - "right": "issue:45198", - "accept": false, - "reason": "torch.export failure due to torch_compilable_check is unrelated to Wav2Vec2 save_pretrained/tokenization problems." + "reason": "LayoutLMv2Tokenizer NER/padding crash and Wav2Vec2 save/tokenization failure affect different tokenizers and paths." }, { "left": "issue:43257", "right": "issue:43278", "accept": false, - "reason": "MoE weight conversion/loading problem vs embedding dtype mismatch. Different layers and failure modes." + "reason": "Qwen3 MoE weight conversion with accelerate/deepspeed is unrelated to embedding dtype changes in eval." }, { - "left": "issue:44361", + "left": "issue:44265", "right": "issue:45198", "accept": false, - "reason": "MLuke tokenizer AttributeError is unrelated to Wav2Vec2 save and tokenization failure." + "reason": "torch.export failure due to torch_compilable_check is not the same as Wav2Vec2 save_pretrained/tokenization failure." }, { - "left": "issue:43994", - "right": "issue:45412", + "left": "issue:44361", + "right": "issue:45198", "accept": false, - "reason": "SigLIP2 wrong predictions and RT-DETR memory reclamation are unrelated." + "reason": "MLukeTokenizer AttributeError and Wav2Vec2 save/tokenization failure are separate tokenizer bugs." }, { "left": "issue:43421", "right": "issue:45491", "accept": false, - "reason": "Runtime special-token post-processor update bug vs Gemma3 NaN embeddings from sliding-window attention. Not the same issue." + "reason": "Runtime post-processor update bug and Gemma3 NaN embeddings from sliding-window attention are unrelated." + }, + { + "left": "issue:43994", + "right": "issue:45412", + "accept": false, + "reason": "SigLIP2 nonsensical outputs and RT-DETR memory leak are completely different problems." }, { "left": "issue:42915", "right": "issue:42994", "accept": false, - "reason": "Both mention quantization, but one is FP8 inference failure and the other is saving quantized models. Different code paths." + "reason": "FineGrainedFP8Config failure on Qwen3Moe and quantized model saving failure are both quantization-related but not the same bug." }, { "left": "issue:44871", "right": "issue:44977", "accept": false, - "reason": "Gemma eos_token_id config mismatch is unrelated to Qwen3.5 flash-attention generation failure." + "reason": "Gemma-3 eos_token_id mismatch and Qwen3.5 flash-attention generation failure do not share a concrete code path." }, { "left": "issue:44206", "right": "issue:44479", "accept": false, - "reason": "Feature extractor center-arg regression and video-input regression for Qwen models are distinct issues." + "reason": "Feature-extractor center-arg regression and Qwen VL video-input regression are different model/input pipelines." }, { "left": "issue:43756", "right": "issue:45440", "accept": false, - "reason": "Smollm3 RoPE-layer mismatch vs DeepSeekV3 implementation divergence are different model-specific bugs." + "reason": "Smollm3 RoPE-layer mismatch and DeepSeekV3MoE divergence are unrelated model-architecture complaints." }, { "left": "issue:43504", "right": "issue:44857", "accept": false, - "reason": "Legacy field blocks BEiT model loading; the other is an AMP/CUDA loss crash. Not the same problem." + "reason": "Beit pretrained-load legacy field bug and LwDetrImageLoss AMP CUDA crash are distinct failures." }, { "left": "issue:29942", "right": "issue:30064", "accept": false, - "reason": "Flash Attention 2 test failures and void segmentation-map processing are unrelated." + "reason": "Flash Attention 2 test failures and void segmentation-map processing failure are unrelated." }, { "left": "issue:44610", "right": "issue:45412", "accept": false, - "reason": "Processor output-size mismatch vs model memory leak; no shared bug." + "reason": "OmDet-Turbo processor image-size mismatch and RT-DETR memory leak are different bugs." }, { "left": "issue:33357", "right": "issue:39692", "accept": false, - "reason": "MacOS bus error on a CLIP model and SigLIP2 doc-example failures are different bugs, not duplicates." + "reason": "MacOS CLIP bus error and SigLIP2 documentation example errors are not the same underlying issue." }, { "left": "issue:45103", "right": "issue:45372", "accept": false, - "reason": "Auto-docstring annotation crash and Gemma processor import error affect different subsystems." + "reason": "auto_docstring future-annotations crash and Gemma 4 processor import failure are unrelated import/runtime errors." }, { "left": "issue:32090", "right": "issue:33290", "accept": false, - "reason": "Trainer GPU broadcast NoneType error and DeepSpeed Adafactor OOM are unrelated training issues." + "reason": "Trainer broadcast NoneType error and DeepSpeed Adafactor OOM are different training-time problems." }, { "left": "issue:29127", "right": "issue:34567", "accept": false, - "reason": "LayoutLMv3 error-message clarity issue and TrainerState token counter not updating are not the same bug." + "reason": "LayoutLMv3 error-message clarity and TrainerState token counter not updating are separate issues." }, { "left": "issue:45072", "right": "issue:45229", "accept": false, - "reason": "Bfloat16 dtype mismatch in inference and Gemma4 multi-GPU CUDA OOM are different failure classes." + "reason": "dtype mismatch in inference wrappers and Gemma4 multi-GPU OOM are not the same defect." } ] }, @@ -7538,8 +7528,8 @@ "issue:43526|issue:44186", "issue:44464|issue:45491", "issue:29942|issue:36010", - "issue:44977|issue:45245", "issue:29127|issue:33357", + "issue:44977|issue:45245", "issue:43262|issue:44206", "issue:29127|issue:35141", "issue:43761|issue:44079", @@ -7549,24 +7539,24 @@ "issue:43526|issue:44857", "issue:44479|issue:45290", "issue:43526|issue:44610", - "issue:43723|issue:45479", "issue:44743|issue:44811", + "issue:43723|issue:45479", "issue:45003|issue:45440", "issue:43575|issue:43701", "issue:43931|issue:45216", "issue:44977|issue:45584", "issue:43646|issue:44077", "issue:43761|issue:44188", - "issue:44964|issue:45440", "issue:43316|issue:43525", - "issue:29942|issue:33666" + "issue:29942|issue:33666", + "issue:44964|issue:45440" ], - "prepared_review_unit_hash": "2e151d5c2013cd7d8412fb91ec44854199de61ce2b72aaf010aa542107a68706", + "prepared_review_unit_hash": "3af25fde097b959b0962bc4954bd24324d853cf36594b0d6fda0bd593eec03e4", "component_budget": { "node_count": 367, "item_count": 367, "soft_pair_count": 1370, - "serialized_chars": 706987, + "serialized_chars": 706986, "estimated_input_tokens": 176747, "estimated_eval_tokens": 353750 }, @@ -7574,7 +7564,7 @@ "node_count": 48, "item_count": 48, "soft_pair_count": 29, - "serialized_chars": 36403, + "serialized_chars": 36402, "estimated_input_tokens": 9101, "estimated_eval_tokens": 18458 }, @@ -7583,186 +7573,186 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is highly heterogeneous: the soft pairs mostly share only broad transformer/modeling themes, not the same underlying bug. I would not merge any of the suggested soft edges.", - "confidence": 0.94, - "canonical_issue_reason": "There is no clear cluster-wide duplicate target. If an anchor is needed, issue 43761 is the strongest standalone report because it describes a specific v5 regression, has a clear expected/actual behavior, and is actionable.", + "summary": "This looks like a loose thematic cluster of Transformers bugs and regressions, not a set of true duplicates. The soft pairs are mostly same-subsystem overlaps (Qwen, tokenizers, generation, configs) but each points to a different concrete failure mode or code path.", + "confidence": 0.66, + "canonical_issue_reason": "issue:43824 has the strongest signal as a representative user-facing regression: high discussion, multiple inbound references, and a concrete import failure with clear reproduction impact. It is the best single issue to anchor the cluster, even though the cluster itself does not collapse into one duplicate set.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 43761 is the best representative issue: concrete, reproducible, and narrowly scoped to one code path; it is a better triage anchor than the more generic or speculative reports.", + "best_issue_reason": "issue:43824 is the most externally validated and broadly relevant issue in the set, with the clearest evidence of impact and follow-up from other references. It is a good global representative for triage.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43824", "right": "issue:44279", "accept": false, - "reason": "ImportError for a missing Qwen2.5-VL class is not the same bug as a vague dependency complaint." + "reason": "Both are import/dependency-adjacent at a high level, but one is a missing model export and the other is a generic dependency issue; different underlying failures." }, { "left": "issue:43381", "right": "issue:43701", "accept": false, - "reason": "Gradient checkpointing in eval mode and resume-from-checkpoint key mismatch are unrelated failure modes." + "reason": "One is about gradient checkpointing in eval mode; the other is a resume-from-checkpoint key mismatch. Same training area, different bug." }, { "left": "issue:43653", "right": "issue:44442", "accept": false, - "reason": "BigBirdTokenizer special-token registration and AutoTokenizer loading FastSpeech2ConformerTokenizer are different tokenizer issues." + "reason": "Both involve tokenizers, but one is a special-token registration/decode bug and the other is an AutoTokenizer load failure for a different tokenizer." }, { "left": "issue:43334", "right": "issue:43531", "accept": false, - "reason": "Missing pad_token_id on Qwen3-VL config is unrelated to the sliding_window behavior in Qwen3-MoE." + "reason": "Both are Qwen-related, but one is a missing pad_token_id on Qwen3-VL config and the other is a sliding_window issue on Qwen3-MoE." }, { "left": "issue:43526", "right": "issue:44186", "accept": false, - "reason": "reduce_labels returning one label and LayoutLMv2Tokenizer NER padding crashes are different bugs." + "reason": "Both touch preprocessing, but one is BeitImageProcessorFast label reduction and the other is a LayoutLMv2Tokenizer crash on NER/batching." }, { "left": "issue:44464", "right": "issue:45491", "accept": false, - "reason": "Compiled-forward chunked generation divergence is not the same as Gemma3 sliding-window NaNs." + "reason": "Both are generation/attention regressions, but they involve different models and different failures: compiled-forward inconsistency vs NaNs from sliding-window attention." }, { "left": "issue:29942", "right": "issue:36010", "accept": false, - "reason": "Flash Attention 2 test failures do not match a GenerationMixin import error." + "reason": "Flash Attention test failures and GenerationMixin import errors are unrelated code paths." }, { - "left": "issue:44977", - "right": "issue:45245", + "left": "issue:29127", + "right": "issue:33357", "accept": false, - "reason": "Qwen3.5 flash-attention generation issues are unrelated to the 2^24 category limit error." + "reason": "Different models and failures: LayoutLMv3 error messaging vs a MacOS bus error in CLIP." }, { - "left": "issue:29127", - "right": "issue:33357", + "left": "issue:44977", + "right": "issue:45245", "accept": false, - "reason": "LayoutLMv3 box-content error messaging is unrelated to a MacOS bus error on CLIP." + "reason": "Qwen3.5 flash-attention generation issues are unrelated to the category-count runtime limit error." }, { "left": "issue:43262", "right": "issue:44206", "accept": false, - "reason": "Audio chat-template sample-rate defaults and an unsupported center arg in LasrFeatureExtractor are different problems." + "reason": "Both are audio preprocessing regressions, but they are different unsupported-argument failures in different components." }, { "left": "issue:29127", "right": "issue:35141", "accept": false, - "reason": "LayoutLMv3 box validation messaging has nothing to do with token embedding reinitialization after resize." + "reason": "One is an error-message clarity issue; the other is a post_init reinitialization bug after resizing token embeddings." }, { "left": "issue:43761", "right": "issue:44079", "accept": false, - "reason": "CLIPVisionModel missing hidden_states and ModelOutput key assignment semantics are distinct code paths." + "reason": "CLIPVisionModel hidden-states regression and ModelOutput key assignment are separate output-handling bugs." }, { "left": "issue:42915", "right": "issue:43931", "accept": false, - "reason": "Qwen3MoE FP8 failure and Qwen3-VL weight-shape mismatch are different model-loading failures." + "reason": "Both are Qwen model issues, but one is FP8 config failure and the other is a weight-shape mismatch during loading." }, { "left": "issue:44442", "right": "issue:45072", "accept": false, - "reason": "Tokenizer loading failure and bfloat16 dtype mismatch are not the same underlying defect." + "reason": "Tokenizer loading and dtype mismatch in inference are unrelated." }, { "left": "issue:43232", "right": "issue:43388", "accept": false, - "reason": "Generation kwargs after sync_gpus and metric-label truncation are unrelated distributed-training bugs." + "reason": "Generation kwargs synchronization and metric-label gathering are different distributed-training bugs." }, { "left": "issue:43526", "right": "issue:44857", "accept": false, - "reason": "Label reduction in BeitImageProcessorFast and float16 CUDA loss crashes are different issues." + "reason": "Label reduction in an image processor and a mixed-precision CUDA crash are not the same bug." }, { "left": "issue:44479", "right": "issue:45290", "accept": false, - "reason": "Video-input regression across Qwen VL variants is unrelated to chat template crashes on empty tool-call messages." + "reason": "Video-input regression for Qwen models and chat-template crashes on empty tool-call messages are different issues." }, { "left": "issue:43526", "right": "issue:44610", "accept": false, - "reason": "Beit label reduction and OmDet-Turbo processor/model size mismatch are unrelated." + "reason": "Both are processor-related, but one is a label-reduction bug and the other is an input-size mismatch." }, { - "left": "issue:43723", - "right": "issue:45479", + "left": "issue:44743", + "right": "issue:44811", "accept": false, - "reason": "AutoTokenizer loading regression in v5 is unrelated to zero-loss single-label classification behavior." + "reason": "Recurrent-state reset with cache and Whisper batch_decode skip_special_tokens are unrelated." }, { - "left": "issue:44743", - "right": "issue:44811", + "left": "issue:43723", + "right": "issue:45479", "accept": false, - "reason": "Qwen3.5 recurrent-state reset and Whisper batch_decode skip_special_tokens handling are different components." + "reason": "Tokenizer loading in v5 and degenerate zero-loss sequence classification are different failures." }, { "left": "issue:45003", "right": "issue:45440", "accept": false, - "reason": "Unsafe sys.modules access in modeling_utils is not the same as DeepSeekV3MoE divergence from remote implementation." + "reason": "Unsafe sys.modules access and DeepSeekV3MoE divergence are unrelated code paths." }, { "left": "issue:43575", "right": "issue:43701", "accept": false, - "reason": "Tensor-parallel OOM when loading Qwen2-57B is unrelated to resume-from-checkpoint key mismatch." + "reason": "Tensor-parallel OOM on Qwen2-57B and checkpoint key mismatch are distinct problems." }, { "left": "issue:43931", "right": "issue:45216", "accept": false, - "reason": "Model weight-shape mismatch on load is not the same as an incorrect saved checkpoint from save_pretrained." + "reason": "Both involve Qwen3.5/VL loading or saving, but one is a load-time shape mismatch and the other is an incorrect saved checkpoint regression." }, { "left": "issue:44977", "right": "issue:45584", "accept": false, - "reason": "Qwen3.5 flash-attention generation problems are unrelated to Whisper empty-transcription generation failures." + "reason": "Different models and symptoms: Qwen3.5 flash-attn generation vs Whisper empty-transcription handling." }, { "left": "issue:43646", "right": "issue:44077", "accept": false, - "reason": "Breaking custom model initialization in v5 is broader and different from patchtsmixer's optional post_init rule." + "reason": "Both concern initialization/post_init behavior, but the broad custom-model-init breakage and the PatchTSMixer-specific post_init policy are not the same concrete bug." }, { "left": "issue:43761", "right": "issue:44188", "accept": false, - "reason": "Hidden_states not returned and attention-kernel divergence under torch.compile are separate bugs." - }, - { - "left": "issue:44964", - "right": "issue:45440", - "accept": false, - "reason": "Phi-4-multimodal loading failure is unrelated to DeepSeekV3MoE parity issues." + "reason": "Hidden_states regression in CLIPVisionModel and attention-kernel divergence under torch.compile are unrelated." }, { "left": "issue:43316", "right": "issue:43525", "accept": false, - "reason": "Gemma3TextConfig API discrepancy and Llama4Config missing pad_token_id are only loosely similar, not the same bug." + "reason": "Both are config-API inconsistencies, but one is a Gemma3TextConfig parity issue and the other is a Llama4Config pad_token_id attribute error." }, { "left": "issue:29942", "right": "issue:33666", "accept": false, - "reason": "Flash Attention 2 test failures are unrelated to Qwen2-VL multi-GPU training." + "reason": "Failing Flash Attention tests and Qwen2-VL multi-GPU training are unrelated." + }, + { + "left": "issue:44964", + "right": "issue:45440", + "accept": false, + "reason": "Phi-4 multimodal loading failure and DeepSeekV3MoE divergence are different model-specific issues." } ] }, @@ -7838,19 +7828,19 @@ "issue:43575|issue:44387", "issue:43994|issue:44206", "issue:39692|issue:45491", - "issue:43673|issue:44743", "issue:33290|issue:34567", "issue:44442|issue:44898", + "issue:43673|issue:44743", "issue:34689|issue:43611", "issue:43381|issue:43873", "issue:43493|issue:43827", "issue:33290|issue:35141", "issue:33290|issue:34689", - "issue:44077|issue:44479", "issue:41628|issue:45372", + "issue:44077|issue:44479", "issue:43720|issue:44442", - "issue:43749|issue:43931", "issue:43643|issue:44913", + "issue:43749|issue:43931", "issue:43873|issue:44871", "issue:44190|issue:44496", "issue:43381|issue:43856", @@ -7864,12 +7854,12 @@ "issue:42175|issue:43316", "issue:43404|issue:43441" ], - "prepared_review_unit_hash": "78be58d9f035125e73b6f13c6620694b1e1841707f84ac84643b94a9f22c666f", + "prepared_review_unit_hash": "2888c4df9578ba1cb91aef0e6a2ff904c93690021c2fab7913c1d266999ba1eb", "component_budget": { "node_count": 367, "item_count": 367, "soft_pair_count": 1370, - "serialized_chars": 706987, + "serialized_chars": 706986, "estimated_input_tokens": 176747, "estimated_eval_tokens": 353750 }, @@ -7877,189 +7867,189 @@ "node_count": 47, "item_count": 47, "soft_pair_count": 28, - "serialized_chars": 35702, - "estimated_input_tokens": 8926, - "estimated_eval_tokens": 18108 + "serialized_chars": 35699, + "estimated_input_tokens": 8925, + "estimated_eval_tokens": 18106 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The cluster is mostly a set of unrelated issue reports that only share broad themes like loading errors, memory use, or config/API regressions. I found no soft-edge pair that looks like the same underlying bug/change, so all candidate merges are rejected.", - "confidence": 0.94, - "canonical_issue_reason": "issue:44387 is a high-signal, well-scoped regression report with strong discussion activity around a concrete OOM/memory behavior, making it the best representative issue in this set.", + "summary": "This cluster is heterogeneous: most items are distinct bug reports about different models, configs, or regressions, with only superficial similarity around loading/memory/cache issues. I do not see any strong duplicate groups; the soft edges should be rejected.", + "confidence": 0.33, + "canonical_issue_reason": "Issue 44387 is the strongest representative only in the sense that it is a high-signal, broadly framed regression report with substantial discussion and impact; however, it is not a true duplicate anchor for the rest of the cluster.", "canonical_pr_reason": null, - "best_issue_reason": "issue:44387 has the clearest, most actionable bug description and the strongest engagement, so it is the best standalone issue here.", + "best_issue_reason": "44387 is the best standalone issue because it is clearly scoped, has strong discussion activity, and describes a concrete regression rather than a narrow model-specific symptom.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43575", "right": "issue:44387", "accept": false, - "reason": "Both mention OOM, but one is tensor-parallel model loading and the other is int4 quantization memory growth; different code paths and root causes." + "reason": "Both mention OOM, but they describe different causes and contexts: TP loading for a specific Qwen model versus increased CUDA reserved memory under int4 quantization." }, { "left": "issue:43994", "right": "issue:44206", "accept": false, - "reason": "Unrelated regressions: SigLIP2 inference quality vs an audio feature extractor argument crash." + "reason": "These are unrelated failures: one is SigLIP2 producing bad outputs with AutoModel/pipeline, the other is a feature extractor argument regression." }, { "left": "issue:39692", "right": "issue:45491", "accept": false, - "reason": "Documentation/example errors for SigLIP2 do not match a Gemma3 GPU NaN embedding bug." - }, - { - "left": "issue:43673", - "right": "issue:44743", - "accept": false, - "reason": "Both involve cache behavior, but one is a missing GenerationMixin cache in v5.0.0 and the other is Qwen3.5 recurrent state reset in a specific modular implementation." + "reason": "Different models and symptoms: SigLIP2 docs example errors versus Gemma-3 NaN embeddings during mixed-length batching." }, { "left": "issue:33290", "right": "issue:34567", "accept": false, - "reason": "Completely different failures: DeepSpeed Adafactor OOM versus TrainerState token-count bookkeeping." + "reason": "OOM with Adafactor in DeepSpeed is unrelated to TrainerState's token counter not updating." }, { "left": "issue:44442", "right": "issue:44898", "accept": false, - "reason": "Different subsystems and symptoms: tokenizer loading failure versus Perceiver image-classification resolution handling." + "reason": "One is a tokenizer load failure for FastSpeech2ConformerTokenizer; the other is a Perceiver image-classification failure on non-default resolution." + }, + { + "left": "issue:43673", + "right": "issue:44743", + "accept": false, + "reason": "Both involve cache semantics, but they affect different code paths and models: v5 chunked prefill cache lookup versus Qwen3.5 recurrent state reset." }, { "left": "issue:34689", "right": "issue:43611", "accept": false, - "reason": "Both are model-loading regressions, but one is a specific Llama 3.2 vision breakage and the other is a base_model_prefix-related v5 loading issue." + "reason": "Both are loading-related, but one is a model-specific Llama 3.2 Vision issue and the other is a general base_model_prefix loading regression." }, { "left": "issue:43381", "right": "issue:43873", "accept": false, - "reason": "Gradient checkpointing in eval mode and quantization offloading are separate training/runtime behaviors with different failures." + "reason": "Gradient checkpointing in eval mode and quantization offloading are different mechanisms with different failure modes." }, { "left": "issue:43493", "right": "issue:43827", "accept": false, - "reason": "SigLIP2 implementation mismatch and docs still referencing pipeline() are unrelated." + "reason": "SigLIP2 implementation drift and outdated pipeline docs are unrelated." }, { "left": "issue:33290", "right": "issue:35141", "accept": false, - "reason": "OOM during optimizer use is not the same bug as output embeddings being reinitialized after resizing." + "reason": "Adafactor/DeepSpeed OOM is not the same bug as output embeddings being reinitialized after resizing token embeddings." }, { "left": "issue:33290", "right": "issue:34689", "accept": false, - "reason": "Only a loose loading/memory theme; the concrete bugs are different." + "reason": "These are separate issues: optimizer/deepspeed OOM versus a specific model-loading breakage." }, { - "left": "issue:44077", - "right": "issue:44479", + "left": "issue:41628", + "right": "issue:45372", "accept": false, - "reason": "PatchTSMixer post_init policy and Qwen video-input regression are unrelated." + "reason": "Both are import errors, but they involve different missing symbols and different affected components." }, { - "left": "issue:41628", - "right": "issue:45372", + "left": "issue:44077", + "right": "issue:44479", "accept": false, - "reason": "Different missing imports for different classes; no shared underlying defect." + "reason": "PatchTSMixer post_init policy and Qwen video-input regression are unrelated." }, { "left": "issue:43720", "right": "issue:44442", "accept": false, - "reason": "BitNet accelerate-loading unpacking bug is unrelated to a tokenizer loading failure." + "reason": "Packed-weight unpacking during accelerate loading is unrelated to tokenizer loading." }, { - "left": "issue:43749", - "right": "issue:43931", + "left": "issue:43643", + "right": "issue:44913", "accept": false, - "reason": "FSDP CPU RAM efficient loading failure is not the same as a Qwen3-VL weight-shape mismatch." + "reason": "Missing fields with trust_remote_code and GPTNeoX rotary_pct reset are both config-related but not the same underlying defect." }, { - "left": "issue:43643", - "right": "issue:44913", + "left": "issue:43749", + "right": "issue:43931", "accept": false, - "reason": "Both involve config persistence, but one is trust_remote_code returning missing fields while the other is GPTNeoX rotary_pct reload behavior." + "reason": "FSDP CPU RAM efficient loading and Qwen3-VL weight-shape mismatch are different loading problems." }, { "left": "issue:43873", "right": "issue:44871", "accept": false, - "reason": "Quantization/offloading behavior and Gemma-3 eos_token_id mismatch are different bugs." + "reason": "Quantization offloading and Gemma-3 eos_token_id mismatch are unrelated." }, { "left": "issue:44190", "right": "issue:44496", "accept": false, - "reason": "Local dataset loading in a training script is unrelated to an unrecognized model/config error." + "reason": "A local dataset loading failure in an example script is not the same as an unrecognized model/config loading error." }, { "left": "issue:43381", "right": "issue:43856", "accept": false, - "reason": "Eval-mode gradient checkpointing and Qwen3 MoE memory inefficiency are distinct issues." + "reason": "One is a gradient-checkpointing eval-mode restriction; the other is a Qwen3 MoE memory-efficiency complaint." }, { "left": "issue:33290", "right": "issue:36331", "accept": false, - "reason": "Adafactor/DeepSpeed OOM is unrelated to a Trainer compute_loss signature change." + "reason": "OOM in DeepSpeed is unrelated to a custom trainer signature change." }, { "left": "issue:44855", "right": "issue:45137", "accept": false, - "reason": "Python parsing/import failure in DebertaV2Model has no overlap with a DeepSpeed ZeRO3 deque error." + "reason": "Python 3.13 parsing/import error in DebertaV2Model is unrelated to a ZeRO3 deque underflow." }, { "left": "issue:34689", "right": "issue:36010", "accept": false, - "reason": "A model-loading regression and a GenerationMixin import error are different breakages." + "reason": "A model-loading regression for Llama 3.2 Vision is not the same as the GenerationMixin import-path break." }, { "left": "issue:43540", "right": "issue:44514", "accept": false, - "reason": "Both touch vision/video processing, but the concrete failures are different processor paths and symptoms." + "reason": "Both are Qwen-related processor failures, but one is video input processing in Qwen3OmniMoe and the other is batched chat-template padding behavior." }, { "left": "issue:43316", "right": "issue:44877", "accept": false, - "reason": "Different config-loading problems with different models and failure modes." + "reason": "Gemma3TextConfig API inconsistency and strict config loading for granite_speech are different config issues." }, { "left": "issue:44162", "right": "issue:44792", "accept": false, - "reason": "ESM2 runtime breakage and a Janus test failure are not the same underlying bug." + "reason": "ESM2 breakage and a janus model test failure do not describe the same underlying bug." }, { "left": "issue:43388", "right": "issue:43688", "accept": false, - "reason": "Metrics-gathering tuple truncation is unrelated to auxiliary-loss normalization." + "reason": "Metric gathering dropping tuple labels and auxiliary-loss normalization are unrelated training bugs." }, { "left": "issue:42175", "right": "issue:43316", "accept": false, - "reason": "Backend packaging for TensorFlow is unrelated to a Gemma3TextConfig API discrepancy." + "reason": "Backend packaging for TensorFlow and a Gemma3TextConfig API discrepancy are unrelated." }, { "left": "issue:43404", "right": "issue:43441", "accept": false, - "reason": "Weight tying in Mistral3 and FlashAttention failure in Ministral-3 are separate model-specific issues." + "reason": "Weight tying in Mistral3ForConditionalGeneration and a FlashAttention failure in Ministral-3 are different issues, even if both concern related model families." } ] }, @@ -8136,10 +8126,10 @@ "issue:16998|issue:22355", "issue:25251|issue:28282", "issue:43316|issue:43716", - "issue:41950|issue:43475", "issue:43065|issue:44617", - "issue:43295|issue:43825", + "issue:41950|issue:43475", "issue:45127|issue:45200", + "issue:43295|issue:43825", "issue:44871|issue:45584", "issue:43761|issue:44479", "issue:44077|issue:44534", @@ -8151,26 +8141,26 @@ "issue:43906|issue:45290", "issue:43452|issue:44496", "issue:43761|issue:44242", - "issue:45563|issue:45584", "issue:43575|issue:43742", - "issue:43931|issue:44060", "issue:42617|issue:43575", "issue:44617|issue:45584", + "issue:45563|issue:45584", + "issue:43931|issue:44060", "issue:44521|issue:45381", "issue:43575|issue:43873", "issue:32090|issue:35141", "issue:44479|issue:44734", "issue:44933|issue:45362", - "issue:44464|issue:45563", "issue:44464|issue:44871", + "issue:44464|issue:45563", "issue:44792|issue:44977" ], - "prepared_review_unit_hash": "7364df1ec800937d31132d982c4fd5fddeb256b72abfed221e3863b8d26eb1e4", + "prepared_review_unit_hash": "083a739e41d2bae7b0e9b3f54f2cfd28b209db1a779577fc4ab29f799a17ccb6", "component_budget": { "node_count": 367, "item_count": 367, "soft_pair_count": 1370, - "serialized_chars": 706987, + "serialized_chars": 706986, "estimated_input_tokens": 176747, "estimated_eval_tokens": 353750 }, @@ -8187,198 +8177,198 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a set of unrelated issues that only look superficially similar because they involve Transformers regressions, multimodal inputs, loading, or configuration. I do not see any true duplicate pair here.", - "confidence": 0.98, - "canonical_issue_reason": "Issue 43873 is the broadest and most active open item in the set, with multiple references and discussion, so it is the closest thing to a hub; however, the cluster is heterogeneous rather than a single duplicate chain.", + "summary": "All suggested soft edges look like false positives. The items span unrelated bugs across loading, generation, multimodal processing, training, and model-specific regressions, with no pair clearly describing the same underlying fix. No pull requests are present.", + "confidence": 0.97, + "canonical_issue_reason": "No single canonical issue stands out: the candidates are about different subsystems and failure modes, so none is a safe representative for deduplication.", "canonical_pr_reason": null, - "best_issue_reason": "43873 is the best representative only in a loose sense: it is open, active, and broadly phrased. It is not a strong canonical duplicate target for the rest of the items because the underlying bugs differ.", + "best_issue_reason": "No issue is a strong global representative because the cluster is not a true duplicate set; choosing one would conflate unrelated reports.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:16998", "right": "issue:22355", "accept": false, - "reason": "Different problems: model_max_length handling for DeBERTa-v3 vs a missing transformers.onnx import/module path." + "reason": "Different problems: DeBERTa max length handling vs missing `transformers.onnx` module import." }, { "left": "issue:25251", "right": "issue:28282", "accept": false, - "reason": "Pipeline top_k nesting bug is unrelated to an AutoModel PyTorch dependency import error." + "reason": "Pipeline `top_k` output nesting bug is unrelated to the PyTorch import error for `AutoModel`." }, { "left": "issue:43316", "right": "issue:43716", "accept": false, - "reason": "Gemma3TextConfig API mismatch is separate from the Mistral-3 image preprocessor dtype mismatch." - }, - { - "left": "issue:41950", - "right": "issue:43475", - "accept": false, - "reason": "Video-classification pipeline using image processors is not the same bug as Sam3VisionEncoderOutput missing fpn_position_embeddings." + "reason": "Gemma3TextConfig API inconsistency is unrelated to the Mistral-3 image preprocessor dtype mismatch." }, { "left": "issue:43065", "right": "issue:44617", "accept": false, - "reason": "Sam3PixelDecoder dummy Conv2d vs Sam3Video CUDA OOM are different SAM 3 issues with different failure modes." + "reason": "`Sam3PixelDecoder` dummy conv issue is a different bug from Sam3Video CUDA OOM." }, { - "left": "issue:43295", - "right": "issue:43825", + "left": "issue:41950", + "right": "issue:43475", "accept": false, - "reason": "Processor.tokenizer access regression is distinct from the pipeline translation-task error message regression." + "reason": "Video-classification processor lookup and missing `fpn_position_embeddings` are not the same code-path." }, { "left": "issue:45127", "right": "issue:45200", "accept": false, - "reason": "LoRA merge with extended vocab causing collapse is unrelated to Gemma 4 mm_token_type_ids defaulting for text-only fine-tuning." + "reason": "LoRA merge/tied embeddings collapse is unrelated to Gemma 4 `mm_token_type_ids` defaults." + }, + { + "left": "issue:43295", + "right": "issue:43825", + "accept": false, + "reason": "Processor/tokenizer regression in v4.57.5 is different from the pipeline translation-task error message." }, { "left": "issue:44871", "right": "issue:45584", "accept": false, - "reason": "Gemma-3 eos_token_id mismatch is separate from Whisper empty-transcription handling after align_special_tokens." + "reason": "Gemma-3 EOS token config mismatch is unrelated to Whisper empty-transcription handling." }, { "left": "issue:43761", "right": "issue:44479", "accept": false, - "reason": "CLIPVisionModel hidden_states regression is not the same as the multi-model video-input regression in v5.3.0." + "reason": "CLIP hidden_states regression and Qwen video-input regression affect different models and failures." }, { "left": "issue:44077", "right": "issue:44534", "accept": false, - "reason": "PatchTSMixer post_init allowance is unrelated to non-persistent buffers being filled with junk." + "reason": "`patchtsmixer` post_init restriction is unrelated to non-persistent buffer junk in v5." }, { "left": "issue:43550", "right": "issue:43701", "accept": false, - "reason": "Bamba-9B-v2 torch.compile/SDPA failure is a different bug than resume_from_checkpoint key mismatch." + "reason": "Bamba torch.compile/SDPA failure and resume_from_checkpoint key mismatch are different bugs." }, { "left": "issue:43295", "right": "issue:44871", "accept": false, - "reason": "Both mention model/processor config regressions, but one is about processor.tokenizer access and the other is about eos_token_id consistency." + "reason": "Custom processor/tokenizer regression is not the same as Gemma-3 EOS token inconsistency." }, { "left": "issue:43606", "right": "issue:43701", "accept": false, - "reason": "CPU offload device-mismatch in bark-small does not match resume_from_checkpoint key mismatch." + "reason": "CPU offload device mismatch for Bark is unrelated to checkpoint key mismatch on resume." }, { "left": "issue:44734", "right": "issue:45406", "accept": false, - "reason": "KV-cache continuation tensor indexing in serve is distinct from Gemma4Processor missing _tokenizer." + "reason": "KV-cache continuation indexing in serve is unrelated to Gemma4Processor missing `_tokenizer`." }, { "left": "issue:43646", "right": "issue:44913", "accept": false, - "reason": "Custom model initialization breakage is not the same as GPTNeoXConfig rotary_pct not persisting on reload." + "reason": "Custom model initialization breakage and GPTNeoX rotary_pct reload loss are different configuration bugs." }, { "left": "issue:43906", "right": "issue:45290", "accept": false, - "reason": "Isolated reproduction of another issue is unrelated to apply_chat_template crashing on tool-call assistant messages." + "reason": "An isolated reproduction of a prior issue is not the same as the tool-call assistant message crash." }, { "left": "issue:43452", "right": "issue:44496", "accept": false, - "reason": "gguf_file loading failures for tokenizer/model are different from an unrecognized model_type in config.json." + "reason": "`gguf_file` loading regression is distinct from the unrecognized-model/config.json error." }, { "left": "issue:43761", "right": "issue:44242", "accept": false, - "reason": "Hidden_states regression in CLIPVisionModel is separate from missing load-balancing loss when output_router_logits=False." - }, - { - "left": "issue:45563", - "right": "issue:45584", - "accept": false, - "reason": "Paged generate warning for num_return_sequences is unrelated to Whisper failing on empty transcription." + "reason": "CLIP hidden_states regression is unrelated to MoE load-balancing loss omission." }, { "left": "issue:43575", "right": "issue:43742", "accept": false, - "reason": "TP loading OOM for Qwen2-57B is not the same as a key error loading facebook/MobileLLM-125M." - }, - { - "left": "issue:43931", - "right": "issue:44060", - "accept": false, - "reason": "Qwen3-VL weight-shape mismatch is different from Qwen3-Next tied-weights warning wiring embed_tokens to dt_bias." + "reason": "Qwen2-57B tensor-parallel OOM and MobileLLM key error are unrelated model-loading failures." }, { "left": "issue:42617", "right": "issue:43575", "accept": false, - "reason": "3d_parallel.py execution failure is unrelated to TP loading OOM for Qwen2-57B." + "reason": "Unable to run `3d_parallel.py` is not the same as Qwen2-57B TP OOM." }, { "left": "issue:44617", "right": "issue:45584", "accept": false, - "reason": "Sam3Video OOM is a resource issue, not the Whisper empty-transcription regression." + "reason": "Sam3Video OOM is unrelated to Whisper empty transcription after token alignment." + }, + { + "left": "issue:45563", + "right": "issue:45584", + "accept": false, + "reason": "Paged generate warning for `num_return_sequences` is unrelated to Whisper empty transcription." + }, + { + "left": "issue:43931", + "right": "issue:44060", + "accept": false, + "reason": "Qwen3-VL shape mismatch and Qwen3-Next tied-weights warning are different model/config bugs." }, { "left": "issue:44521", "right": "issue:45381", "accept": false, - "reason": "All-zero assistant masks for multimodal apply_chat_template is a different bug from qwen2.5-vl video vision_position_ids being wrong." + "reason": "Assistant-mask failure for multimodal chat is not the same as qwen2.5-vl video position-id regression." }, { "left": "issue:43575", "right": "issue:43873", "accept": false, - "reason": "Qwen2-57B TP OOM and offloading-with-quantization are both memory-related but are different concrete failure paths." + "reason": "Qwen2-57B OOM is a different symptom from offloading/quantization behavior." }, { "left": "issue:32090", "right": "issue:35141", "accept": false, - "reason": "Trainer GPU broadcast NoneType error is unrelated to output embeddings being reinitialized after resizing token embeddings." + "reason": "Trainer `_gpu_broadcast_one` NoneType error and embedding reinitialization after resize are unrelated." }, { "left": "issue:44479", "right": "issue:44734", "accept": false, - "reason": "Video-input regression across several models is distinct from serve KV-cache continuation tensor indexing." + "reason": "Qwen video-input regression and serve KV-cache indexing crash affect different paths." }, { "left": "issue:44933", "right": "issue:45362", "accept": false, - "reason": "Missing import from image_utils is unrelated to Qwen3.5-35B crashing with transformers chat." + "reason": "Nonexistent `image_utils` import and Qwen3.5 chat crash are unrelated." }, { "left": "issue:44464", - "right": "issue:45563", + "right": "issue:44871", "accept": false, - "reason": "Chunked generation inconsistency under compiled forward is not the same as the paged-generate stale warning." + "reason": "Chunked generation inconsistency with compiled forward is unrelated to Gemma-3 EOS config mismatch." }, { "left": "issue:44464", - "right": "issue:44871", + "right": "issue:45563", "accept": false, - "reason": "Compiled-forward generation inconsistency is different from Gemma-3 eos_token_id configuration mismatch." + "reason": "Chunked generation/compiled forward differs from the stale paged-generate warning." }, { "left": "issue:44792", "right": "issue:44977", "accept": false, - "reason": "Janus test failure for image generation is unrelated to Qwen3.5 flash-attention generation issues." + "reason": "Janus test failure and Qwen3.5 flash-attention generation bug are different model-specific issues." } ] }, @@ -8403,7 +8393,6 @@ "cluster_id": "cluster-15354-367", "nodes": [ "issue:36246", - "issue:38175", "issue:38617", "issue:39692", "issue:41628", @@ -8432,6 +8421,7 @@ "issue:44190", "issue:44220", "issue:44355", + "issue:44466", "issue:44496", "issue:44534", "issue:44610", @@ -8452,20 +8442,20 @@ "issue:45584" ], "soft_pairs": [ - "issue:43065|issue:44610", "issue:43827|issue:43994", + "issue:43065|issue:44610", "issue:44534|issue:44913", "issue:41720|issue:45245", "issue:44938|issue:45542", "issue:44190|issue:44964", - "issue:43388|issue:43582", "issue:43650|issue:44220", + "issue:43388|issue:43582", "issue:45127|issue:45357", "issue:43761|issue:44743", "issue:45276|issue:45584", "issue:43644|issue:44913", - "issue:44355|issue:44855", "issue:44077|issue:44079", + "issue:44355|issue:44855", "issue:43540|issue:44734", "issue:43976|issue:44855", "issue:44496|issue:45372", @@ -8481,14 +8471,14 @@ "issue:36246|issue:41628", "issue:44186|issue:44734", "issue:36246|issue:38617", - "issue:38175|issue:43065" + "issue:44466|issue:44913" ], - "prepared_review_unit_hash": "47a7bcc3fac626a93685d66b4e5ec102015e19b34d43f66809d374f45fc49853", + "prepared_review_unit_hash": "a60c003ffb334595e4457c551ee93cf0ce67a0ec85a04ca83cde65a9dddebff4", "component_budget": { "node_count": 367, "item_count": 367, "soft_pair_count": 1370, - "serialized_chars": 706987, + "serialized_chars": 706986, "estimated_input_tokens": 176747, "estimated_eval_tokens": 353750 }, @@ -8496,195 +8486,201 @@ "node_count": 48, "item_count": 48, "soft_pair_count": 30, - "serialized_chars": 36999, - "estimated_input_tokens": 9250, - "estimated_eval_tokens": 18756 + "serialized_chars": 37070, + "estimated_input_tokens": 9268, + "estimated_eval_tokens": 18792 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Mostly unrelated Transformers issues with a few superficial overlaps around multimodal processors, pipelines, and version regressions. None of the soft-edge pairs look like the same concrete bug/change, so I would not merge any of them as duplicates.", - "confidence": 0.72, - "canonical_issue_reason": "issue:43295 is the broadest, clearest regression here: a processor/tokenizer API break that affects custom multimodal code and has a concrete, user-facing failure mode.", + "summary": "This looks like a grab-bag of unrelated Transformers bugs. None of the proposed soft edges appear to be the same underlying issue, so I would not merge any of these items into a duplicate cluster.", + "confidence": 0.96, + "canonical_issue_reason": "No clear canonical issue: the items span distinct failures (imports, config loading, model serialization, training scripts, and serving/runtime bugs) rather than one shared bug.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43295 is the strongest triage anchor: explicit regression, clear repro surface, and broader applicability than the more model-specific or narrowly scoped reports.", + "best_issue_reason": "If forced to pick a representative, issue 43295 is the broadest user-facing regression, but it still does not represent the other issues well enough to serve as a true canonical duplicate.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:43065", - "right": "issue:44610", + "left": "issue:43827", + "right": "issue:43994", "accept": false, - "reason": "Both are vision/model-related, but one is about a dummy Conv2d in Sam3PixelDecoder and the other is an OmDet-Turbo processor size mismatch; different code paths and failures." + "reason": "Docs still referencing pipeline() removals is unrelated to SigLIP2 producing bad outputs; different symptom and code path." }, { - "left": "issue:43827", - "right": "issue:43994", + "left": "issue:43065", + "right": "issue:44610", "accept": false, - "reason": "One is documentation still using removed pipeline() calls; the other is bad SigLIP2 inference results with AutoModel/pipeline. Same ecosystem, not the same bug." + "reason": "Sam3PixelDecoder dummy conv2d and OmDet-Turbo processor size mismatch are separate vision-model bugs." }, { "left": "issue:44534", "right": "issue:44913", "accept": false, - "reason": "Non-persistent buffer initialization junk is unrelated to GPTNeoX rotary_pct not persisting on reload." + "reason": "Non-persistent buffer junk serialization and GPTNeoX rotary_pct reload are different configuration/serialization problems." }, { "left": "issue:41720", "right": "issue:45245", "accept": false, - "reason": "CUDA assert on Qwen3 auto device mapping and a categories-limit RuntimeError are unrelated failures." + "reason": "Qwen3 auto device mapping CUDA assert is unrelated to the category-count limit RuntimeError." }, { "left": "issue:44938", "right": "issue:45542", "accept": false, - "reason": "Python 3.14 import/load failure and a TensorBoard-without-TensorFlow backend error are different issues." + "reason": "Python 3.14 import/load failure and missing TensorFlow backend error are distinct environment/package issues." }, { "left": "issue:44190", "right": "issue:44964", "accept": false, - "reason": "Local dataset loading in an image-classification script is unrelated to loading a Phi-4 multimodal checkpoint." + "reason": "Local dataset loading in an image-classification script is not the same bug as loading Phi-4 multimodal models." }, { - "left": "issue:43388", - "right": "issue:43582", + "left": "issue:43650", + "right": "issue:44220", "accept": false, - "reason": "Metrics gathering tuple-label truncation and an Apple Silicon allocator warmup TypeError do not share a code path." + "reason": "The placeholder \"ADD THE DATA\" issue and the fbank feature extraction bug do not match." }, { - "left": "issue:43650", - "right": "issue:44220", + "left": "issue:43388", + "right": "issue:43582", "accept": false, - "reason": "The placeholder 'ADD THE DATA' issue does not match the audio feature extraction bug." + "reason": "Metric label truncation and Apple Silicon allocator warmup TypeError are unrelated." }, { "left": "issue:45127", "right": "issue:45357", "accept": false, - "reason": "Both mention Qwen3.5 serialization/weight handling, but one is LoRA + extended vocab collapse and the other is wrong visual encoder keys on save_pretrained; distinct root causes." + "reason": "Both involve Qwen models, but one is LoRA/extended-vocab merge collapse and the other is incorrect visual-encoder save keys; different concrete failures." }, { "left": "issue:43761", "right": "issue:44743", "accept": false, - "reason": "CLIPVisionModel hidden_states regression is unrelated to Qwen3 recurrent states resetting with cache." + "reason": "CLIPVisionModel hidden_states regression and Qwen3 modular recurrent-state reset are separate model behaviors." }, { "left": "issue:45276", "right": "issue:45584", "accept": false, - "reason": "resize_token_embeddings not updating Gemma4 embeddings/output is unrelated to Whisper empty-transcription generation after align_special_tokens." + "reason": "Gemma4 embedding resize behavior and Whisper empty transcription after align_special_tokens are unrelated." }, { "left": "issue:43644", "right": "issue:44913", "accept": false, - "reason": "Global non-persistent buffer junk and GPTNeoX config reload losing rotary_pct are different bugs." + "reason": "Non-persistent buffer junk in v5 and GPTNeoX rotary_pct reload are different bugs." }, { - "left": "issue:44355", - "right": "issue:44855", + "left": "issue:44077", + "right": "issue:44079", "accept": false, - "reason": "Compiled Python file errors are not the same as DebertaV2Model import parsing failure on Python 3.13." + "reason": "Optional post_init handling in patchtsmixer is not the same as ModelOutput key assignment when values were previously None." }, { - "left": "issue:44077", - "right": "issue:44079", + "left": "issue:44355", + "right": "issue:44855", "accept": false, - "reason": "A model-specific post_init constraint and ModelOutput key assignment semantics are unrelated." + "reason": "Compiled Python file errors and DebertaV2Model import indentation parsing are different failure modes." }, { "left": "issue:43540", "right": "issue:44734", "accept": false, - "reason": "Qwen3OmniMoe video input validation and transformers serve KV-cache tensor indexing are separate code paths." + "reason": "Qwen3OmniMoe video-input validation and serve KV-cache continuation indexing affect different code paths." }, { "left": "issue:43976", "right": "issue:44855", "accept": false, - "reason": "Python version compatibility problem and @torch.jit.script parsing/indentation import error are different failure modes." + "reason": "Python version compatibility for Transformers 5.1.0 is unrelated to the DebertaV2 @torch.jit.script parsing error." }, { "left": "issue:44496", "right": "issue:45372", "accept": false, - "reason": "Unrecognized model_type in config.json is unrelated to a mistral_common ReasoningEffort import breaking Gemma 4 processor loading." + "reason": "Unrecognized model/config loading and mistral_common ReasoningEffort import breakage are different loader issues." }, { "left": "issue:43335", "right": "issue:44855", "accept": false, - "reason": "SwitchTransformers sparse-layer construction and DebertaV2 Python 3.13 import parsing are not the same bug." + "reason": "Sparse-layer creation in SwitchTransformersConfig is unrelated to the DebertaV2 import parsing problem." }, { "left": "issue:43550", "right": "issue:44190", "accept": false, - "reason": "torch.compile SDPA failure in Bamba-9B-v2 is unrelated to local dataset loading in an image-classification example." + "reason": "Bamba torch.compile+SDPA failure and local dataset loading in an image-classification script are unrelated." }, { "left": "issue:43404", "right": "issue:45216", "accept": false, - "reason": "Tied lm_head weights in Mistral3 and incorrect Qwen3.5 save_pretrained checkpoint contents are different serialization/model bugs." + "reason": "lm_head weight tying in Mistral3ForConditionalGeneration and Qwen3.5 save_pretrained regression are both weight/save issues but not the same concrete bug." }, { "left": "issue:43295", "right": "issue:45276", "accept": false, - "reason": "Processor.tokenizer/images-to-tokenizer regression is unrelated to Gemma4 resize_token_embeddings not propagating." + "reason": "Processor.tokenizer/image-to-tokenizer regression and Gemma4 embedding resize not propagating are different API regressions." }, { "left": "issue:43116", "right": "issue:44190", "accept": false, - "reason": "Multi-label classification returning empty results is unrelated to loading a local dataset in a different example script." + "reason": "Multilabel classification returning empty results is not the same as failing to load a local dataset." + }, + { + "left": "issue:45245", + "right": "issue:45584", + "accept": false, + "reason": "Category-cardinality limit and Whisper empty transcription are unrelated errors." }, { "left": "issue:39692", "right": "issue:44190", "accept": false, - "reason": "SigLIP2 docs/model mismatch and quantization example problems are not the same as local dataset loading failures." + "reason": "SigLIP2 docs example errors and the local image-classification dataset loading bug are different issues." }, { "left": "issue:41950", "right": "issue:44016", "accept": false, - "reason": "Video-classification pipeline using image processors and a syntax error in a notebook are unrelated." + "reason": "Video-classification pipeline processor lookup and a notebook syntax error are unrelated." }, { "left": "issue:43479", "right": "issue:45375", "accept": false, - "reason": "Both involve multimodal config fields, but one is default vision/audio config initialization and the other is a missing deepstack_visual_indexes field dropped by strict serialization; not the same bug." + "reason": "Phi4MultimodalConfig default-init behavior and Qwen3_5MoeVisionConfig missing field handling are separate config bugs." }, { "left": "issue:36246", "right": "issue:41628", "accept": false, - "reason": "Cannot import Qwen2_5_VLImageProcessor and cannot import AutoImageProcessor are different import failures in different APIs." + "reason": "Both are import errors, but they refer to different missing symbols and different import surfaces; not the same bug." }, { "left": "issue:44186", "right": "issue:44734", "accept": false, - "reason": "LayoutLMv2Tokenizer padding/truncation crashes and serve KV-cache continuation tensor indexing are unrelated." + "reason": "LayoutLMv2Tokenizer NER/padding crashes are unrelated to serve KV-cache continuation indexing." }, { "left": "issue:36246", "right": "issue:38617", "accept": false, - "reason": "Importing Qwen2_5_VLImageProcessor and missing layer_type_validation in configuration_utils are distinct import errors." + "reason": "Different missing imports in different modules; no shared underlying fix is evident." }, { - "left": "issue:38175", - "right": "issue:43065", + "left": "issue:44466", + "right": "issue:44913", "accept": false, - "reason": "SigLIP2 zero probabilities and a dummy Conv2d in Sam3PixelDecoder do not describe the same underlying problem." + "reason": "Device-dependent lm_head serialization inconsistency and rotary_pct reload persistence are different configuration bugs." } ] }, @@ -8708,8 +8704,10 @@ "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-15354-367", "nodes": [ + "issue:38175", "issue:42175", "issue:42907", + "issue:43065", "issue:43295", "issue:43299", "issue:43441", @@ -8734,12 +8732,10 @@ "issue:44336", "issue:44351", "issue:44368", - "issue:44466", "issue:44496", "issue:44521", "issue:44683", "issue:44792", - "issue:44913", "issue:44933", "issue:44964", "issue:44987", @@ -8757,15 +8753,15 @@ "issue:45584" ], "soft_pairs": [ - "issue:44466|issue:44913", + "issue:38175|issue:43065", "issue:44683|issue:45446", "issue:44016|issue:44220", "issue:44246|issue:45003", "issue:42175|issue:44279", "issue:44112|issue:44792", - "issue:43295|issue:45335", "issue:43643|issue:44496", "issue:43299|issue:45216", + "issue:43295|issue:45335", "issue:42907|issue:43441", "issue:44016|issue:44062", "issue:44112|issue:44964", @@ -8788,12 +8784,12 @@ "issue:44351|issue:45278", "issue:43479|issue:43525" ], - "prepared_review_unit_hash": "5877d7e3888fca4554eff7f9c425f18b6b33dcd46d5cce5411c6ca46c3ada4d6", + "prepared_review_unit_hash": "b373b4767fd548755848602c879d709fd65c603b800bacc49d79b57ed17283e3", "component_budget": { "node_count": 367, "item_count": 367, "soft_pair_count": 1370, - "serialized_chars": 706987, + "serialized_chars": 706986, "estimated_input_tokens": 176747, "estimated_eval_tokens": 353750 }, @@ -8801,201 +8797,201 @@ "node_count": 47, "item_count": 47, "soft_pair_count": 30, - "serialized_chars": 36077, - "estimated_input_tokens": 9020, - "estimated_eval_tokens": 18296 + "serialized_chars": 35976, + "estimated_input_tokens": 8994, + "estimated_eval_tokens": 18244 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These issues are mostly loose semantic neighbors around Transformers regressions, imports, model loading, config handling, and backend dependencies, but they do not describe the same underlying bug or fix. None of the soft-similarity pairs are strong enough to merge as duplicates.", - "confidence": 0.86, - "canonical_issue_reason": "issue:45278 is the broadest representative: it reports widespread import failures after upgrading to 5.5.0, which makes it the most umbrella-like issue in this set. It is still not a duplicate of the others, but it is the least specific and best canonical anchor if one must be chosen.", + "summary": "This cluster is mostly a grab bag of Transformers regression/import/model-specific issues rather than true duplicates. All soft-edge pairs look like false positives; several share a subsystem or symptom class, but not the same concrete bug.", + "confidence": 0.84, + "canonical_issue_reason": "issue:45278 is the broadest representative of the cluster\u2019s recurring theme: post-upgrade import/runtime breakage. It\u2019s still not a true duplicate of the others, but it best captures the generalized failure mode compared with the many narrower model- or API-specific reports.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45278 is the best single issue representative because it is user-facing, broad, and captures a whole class of regression complaints rather than one narrow model-specific case.", + "best_issue_reason": "issue:45278 is the best single issue to stand in for this cluster because it is the most umbrella-like and most likely to subsume related import failures after upgrading Transformers.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:44466", - "right": "issue:44913", + "left": "issue:38175", + "right": "issue:43065", "accept": false, - "reason": "Both involve config/serialization behavior, but one is about tied-weight serialization on save and the other about rotary_pct not persisting on reload; different code paths and symptoms." + "reason": "Different models and different failures: zero probabilities in SigLIP2 versus a dummy Conv2d in Sam3PixelDecoder." }, { "left": "issue:44683", "right": "issue:45446", "accept": false, - "reason": "Both mention flex_attention, but one is a runtime failure on torch>=2.9 and the other is a bad version check for an import path; not the same concrete bug." + "reason": "Both involve flex_attention, but they describe different concrete problems: a torch>=2.9 compiled flex_attention failure versus a PyTorch version check for AuxRequest import." }, { "left": "issue:44016", "right": "issue:44220", "accept": false, - "reason": "Notebook syntax error versus feature extraction helper issue; no shared underlying bug." + "reason": "A notebook syntax error and an audio feature-extraction bug are unrelated." }, { "left": "issue:44246", "right": "issue:45003", "accept": false, - "reason": "Both touch import-time behavior, but one is intermittent slow imports and the other is unsafe sys.modules access; related area, not the same defect." + "reason": "Import slowdowns and unsafe sys.modules access are not the same bug; one is performance, the other is unsafe import handling." }, { "left": "issue:42175", "right": "issue:44279", "accept": false, - "reason": "One is a specific extras/install dependency packaging bug, the other is a generic dependency issue with no matching failure mode." + "reason": "The first is a specific missing TensorFlow backend in a torch install; the second is a vague dependency issue with no evidence of the same root cause." }, { "left": "issue:44112", "right": "issue:44792", "accept": false, - "reason": "Both are CI/test failures, but they are for different models and different failure surfaces." - }, - { - "left": "issue:43295", - "right": "issue:45335", - "accept": false, - "reason": "Processor/tokenizer regression versus t5gemma embedding resize bug; unrelated functionality." + "reason": "Different failing tests in different models; same general CI symptom, not the same underlying defect." }, { "left": "issue:43643", "right": "issue:44496", "accept": false, - "reason": "Both are config-loading errors, but one is about trust_remote_code missing fields and the other about missing model_type in config.json; distinct causes." + "reason": "One is a trust_remote_code/config field propagation issue; the other is an unrecognized model config problem." }, { "left": "issue:43299", "right": "issue:45216", "accept": false, - "reason": "One is a load-time breakage for Qwen3VL MoE models, the other is a save_pretrained checkpoint regression for Qwen3.5; different operations and bug types." + "reason": "Both are Qwen-related regressions, but one is a model loading failure and the other is a save_pretrained checkpoint serialization bug." + }, + { + "left": "issue:43295", + "right": "issue:45335", + "accept": false, + "reason": "Different APIs and code paths: processor/tokenizer access regression versus resize_token_embeddings not updating decoder embeddings." }, { "left": "issue:42907", "right": "issue:43441", "accept": false, - "reason": "Same model family, but one is about saving dequantized models and the other is about FlashAttention loading/runtime failure." + "reason": "Both mention Ministral, but one is about saving dequantized models and the other about FlashAttention failures." }, { "left": "issue:44016", "right": "issue:44062", "accept": false, - "reason": "Notebook syntax error versus AddedToken keyword conflict; unrelated bugs." + "reason": "Notebook syntax error versus AddedToken special-argument TypeError; unrelated." }, { "left": "issue:44112", "right": "issue:44964", "accept": false, - "reason": "Both are model-specific failures, but for different models and different error modes." + "reason": "Different models and failures; a GraniteSpeech CI test issue is not the same as Phi-4 multimodal loading failure." }, { "left": "issue:45335", "right": "issue:45584", "accept": false, - "reason": "Embedding resize issue versus Whisper generation on empty transcription; different subsystems." + "reason": "Tokenizer resize behavior and Whisper empty-transcription generation are different bugs." }, { "left": "issue:43673", "right": "issue:44336", "accept": false, - "reason": "Generation cache regression versus ANSI escape leakage in loading_report; no common bug." + "reason": "Generation cache missing during chunked prefill and ANSI code leakage in loading_report are unrelated." }, { "left": "issue:43976", "right": "issue:44987", "accept": false, - "reason": "Python version compatibility problem versus model-loading failure on a specific checkpoint; not the same change." + "reason": "A Python-version compatibility issue is not the same as a model-loading failure for physical-intelligence/fast." }, { "left": "issue:43854", "right": "issue:44964", "accept": false, - "reason": "Two model-loading failures, but on different models and with different likely root causes." + "reason": "Different model families and different failures: GLM-4.7-Flash unit-test loading versus Phi-4 multimodal loading." }, { "left": "issue:44496", "right": "issue:45042", "accept": false, - "reason": "Config/model-type recognition bug versus PIL backend unnecessarily requiring torchvision; unrelated." + "reason": "An unrecognized model config and a torchvision dependency regression in PIL image processors are unrelated." }, { "left": "issue:43824", "right": "issue:44933", "accept": false, - "reason": "Both are import errors, but they concern different symbols/modules and different causes." + "reason": "Both are import errors, but for different missing symbols and different modules; too generic to treat as duplicates." }, { "left": "issue:43526", "right": "issue:43901", "accept": false, - "reason": "ImageProcessorFast label reduction bug versus pipeline docs/behavior mismatch; not the same issue." + "reason": "A label-reduction bug in BeitImageProcessorFast is unrelated to stale pipeline documentation." }, { "left": "issue:43976", "right": "issue:44279", "accept": false, - "reason": "Generic dependency complaint versus Python-version incompatibility; too vague to merge." + "reason": "A Python compatibility problem is not the same as a vague dependency issue." }, { "left": "issue:43299", "right": "issue:45357", "accept": false, - "reason": "Both concern Qwen3.5/Qwen3.5-VL save/load regressions, but one is loading MoE models and the other is incorrect visual encoder keys during save; distinct bugs." + "reason": "Both are Qwen3.5-related, but one is a loading regression and the other is incorrect visual-encoder key saving." }, { "left": "issue:43577", "right": "issue:45216", "accept": false, - "reason": "Both involve model dtype/save behavior, but one is Blip2 loading dtype propagation and the other is Qwen3.5 checkpoint serialization." + "reason": "Blip2 dtype propagation and Qwen3.5 checkpoint saving are different code paths and different bugs." }, { "left": "issue:44075", "right": "issue:44368", "accept": false, - "reason": "Optimizer argument handling versus a warning about tie_word_embeddings; no shared code path." + "reason": "Optimizer argument handling and a tie_word_embeddings warning are unrelated." }, { "left": "issue:44521", "right": "issue:45325", "accept": false, - "reason": "Both are multimodal/chat-template related, but one is all-zero assistant masks and the other is rope-index temporal scaling for Qwen2.5-VL." + "reason": "Both are multimodal, but assistant mask generation and rope/position-id scaling are different failures." }, { "left": "issue:43901", "right": "issue:44496", "accept": false, - "reason": "Docs/behavior mismatch for a pipeline versus model config loading error; unrelated." + "reason": "Documentation mismatch versus model config recognition failure; unrelated." }, { "left": "issue:44242", "right": "issue:45200", "accept": false, - "reason": "Mixture-of-experts router loss bug versus mm_token_type_ids defaulting issue; different training paths." + "reason": "Load-balancing loss behavior in MoE models is not the same as mm_token_type_ids defaulting for text-only fine-tuning." }, { "left": "issue:44683", "right": "issue:45542", "accept": false, - "reason": "Flex attention PyTorch compatibility versus a missing TensorFlow backend dependency; not the same bug." + "reason": "Compiled flex_attention on torch>=2.9 and missing TensorFlow backend installation are unrelated." }, { "left": "issue:43976", "right": "issue:45405", "accept": false, - "reason": "Python support mismatch versus an over-bumped PEFT minimum version; distinct dependency issues." + "reason": "Python version support and a PEFT minimum-version bump are separate dependency issues." }, { "left": "issue:44351", "right": "issue:45278", "accept": false, - "reason": "Importing a missing symbol and broad import errors after upgrade are both import-related, but the specific failure modes are different." + "reason": "45278 is a broad umbrella for many import errors after upgrade, but 44351 is a specific HybridCache import failure; not enough evidence they share one exact root cause." }, { "left": "issue:43479", "right": "issue:43525", "accept": false, - "reason": "Phi4Multimodal config default initialization versus Llama4Config missing pad_token_id; both config-related but unrelated fields and models." + "reason": "Different config bugs: default multimodal sub-config initialization versus missing pad_token_id on Llama4Config." } ] }, @@ -9070,9 +9066,9 @@ ], "soft_pairs": [ "issue:41950|issue:43650", + "issue:44496|issue:44877", "issue:44462|issue:45081", "issue:41950|issue:45020", - "issue:44496|issue:44877", "issue:44297|issue:44462", "issue:43550|issue:44964", "issue:44279|issue:45405", @@ -9084,10 +9080,10 @@ "issue:43650|issue:45290", "issue:44016|issue:44291", "issue:43723|issue:44162", - "issue:43441|issue:44683", "issue:43450|issue:44279", - "issue:43262|issue:44521", + "issue:43441|issue:44683", "issue:44466|issue:45276", + "issue:43262|issue:44521", "issue:42994|issue:43317", "issue:44261|issue:44263", "issue:43065|issue:44898", @@ -9099,12 +9095,12 @@ "issue:43673|issue:43761", "issue:44987|issue:45542" ], - "prepared_review_unit_hash": "4a65afa3f32885f487695668eea4a7a6fca8755546cee45384afc5c4b5a8703f", + "prepared_review_unit_hash": "2c0faa0868625c96c14ea576dc70a798fbb0720820a4f0bd530d6338b3b9a9f1", "component_budget": { "node_count": 367, "item_count": 367, "soft_pair_count": 1370, - "serialized_chars": 706987, + "serialized_chars": 706986, "estimated_input_tokens": 176747, "estimated_eval_tokens": 353750 }, @@ -9112,7 +9108,7 @@ "node_count": 48, "item_count": 48, "soft_pair_count": 29, - "serialized_chars": 36344, + "serialized_chars": 36343, "estimated_input_tokens": 9086, "estimated_eval_tokens": 18428 }, @@ -9121,186 +9117,186 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Mostly unrelated issues bundled together with a few superficial overlaps around recent Transformers regressions, tokenization, and model loading. None of the soft pairs look like the same underlying bug/change, so I would not merge any of them as duplicates. 45020 is the broadest umbrella issue, but the cluster is too heterogeneous for a clean single canonical sink.", - "confidence": 0.96, - "canonical_issue_reason": "issue:45020 is the broadest umbrella for the shared \u2018latest transformers breaks loading/custom models\u2019 theme, but only as a loose anchor; the cluster still spans many distinct bugs.", + "summary": "Mostly unrelated issues with superficial overlap around Transformers v5 regressions, tokenizers, processors, and model loading. None of the soft pairs look like the same underlying bug/change.", + "confidence": 0.91, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "issue:45020 is the most central and general issue in this set, so it is the best available triage anchor even though it is not a true duplicate target for the rest.", + "best_issue_reason": null, "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:41950", "right": "issue:43650", "accept": false, - "reason": "Different problems: video pipeline processor selection vs a nonsensical placeholder issue title." + "reason": "Video pipeline processor lookup is unrelated to the vague \"ADD THE DATA\" report." + }, + { + "left": "issue:44496", + "right": "issue:44877", + "accept": false, + "reason": "Both are loading/config issues, but they fail for different reasons and different models." }, { "left": "issue:44462", "right": "issue:45081", "accept": false, - "reason": "Both involve tokenizers, but one ignores repo tokenizer.json while the other crashes in a Mistral regex patch path." + "reason": "AutoTokenizer ignoring repo tokenizer.json is a different bug from Mistral regex patch crashing on backend_tokenizer." }, { "left": "issue:41950", "right": "issue:45020", "accept": false, - "reason": "One is a specific video-classification processor bug; the other is a broad remote_code loading regression, not the same failure." - }, - { - "left": "issue:44496", - "right": "issue:44877", - "accept": false, - "reason": "Both are model/config load failures, but they affect different models and have different root causes (missing model_type vs strict config handling)." + "reason": "A video-classification processor bug is not the same as a broad remote_code compatibility regression." }, { "left": "issue:44297", "right": "issue:44462", "accept": false, - "reason": "Related tokenizer serialization/loading area, but save_pretrained metadata mismatch is distinct from AutoTokenizer ignoring tokenizer.json." + "reason": "Tokenizer save metadata mismatch and tokenizer.json precedence are distinct serialization vs loading issues." }, { "left": "issue:43550", "right": "issue:44964", "accept": false, - "reason": "Different code paths and models: torch.compile SDPA on Bamba vs loading Phi-4 multimodal instruct." + "reason": "Different models and failure modes: torch.compile/SDPA vs model loading for Phi-4 multimodal." }, { "left": "issue:44279", "right": "issue:45405", "accept": false, - "reason": "Both mention dependencies, but they are unrelated dependency/version issues with different packages and symptoms." + "reason": "Unrelated dependency problems: transformers dependency issue vs an unreleased PEFT version bump." }, { "left": "issue:43299", "right": "issue:43854", "accept": false, - "reason": "Both are latest-version loading regressions, but one is Qwen3VL MoE model loading and the other is a sentence-transformers import NameError." + "reason": "Both are model-loading problems, but they affect different models and code paths." }, { "left": "issue:43295", "right": "issue:44779", "accept": false, - "reason": "Both are v5 regressions in tokenizer-related behavior, but one is processor.tokenizer access/images handling and the other is Deepseek tokenization output drift." + "reason": "Processor.tokenizer regression is not the same as Deepseek tokenizer output drift." }, { "left": "issue:44462", "right": "issue:44779", "accept": false, - "reason": "Tokenizer-related, but ignoring tokenizer.json and Deepseek tokenization corruption are different bugs." + "reason": "Repo tokenizer.json selection and Deepseek tokenization changes are different tokenizer bugs." }, { "left": "issue:39692", "right": "issue:43901", "accept": false, - "reason": "Both are docs/example issues, but one is SigLIP2 example errors and the other is a pipeline docs mismatch." + "reason": "Both mention docs, but one is a broken SigLIP2 example and the other is stale pipeline documentation." }, { "left": "issue:38175", "right": "issue:44617", "accept": false, - "reason": "Different model families and failures: unexpected zero probabilities in SigLIP2 vs CUDA OOM in Sam3Video." + "reason": "Unexpected zero probabilities in SigLIP2 is unrelated to Sam3Video CUDA OOM." }, { "left": "issue:43650", "right": "issue:45290", "accept": false, - "reason": "No substantive overlap; placeholder title vs apply_chat_template crash with tool calls." + "reason": "\"ADD THE DATA\" is not a substantive duplicate target for a chat-template tool-call crash." }, { "left": "issue:44016", "right": "issue:44291", "accept": false, - "reason": "Notebook syntax error and init_empty_weights argument mismatch are unrelated." + "reason": "Notebook syntax error and init_empty_weights TypeError are unrelated failures." }, { "left": "issue:43723", "right": "issue:44162", "accept": false, - "reason": "Both are loading/compatibility complaints, but AutoTokenizer.from_pretrained and ESM2 breakage are distinct issues." - }, - { - "left": "issue:43441", - "right": "issue:44683", - "accept": false, - "reason": "Both touch attention paths, but Ministral FlashAttention failure and compiled flex_attention on torch>=2.9 are separate problems." + "reason": "Tokenizer loading regression and ESM2 breakage are different model/component issues." }, { "left": "issue:43450", "right": "issue:44279", "accept": false, - "reason": "Video processor batch-shape bug vs dependency issue; no shared code-path." + "reason": "Batched video processor shape bugs are unrelated to a generic dependency issue." }, { - "left": "issue:43262", - "right": "issue:44521", + "left": "issue:43441", + "right": "issue:44683", "accept": false, - "reason": "Both concern chat_template/multimodal processing, but default audio sampling rate and zero assistant masks are different bugs." + "reason": "FlashAttention failure and compiled flex_attention failure are different kernels and triggers." }, { "left": "issue:44466", "right": "issue:45276", "accept": false, - "reason": "Different model serialization behaviors: device-dependent lm_head.weight saving vs resize_token_embeddings not updating Gemma4 embeddings." + "reason": "lm_head serialization inconsistency and resize_token_embeddings propagation are different weight-handling bugs." + }, + { + "left": "issue:43262", + "right": "issue:44521", + "accept": false, + "reason": "Audio chat-template sampling-rate default and multimodal assistant_mask generation are different failures in the same API." }, { "left": "issue:42994", "right": "issue:43317", "accept": false, - "reason": "Both are quantization/offload related, but one is saving a quantized model and the other is loading a dequantized model with device_map=auto." + "reason": "Saving quantized models and loading dequantized offloaded models are separate quantization workflows." }, { "left": "issue:44261", "right": "issue:44263", "accept": false, - "reason": "Completely different failures: missing rms_norm_eps in MLA q_a_layernorm vs torch.split return handling in GlmMoeDsaIndexer." + "reason": "Different low-level bugs: missing config.rms_norm_eps vs torch.split return-value handling." }, { "left": "issue:43065", "right": "issue:44898", "accept": false, - "reason": "Different vision-model bugs in different architectures: dummy Conv2d in Sam3PixelDecoder vs Perceiver non-default resolution failure." + "reason": "Dummy Conv2d in Sam3PixelDecoder and Perceiver resize/interpolation failure are unrelated vision-model issues." }, { "left": "issue:44623", "right": "issue:45397", "accept": false, - "reason": "processor.save_pretrained missing files and gemma-4 zero3 from_pretrained are unrelated." + "reason": "Processor.save_pretrained missing files is not the same as a zero3 from_pretrained loading failure." }, { "left": "issue:43122", "right": "issue:43650", "accept": false, - "reason": "Tokenization regression vs placeholder issue; no real overlap." + "reason": "Tokenizer version drift is unrelated to the non-specific \"ADD THE DATA\" issue." }, { "left": "issue:44623", "right": "issue:44749", "accept": false, - "reason": "Different areas and symptoms: missing processor save files vs slow filtering after upgrade." + "reason": "Missing save_pretrained files and a performance regression after upgrade are different classes of bugs." }, { "left": "issue:43784", "right": "issue:45542", "accept": false, - "reason": "sentence-transformers import NameError and tensorboard/TensorFlow backend selection are unrelated." + "reason": "Importing sentence-transformers breaks for nn, while the other issue is about tensorboard/TensorFlow backend detection." }, { "left": "issue:44671", "right": "issue:44779", "accept": false, - "reason": "Both are v5 output regressions, but CamemBERT masked LM predictions and Deepseek tokenizer results are different model-specific bugs." + "reason": "CamemBERT masked-LM prediction regression is a different bug from Deepseek tokenizer output changes." }, { "left": "issue:43673", "right": "issue:43761", "accept": false, - "reason": "Generation cache handling during chunked_prefill is unrelated to CLIPVisionModel hidden_states returning None." + "reason": "Chunked-prefill cache regression and CLIPVisionModel hidden_states regression are separate v5 breakages." }, { "left": "issue:44987", "right": "issue:45542", "accept": false, - "reason": "Model loading failure for physical-intelligence/fast vs tensorboard-only backend error; no shared underlying bug." + "reason": "Model-loading failure for physical-intelligence/fast is unrelated to the tensorboard/TensorFlow backend issue." } ] }, @@ -9374,16 +9370,16 @@ ], "soft_pairs": [ "issue:43901|issue:43994", - "issue:44297|issue:45397", "issue:42175|issue:43976", + "issue:44297|issue:45397", "issue:45561|issue:45563", "issue:43475|issue:44016", "issue:43824|issue:43976", "issue:43352|issue:45250", "issue:43525|issue:43881", "issue:44589|issue:45362", - "issue:44188|issue:44336", "issue:43295|issue:44617", + "issue:44188|issue:44336", "issue:43901|issue:44246", "issue:41628|issue:43976", "issue:44295|issue:45397", @@ -9403,12 +9399,12 @@ "issue:44162|issue:45200", "issue:45292|issue:45584" ], - "prepared_review_unit_hash": "e9cbd6213fbfb1ff98cd6200431511bb1a20d81e12beb389add6d05b85d8d583", + "prepared_review_unit_hash": "35c0d608082a1ee9f0f70fb8748080b4f90df9ed54ac335cc012f9d412ca0ce5", "component_budget": { "node_count": 367, "item_count": 367, "soft_pair_count": 1370, - "serialized_chars": 706987, + "serialized_chars": 706986, "estimated_input_tokens": 176747, "estimated_eval_tokens": 353750 }, @@ -9416,195 +9412,195 @@ "node_count": 47, "item_count": 47, "soft_pair_count": 29, - "serialized_chars": 35484, - "estimated_input_tokens": 8871, - "estimated_eval_tokens": 17998 + "serialized_chars": 35485, + "estimated_input_tokens": 8872, + "estimated_eval_tokens": 18000 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a loose collection of Transformers bug reports, with a noticeable subgroup of import/loading regressions. The most representative issue is the Qwen2_5_VL import failure (#43824); none of the soft pairs look like the same concrete bug or change.", - "confidence": 0.34, - "canonical_issue_reason": "#43824 is the most central and broadly useful representative here: it has the strongest discussion/inbound-reference signal and is a clear import/export regression affecting a core model entrypoint.", + "summary": "This cluster is a loose mix of unrelated Transformers issues. The soft-similarity pairs are mostly same-area but different bugs, regressions, or docs/runtime symptoms, so none should be merged as duplicates.", + "confidence": 0.94, + "canonical_issue_reason": "If forced to pick one anchor, issue 43299 is the clearest concrete regression report and the most representative of the Qwen3-MoE loading-breakage subgroup; there is no single true canonical issue for the whole set.", "canonical_pr_reason": null, - "best_issue_reason": "#43824 best fits the cluster as a canonical issue because it is a direct, user-facing import failure with broad relevance and stronger evidence than the more niche reports.", + "best_issue_reason": "Issue 43299 is the best overall representative: it is specific, reproducible, and describes a real model-loading regression rather than a generic symptom or unrelated API/docs problem.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43901", "right": "issue:43994", "accept": false, - "reason": "Docs mismatch for text classification vs a SIGLIP2 behavior bug; different code paths and symptoms." + "reason": "One is a docs/behavior mismatch for text classification; the other is a model quality/loading problem for SigLIP2. Different failure modes." }, { - "left": "issue:44297", - "right": "issue:45397", + "left": "issue:42175", + "right": "issue:43976", "accept": false, - "reason": "Tokenizer save metadata bug is unrelated to Gemma-4 Zero3 from_pretrained loading." + "reason": "TensorFlow backend omission from a pip extra is unrelated to the Python-version support regression." }, { - "left": "issue:42175", - "right": "issue:43976", + "left": "issue:44297", + "right": "issue:45397", "accept": false, - "reason": "TensorFlow backend packaging and Python version support are different release/configuration problems." + "reason": "Tokenizer class metadata mismatch vs gemma-4 ZeRO3 from_pretrained failure; different code paths and symptoms." }, { "left": "issue:45561", "right": "issue:45563", "accept": false, - "reason": "xdist file race and stale generation warning are unrelated test/runtime issues." + "reason": "A pytest-xdist file race is unrelated to a stale generation warning in paged generate()." }, { "left": "issue:43475", "right": "issue:44016", "accept": false, - "reason": "SAM 3 Video attribute error and notebook syntax error are not the same bug." + "reason": "SAM3 video attribute bug and a notebook syntax error are entirely separate issues." }, { "left": "issue:43824", "right": "issue:43976", "accept": false, - "reason": "Missing export for a model class vs Python 3.9/3.10 compatibility; unrelated failures." + "reason": "Import export failure for a Qwen2.5-VL symbol is not the same as a Python compatibility issue." }, { "left": "issue:43352", "right": "issue:45250", "accept": false, - "reason": "Generic Flash Attention topic versus a specific model not yet supporting FA2; too broad to merge as one bug." + "reason": "A specific model lacking Flash Attention 2.0 support is not the same as the general Flash Attention 2.0 topic." }, { "left": "issue:43525", "right": "issue:43881", "accept": false, - "reason": "Llama4Config missing pad_token_id and GLM-4v loading failure are distinct model-specific regressions." + "reason": "Missing pad_token_id on Llama4Config and glm-4v-9b loading failure are different bugs." }, { "left": "issue:44589", "right": "issue:45362", "accept": false, - "reason": "Float8 storage deserialization error is different from Qwen3.5 chat crash." + "reason": "Float8 storage deserialization failure and Qwen3.5 chat crash are unrelated runtime errors." }, { - "left": "issue:44188", - "right": "issue:44336", + "left": "issue:43295", + "right": "issue:44617", "accept": false, - "reason": "torch.compile attention-kernel divergence is unrelated to ANSI code output in loading_report." + "reason": "Processor/tokenizer regression and SAM3Video OOM are different problems." }, { - "left": "issue:43295", - "right": "issue:44617", + "left": "issue:44188", + "right": "issue:44336", "accept": false, - "reason": "Processor/tokenizer regression is unrelated to Sam3Video CUDA OOM." + "reason": "Torch.compile attention-kernel divergence has nothing to do with ANSI codes in loading_report." }, { "left": "issue:43901", "right": "issue:44246", "accept": false, - "reason": "Pipeline docs mismatch and slow import are not the same underlying issue." + "reason": "Docs mismatch for return_all_scores is unrelated to intermittent slow imports." }, { "left": "issue:41628", "right": "issue:43976", "accept": false, - "reason": "Missing AutoImageProcessor export and Python version incompatibility are separate issues." + "reason": "AutoImageProcessor import failure and Python version support are separate import/runtime issues." }, { "left": "issue:44295", "right": "issue:45397", "accept": false, - "reason": "position_ids buffer access bug is unrelated to Gemma-4 Zero3 loading." + "reason": "position_ids buffer access error is unrelated to gemma-4 zero3 from_pretrained loading." }, { "left": "issue:42491", "right": "issue:43299", "accept": false, - "reason": "Qwen3-MoE LoRA compatibility across versions and Qwen3VL-MoE loading are related by model family only, not the same concrete bug." + "reason": "Both involve Qwen3-MoE family breakage, but one is a LoRA/hf4.x-to-hf5.x compatibility report and the other is a specific Qwen3VL-MoE loading regression; not the same concrete bug." }, { "left": "issue:41950", "right": "issue:44220", "accept": false, - "reason": "Video-classification pipeline image-processor lookup and fbank extraction are different pipeline bugs." + "reason": "Video-classification pipeline processor lookup and torch fbank feature extraction are different audio/video preprocessing bugs." }, { "left": "issue:44336", "right": "issue:44373", "accept": false, - "reason": "Unwanted ANSI codes and a wrong docstring are unrelated." + "reason": "Terminal ANSI output leakage and a wrong docstring for position_ids are unrelated." }, { "left": "issue:44462", "right": "issue:45356", "accept": false, - "reason": "AutoTokenizer ignoring tokenizer.json is not the same as Kimi-K2.5 codec/regex regression." + "reason": "AutoTokenizer ignoring tokenizer.json and Kimi tokenizer codec regression are both tokenizer issues, but not the same underlying bug." }, { "left": "issue:44246", "right": "issue:44496", "accept": false, - "reason": "Slow import performance and unrecognized model/config errors do not share the same code path." + "reason": "Slow import behavior and unrecognized model config are unrelated." }, { "left": "issue:44661", "right": "issue:45542", "accept": false, - "reason": "TOKENIZER_MAPPING_NAMES failure is unrelated to TensorFlow backend installation issues." + "reason": "Tokenizer mapping validation and missing TensorFlow backend handling are separate packaging/runtime issues." }, { "left": "issue:45003", "right": "issue:45561", "accept": false, - "reason": "sys.modules access safety bug and xdist captured_info race are separate concurrency problems." + "reason": "Unsafe sys.modules access in modeling_utils is unrelated to a captured_info.txt xdist race." }, { "left": "issue:43531", "right": "issue:45362", "accept": false, - "reason": "Qwen3-MoE sliding_window behavior and Qwen3.5 chat crash are different runtime failures." + "reason": "Qwen3-MoE sliding_window behavior and Qwen3.5 chat crashes are not the same code-path problem." }, { "left": "issue:43526", "right": "issue:45003", "accept": false, - "reason": "BeitImageProcessorFast label reduction bug is unrelated to sys.modules access in modeling_utils." + "reason": "BEiT image processor label reduction bug and modeling_utils sys.modules access are unrelated." }, { "left": "issue:36246", "right": "issue:42757", "accept": false, - "reason": "Transformers model import error and huggingface_hub import error are different packages and failures." + "reason": "Both are import errors, but they involve different packages and different missing symbols." }, { "left": "issue:36246", "right": "issue:44351", "accept": false, - "reason": "Both are import errors, but for different symbols and likely different export paths; not enough evidence of one concrete bug." + "reason": "Qwen2.5-VL image processor import failure is not the same as HybridCache export failure." }, { "left": "issue:44336", "right": "issue:44485", "accept": false, - "reason": "Loading-report ANSI output and GLM-5 RoPE implementation are unrelated." + "reason": "ANSI output leakage and GLM-5 RoPE implementation discussion are unrelated." }, { "left": "issue:43673", "right": "issue:44561", "accept": false, - "reason": "Cache missing during chunked prefill and removal of is_torch_fx_available affect different features." + "reason": "Chunked-prefill cache handling and removal of is_torch_fx_available affect different subsystems and code paths." }, { "left": "issue:44162", "right": "issue:45200", "accept": false, - "reason": "ESM2 breakage and Gemma-4 mm_token_type_ids defaulting are unrelated model issues." + "reason": "ESM2 breakage and Gemma 4 mm_token_type_ids defaults are separate model-specific bugs." }, { "left": "issue:45292", "right": "issue:45584", "accept": false, - "reason": "resize_token_embeddings/output_embeddings mismatch and Whisper empty transcription handling are different generation/model-editing bugs." + "reason": "resize_token_embeddings not updating output embeddings is unrelated to Whisper empty-transcription generation failure." } ] }, @@ -9684,15 +9680,15 @@ "issue:44779|issue:45292", "issue:44877|issue:45375", "issue:43441|issue:43540", - "issue:43066|issue:44779", "issue:36246|issue:43596", + "issue:43066|issue:44779", "issue:44987|issue:45459", "issue:42886|issue:45229", "issue:41628|issue:41950", "issue:44995|issue:45200", "issue:43901|issue:45003", - "issue:43065|issue:43408", "issue:43257|issue:45137", + "issue:43065|issue:43408", "issue:43232|issue:44336", "issue:43295|issue:45292", "issue:44496|issue:45003", @@ -9704,15 +9700,15 @@ "issue:44368|issue:44977", "issue:44368|issue:45563", "issue:44933|issue:45459", - "issue:43673|issue:44734", - "issue:43441|issue:43606" + "issue:43441|issue:43606", + "issue:43673|issue:44734" ], - "prepared_review_unit_hash": "75087552abbf279c70128ef0cb358dc578a0014aa79d261e135747e01ba2ebdc", + "prepared_review_unit_hash": "0d8d9c917531b1966a671a19eec80590d91b0b0db6208db571e901442a629241", "component_budget": { "node_count": 367, "item_count": 367, "soft_pair_count": 1370, - "serialized_chars": 706987, + "serialized_chars": 706986, "estimated_input_tokens": 176747, "estimated_eval_tokens": 353750 }, @@ -9720,189 +9716,189 @@ "node_count": 48, "item_count": 48, "soft_pair_count": 28, - "serialized_chars": 36215, - "estimated_input_tokens": 9054, - "estimated_eval_tokens": 18364 + "serialized_chars": 36206, + "estimated_input_tokens": 9052, + "estimated_eval_tokens": 18360 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly heterogeneous: import/export regressions, tokenizer/config issues, generation bugs, and model-specific failures that do not look like true duplicates. I rejected all soft edges under a strict same-underlying-bug standard.", - "confidence": 0.24, - "canonical_issue_reason": "issue:41628 is the closest thing to a cluster representative because it is a direct, broadly user-facing import/export failure, but the surrounding issues are still mostly unrelated.", + "summary": "The cluster is mostly heterogeneous; the soft pairs look like superficially similar Transformers issues but not the same underlying bug or fix. I would anchor on the broad import-performance report as the best issue representative, but I would not merge any of the suggested soft edges.", + "confidence": 0.57, + "canonical_issue_reason": "issue:44246 is the broadest, most general user-facing report in the set and best matches the cluster\u2019s core import/runtime theme, even though it is not a duplicate of the other items.", "canonical_pr_reason": null, - "best_issue_reason": "issue:41628 is the most generic and broadly impactful issue in the set, so it is the best representative even though it is not a true duplicate hub.", + "best_issue_reason": "issue:44246 has the widest scope and strongest general relevance among the issues here; it is the most plausible anchor for a mixed cluster of Transformers import/runtime complaints.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44242", "right": "issue:44857", "accept": false, - "reason": "Different bugs: router loss handling vs AMP/CUDA crash in a loss class." + "reason": "Different bugs: one is router/load-balancing loss behavior, the other is AMP/CUDA crash in LwDetrImageLoss." }, { "left": "issue:44246", "right": "issue:44871", "accept": false, - "reason": "Import performance complaint vs Gemma EOS-token config inconsistency." + "reason": "Import slowdown vs Gemma-3 EOS config inconsistency; unrelated failure modes." }, { "left": "issue:44295", "right": "issue:44749", "accept": false, - "reason": "Buffer/position_ids access bug vs a preprocessing slowdown regression." + "reason": "Position_ids buffer handling is not the same as the reported data filtering slowdown." }, { "left": "issue:44779", "right": "issue:45292", "accept": false, - "reason": "DeepSeek tokenizer correctness vs resize_token_embeddings not updating output embeddings." + "reason": "Tokenizer output regression vs resize_token_embeddings not updating output embeddings; different code paths." }, { "left": "issue:44877", "right": "issue:45375", "accept": false, - "reason": "Strict config loading for one model family vs a missing field in a different config class." + "reason": "Both involve config handling, but one is strict config loading and the other is a missing model field; not the same bug." }, { "left": "issue:43441", "right": "issue:43540", "accept": false, - "reason": "Different code paths: FlashAttention failure in Ministral-3 vs video input validation in Qwen3OmniMoe." + "reason": "FlashAttention failure for one model family vs video-input ValueError in another; unrelated." }, { - "left": "issue:43066", - "right": "issue:44779", + "left": "issue:36246", + "right": "issue:43596", "accept": false, - "reason": "Both are tokenizer-related, but they describe different model-specific failures and different symptoms." + "reason": "ImportError for Qwen2.5-VL image processor vs DeepSpeed ZeRO3/BertModel index error; no shared code-path." }, { - "left": "issue:36246", - "right": "issue:43596", + "left": "issue:43066", + "right": "issue:44779", "accept": false, - "reason": "Missing Qwen2.5-VL import vs DeepSpeed ZeRO3/BertModel index error." + "reason": "Tokenizer decoder-type warning/regression vs incorrect DeepSeek tokenizer outputs; same area, but not the same underlying defect." }, { "left": "issue:44987", "right": "issue:45459", "accept": false, - "reason": "Model loading failure vs tokenizer error masking when protobuf is absent." + "reason": "Model loading failure for physical-intelligence/fast is unrelated to protobuf exception masking in tokenizer decode errors." }, { "left": "issue:42886", "right": "issue:45229", "accept": false, - "reason": "Offline cache tokenizer load bug vs Gemma4 multi-GPU OOM are unrelated." + "reason": "Offline cache tokenizer loading vs Gemma4 multi-GPU OOM; completely different problems." }, { "left": "issue:41628", "right": "issue:41950", "accept": false, - "reason": "Top-level import failure vs a pipeline selecting the wrong processor type." + "reason": "Importing AutoImageProcessor vs video-classification pipeline choosing image processors; related subsystem, but not the same bug." }, { "left": "issue:44995", "right": "issue:45200", "accept": false, - "reason": "GLM-MoE stale cache crash vs Gemma4 missing mm_token_type_ids default." + "reason": "Stale indexer cache crash vs missing mm_token_type_ids defaults; different model behaviors." }, { "left": "issue:43901", "right": "issue:45003", "accept": false, - "reason": "Docs mismatch for a pipeline flag vs unsafe sys.modules access in modeling_utils." + "reason": "Pipeline docs mismatch vs unsafe sys.modules access in modeling_utils; unrelated." }, { - "left": "issue:43065", - "right": "issue:43408", + "left": "issue:43257", + "right": "issue:45137", "accept": false, - "reason": "Dummy Conv2d in SAM3 pixel decoder vs a model-type warning between sam3_video and sam3_tracker." + "reason": "Qwen3 MoE weight conversion issue vs ZeRO3 deque IndexError; different failure mechanisms." }, { - "left": "issue:43257", - "right": "issue:45137", + "left": "issue:43065", + "right": "issue:43408", "accept": false, - "reason": "Qwen3 MoE weight conversion bug vs a separate ZeRO3 deque/pop crash." + "reason": "Dummy Conv2d implementation issue vs model-type warning between sam3_video and sam3_tracker; not the same underlying defect." }, { "left": "issue:43232", "right": "issue:44336", "accept": false, - "reason": "Generation kwargs update bug vs ANSI escape codes in loading_report." + "reason": "Generation kwargs/cache issue vs ANSI codes in loading_report; unrelated." }, { "left": "issue:43295", "right": "issue:45292", "accept": false, - "reason": "Processor.tokenizer/images regression vs resize_token_embeddings not affecting output embeddings." + "reason": "Processor/tokenizer regression with images vs resize_token_embeddings not updating output embeddings; different API surfaces." }, { "left": "issue:44496", "right": "issue:45003", "accept": false, - "reason": "Unrecognized model/config loading failure vs unsafe sys.modules access." + "reason": "Unrecognized model/config loading error vs unsafe sys.modules access; not the same bug." }, { "left": "issue:44060", "right": "issue:44077", "accept": false, - "reason": "Different warnings in different subsystems: tied-weights warning vs optional post_init restriction." + "reason": "Tied-weights warning bug vs optional post_init restriction for patchtsmixer; unrelated." }, { "left": "issue:43262", "right": "issue:44246", "accept": false, - "reason": "Audio chat-template sample-rate default bug vs import latency complaint." + "reason": "Audio chat-template sample-rate default vs import time regression; different subsystems." }, { "left": "issue:43475", "right": "issue:43540", "accept": false, - "reason": "SAM3 vision encoder output attribute mismatch vs Qwen3OmniMoe video processing ValueError." + "reason": "Missing fpn_position_embeddings attribute vs Qwen3OmniMoe video-input processing error; not the same code path." }, { "left": "issue:43653", "right": "issue:44162", "accept": false, - "reason": "BigBirdTokenizer special-token registration bug vs a broad ESM2 regression report." + "reason": "BigBirdTokenizer special-token registration vs ESM2 model breakage; not a duplicate." }, { "left": "issue:44466", "right": "issue:45335", "accept": false, - "reason": "Both involve embedding/weight tying, but the concrete failure modes and affected code paths differ." + "reason": "Serialization inconsistency for tied weights vs resize_token_embeddings not affecting decoder.embed_tokens; related theme, but distinct bugs." }, { "left": "issue:44368", "right": "issue:44977", "accept": false, - "reason": "Qwen3.5 tied-embedding warning vs flash-attention generation failure are different issues." + "reason": "Tie_word_embeddings warning during LoRA fine-tuning vs flash-attention generation failure; different problems." }, { "left": "issue:44368", "right": "issue:45563", "accept": false, - "reason": "Different generation warnings: tied embeddings vs stale num_return_sequences warning." + "reason": "Tie_word_embeddings warning vs stale num_return_sequences warning in paged generate; unrelated." }, { "left": "issue:44933", "right": "issue:45459", "accept": false, - "reason": "Missing import export vs protobuf-related tokenizer error masking." + "reason": "Missing image_utils import vs protobuf decode-error masking in tokenizers; no overlap in root cause." }, { - "left": "issue:43673", - "right": "issue:44734", + "left": "issue:43441", + "right": "issue:43606", "accept": false, - "reason": "Both mention cache, but one is missing GenerationMixin cache state and the other is wrong KV-cache indexing in serve." + "reason": "FlashAttention model failure vs CPU offload device mismatch; different runtime issues." }, { - "left": "issue:43441", - "right": "issue:43606", + "left": "issue:43673", + "right": "issue:44734", "accept": false, - "reason": "Different failures: FlashAttention on Ministral-3 vs CPU offload device mismatch on bark-small." + "reason": "Chunked-prefill cache missing in GenerationMixin vs server KV-cache continuation tensor indexing; both cache-related but not the same fixable defect." } ] }, @@ -9983,8 +9979,8 @@ "issue:43673|issue:45278", "issue:43646|issue:44162", "issue:44361|issue:44749", - "issue:42673|issue:43704", "issue:28282|issue:30990", + "issue:42673|issue:43704", "issue:44749|issue:45310", "issue:45230|issue:45397", "issue:44038|issue:44623", @@ -9993,28 +9989,28 @@ "issue:43701|issue:44749", "issue:43475|issue:45381", "issue:42371|issue:45092", - "issue:45375|issue:45478", "issue:45081|issue:45103", + "issue:45375|issue:45478", "issue:41628|issue:44933", "issue:45310|issue:45375", - "issue:42222|issue:45200", "issue:42994|issue:43525", + "issue:42222|issue:45200", "issue:43475|issue:44661", "issue:43873|issue:44945", - "issue:42831|issue:44222", "issue:45362|issue:45375", + "issue:42831|issue:44222", "issue:45310|issue:45397", "issue:43232|issue:44188", "issue:44222|issue:45310", "issue:41628|issue:44351", "issue:43504|issue:45561" ], - "prepared_review_unit_hash": "a63f43076ea21ddb4f65a6a0a22932e6d45132467c43a048b2e5d52a26a770c0", + "prepared_review_unit_hash": "e16a39bf9904bc734a02553e028b64a3998b12d9ccfc47dc3e3e57e560c9f3a4", "component_budget": { "node_count": 367, "item_count": 367, "soft_pair_count": 1370, - "serialized_chars": 706987, + "serialized_chars": 706986, "estimated_input_tokens": 176747, "estimated_eval_tokens": 353750 }, @@ -10022,7 +10018,7 @@ "node_count": 48, "item_count": 48, "soft_pair_count": 32, - "serialized_chars": 36997, + "serialized_chars": 36998, "estimated_input_tokens": 9250, "estimated_eval_tokens": 18756 }, @@ -10031,210 +10027,210 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Only one soft edge is a true duplicate: the two Qwen3ForCausalLM VRAM-leak reports. The rest are topical similarities but describe different bugs or code paths, so they should remain separate.", - "confidence": 0.83, - "canonical_issue_reason": "issue:42673 is the better canonical for the duplicate VRAM-leak report because it is the older issue and has more discussion activity than issue:43704.", + "summary": "The cluster is mostly a mix of Transformers v5 import/loading/config regressions plus a few model-specific bugs. There are two clear duplicate pairs in the packet: issue:42673/issue:43704 and issue:45310/issue:45478; the rest of the soft links are topical at best, not true duplicates.", + "confidence": 0.81, + "canonical_issue_reason": "issue:45310 is a clean exact duplicate pair with issue:45478 and is a representative concrete v5 from_pretrained regression within the cluster.", "canonical_pr_reason": null, - "best_issue_reason": "issue:42673 is the best representative issue because it matches the accepted duplicate pair and has the stronger discussion signal in that pair.", + "best_issue_reason": "issue:45278 is the broadest umbrella for the cluster\u2019s v5-era import/loading regression reports, even though it is vague.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44869", "right": "issue:45356", "accept": false, - "reason": "Different bugs: Whisper timestamp decoding crash vs Kimi tokenizer regression/regex warning." + "reason": "Different bugs: Whisper timestamp decode crash vs Kimi-K2.5 tokenizer codec/regex regression." }, { "left": "issue:43931", "right": "issue:44368", "accept": false, - "reason": "Different failure modes: Qwen3-VL shape mismatch vs Qwen3.5 LoRA warning about tie_word_embeddings." + "reason": "Different failures: Qwen3-VL weight-shape loading error vs Qwen3.5 LoRA warning about tie_word_embeddings." }, { "left": "issue:43232", "right": "issue:43673", "accept": false, - "reason": "Both concern generation internals, but one is sync_gpus/update_model_kwargs and the other is chunked_prefill cache handling." + "reason": "Both involve generation, but they affect different code paths: sync_gpus kwarg handling vs chunked_prefill cache regression." }, { "left": "issue:44297", "right": "issue:44623", "accept": false, - "reason": "Both are save_pretrained issues, but tokenizer metadata mismatch and processor file omission are distinct bugs." + "reason": "Both are save_pretrained-related, but tokenizer metadata mismatch and missing processor files are distinct bugs." }, { "left": "issue:43673", "right": "issue:45278", "accept": false, - "reason": "Generation cache/chunked_prefill issue vs a broad import-error report after upgrade; not the same code path." + "reason": "The latter is a broad import-error umbrella after upgrade, not the same concrete cache-missing bug." }, { "left": "issue:43646", "right": "issue:44162", "accept": false, - "reason": "Transformer v5 custom model initialization breakage vs ESM2-specific breakage; different underlying problems." + "reason": "Transformers 5 custom-model init breakage is broader than the specific ESM2 failure; not enough evidence they are the same defect." }, { "left": "issue:44361", "right": "issue:44749", "accept": false, - "reason": "Tokenizer AttributeError on tasks vs a performance regression after upgrade; unrelated symptoms." - }, - { - "left": "issue:42673", - "right": "issue:43704", - "accept": true, - "reason": "Same concrete bug: Qwen3ForCausalLM leaks VRAM when used in multiple dataloader threads; titles match exactly." + "reason": "Tokenizer AttributeError and a severe filtering slowdown are unrelated." }, { "left": "issue:28282", "right": "issue:30990", "accept": false, - "reason": "ImportError for missing PyTorch vs Sentence Transformers hanging while loading; not the same issue." + "reason": "PyTorch missing ImportError and a sentence-transformers loading hang are different issues." + }, + { + "left": "issue:42673", + "right": "issue:43704", + "accept": true, + "reason": "Same title and same concrete bug: Qwen3ForCausalLM VRAM leak when used across multiple dataloader threads." }, { "left": "issue:44749", "right": "issue:45310", "accept": false, - "reason": "Data-filtering slowdown after upgrade vs Qwen3.5-Moe from_pretrained error; different behaviors and code paths." + "reason": "Data filtering slowdown vs Qwen3.5-Moe from_pretrained error are unrelated." }, { "left": "issue:45230", "right": "issue:45397", "accept": false, - "reason": "Generic bug report vs gemma-4 zero3 from_pretrained failure; no specific shared bug." + "reason": "Vague bug report vs Gemma-4 ZeRO3 from_pretrained failure; no shared concrete bug." }, { "left": "issue:44038", "right": "issue:44623", "accept": false, - "reason": "Qwen3.5 MoE/transformers v5 bug vs processor save_pretrained missing files; different artifacts and failures." + "reason": "Qwen3-VL-Moe transformers v5 bug and processor.save_pretrained missing files are different problems." }, { "left": "issue:42994", "right": "issue:43482", "accept": false, - "reason": "Quantized model saving failure vs Qwen2.5-GGUF loading failure; save vs load, distinct paths." + "reason": "Quantized model saving failure and Qwen2.5-GGUF loading failure are distinct save/load bugs." }, { "left": "issue:43475", "right": "issue:43643", "accept": false, - "reason": "SAM 3 vision encoder output attribute error vs trust_remote_code missing fields in AutoConfig; unrelated." + "reason": "SAM 3 Video missing attribute and trust_remote_code missing fields are different model-loading issues." }, { "left": "issue:43701", "right": "issue:44749", "accept": false, - "reason": "resume_from_checkpoint key mismatch vs a training-data slowdown regression; different subsystems." + "reason": "Checkpoint resume key mismatch and data filtering slowdown are unrelated." }, { "left": "issue:43475", "right": "issue:45381", "accept": false, - "reason": "Different model families and bugs: SAM 3 attribute error vs qwen2.5-vl video position-id issue." + "reason": "Both are vision/video model issues, but the specific missing attribute and wrong vision_position_ids bugs are not the same." }, { "left": "issue:42371", "right": "issue:45092", "accept": false, - "reason": "TF32 API-setting warning vs InternVL2 meta-init incompatibility; unrelated changes." + "reason": "TF32 API settings warning and InternVL2 meta-init incompatibility are unrelated." }, { - "left": "issue:45375", - "right": "issue:45478", + "left": "issue:45081", + "right": "issue:45103", "accept": false, - "reason": "Both involve Qwen3.5 MoE, but one is a missing config field and the other is a from_pretrained error; not enough to merge." + "reason": "Tokenizer regex patch crash and auto_docstring annotation crash are separate code paths." }, { - "left": "issue:45081", - "right": "issue:45103", + "left": "issue:45375", + "right": "issue:45478", "accept": false, - "reason": "Tokenizer regex patch crash vs auto_docstring crash with future annotations; different components." + "reason": "Missing vision config field and Qwen3.5-Moe from_pretrained error are not the same concrete bug." }, { "left": "issue:41628", "right": "issue:44933", "accept": false, - "reason": "Different import errors in different modules; similar wording only." + "reason": "Different missing-import symptoms: AutoImageProcessor vs an image_utils import." }, { "left": "issue:45310", "right": "issue:45375", "accept": false, - "reason": "Related Qwen3.5 MoE area, but one is a loading error and the other is a missing config field; distinct bugs." + "reason": "Qwen3.5-Moe load error and missing vision config field are not the same issue." }, { - "left": "issue:42222", - "right": "issue:45200", + "left": "issue:42994", + "right": "issue:43525", "accept": false, - "reason": "vitpose model loading breakage vs Gemma 4 mm_token_type_ids defaulting; unrelated model families." + "reason": "Quantized saving failure and Llama4Config missing pad_token_id are different bugs." }, { - "left": "issue:42994", - "right": "issue:43525", + "left": "issue:42222", + "right": "issue:45200", "accept": false, - "reason": "Quantized model saving failure vs Llama4Config missing pad_token_id; different failures." + "reason": "Vitpose model breakage and Gemma 4 mm_token_type_ids handling are separate model-specific problems." }, { "left": "issue:43475", "right": "issue:44661", "accept": false, - "reason": "SAM 3 encoder output attribute error vs add-new-model-like tokenizer mapping failure; unrelated." + "reason": "SAM 3 Video attribute error and add-new-model-like/tokenizer-mapping failure are unrelated." }, { "left": "issue:43873", "right": "issue:44945", "accept": false, - "reason": "Quantization/offloading behavior vs incorrect pipeline-parallel LLM output; distinct runtime bugs." + "reason": "Quantization offloading bug and pipeline-parallel output corruption are different concerns." }, { - "left": "issue:42831", - "right": "issue:44222", + "left": "issue:45362", + "right": "issue:45375", "accept": false, - "reason": "FineGrainedFP8 accuracy issue vs FP8 save_pretrained moe failure; same theme only, not same bug." + "reason": "Qwen3.5-35B chat crash and missing Qwen3_5MoeVisionConfig field are different failures." }, { - "left": "issue:45362", - "right": "issue:45375", + "left": "issue:42831", + "right": "issue:44222", "accept": false, - "reason": "Qwen3.5 chat crash vs missing deepstack_visual_indexes in config; different defects." + "reason": "FineGrainedFP8 accuracy issue and FP8 save_pretrained bug are not the same code-path problem." }, { "left": "issue:45310", "right": "issue:45397", "accept": false, - "reason": "Qwen3.5 MoE loading error vs gemma-4 zero3 loading error; different models and paths." + "reason": "Same from_pretrained wording, but different models and different failure modes." }, { "left": "issue:43232", "right": "issue:44188", "accept": false, - "reason": "Generation kwargs/cache issue vs diverging attention kernels under torch.compile; separate problems." + "reason": "Generation kwarg update bug and torch.compile attention-kernel divergence are distinct." }, { "left": "issue:44222", "right": "issue:45310", "accept": false, - "reason": "FP8 save_pretrained bug vs Qwen3.5-Moe from_pretrained error; different operations." + "reason": "FP8 save_pretrained bug and Qwen3.5-Moe load error are unrelated." }, { "left": "issue:41628", "right": "issue:44351", "accept": false, - "reason": "Cannot import AutoImageProcessor vs cannot import HybridCache; both are import errors but not the same symbol or fix." + "reason": "Both are import errors, but they are for different symbols and likely different fixes." }, { "left": "issue:43504", "right": "issue:45561", "accept": false, - "reason": "BEiT legacy-field loading bug vs pytest-xdist race in testing utils; unrelated." + "reason": "Beit pretrained-load legacy-field issue and pytest-xdist captured_info.txt race are unrelated." } ] }, "evaluator_result": { "accept": false, - "feedback": "The duplicate call on issue:42673 \u2194 issue:43704 is grounded, but the canonical_issue_reason overstates the evidence by claiming \"more discussion and inbound references.\" The packet shows more discussion activity for 42673, but inbound_references are 0 for both issues. Please remove unsupported claims and keep the rationale limited to what the packet directly supports." + "feedback": "The summary is not fully grounded: the packet contains two clear duplicate pairs, not just one. In addition to issue:45310/issue:45478, issue:42673/issue:43704 is the same Qwen3ForCausalLM VRAM leak report with identical titles. The broader characterization of the cluster as mostly v5 import/loading/config regressions is reasonable, but the duplicate-count claim should be corrected." }, "evaluator_used": true, "retried": true, @@ -10305,14 +10301,14 @@ "issue:45561" ], "soft_pairs": [ - "issue:44496|issue:45561", "issue:42994|issue:44164", + "issue:44496|issue:45561", "issue:43526|issue:45561", "issue:44038|issue:45310", "issue:43232|issue:44792", "issue:43377|issue:44242", - "issue:43673|issue:44188", "issue:30990|issue:35141", + "issue:43673|issue:44188", "issue:36010|issue:41628", "issue:42898|issue:44568", "issue:43479|issue:44079", @@ -10337,12 +10333,12 @@ "issue:44222|issue:44297", "issue:42898|issue:43295" ], - "prepared_review_unit_hash": "44d73fb3c8140dfc6ccdb5f33c3ab9dc5f36681de161c2b6e1933615125245e1", + "prepared_review_unit_hash": "c8b871d9dfba5f683f9beb8e264bbe434410e43e33d226e303de8b76ae5d25c4", "component_budget": { "node_count": 367, "item_count": 367, "soft_pair_count": 1370, - "serialized_chars": 706987, + "serialized_chars": 706986, "estimated_input_tokens": 176747, "estimated_eval_tokens": 353750 }, @@ -10359,198 +10355,198 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is a grab-bag of unrelated Transformers issues (tokenizers, multimodal configs, generation, save/load, imports, and testing). None of the soft pairs look like the same underlying bug or change, so I reject all soft edges and do not see a single canonical issue/PR for the cluster.", - "confidence": 0.97, - "canonical_issue_reason": null, + "summary": "The cluster is very heterogeneous: most items are unrelated Transformers v5 regression reports, import/load failures, tokenizer quirks, and a few model-specific bugs. None of the proposed soft pairs look like the same underlying bug/change closely enough to merge, and there are no PRs in the set.", + "confidence": 0.33, + "canonical_issue_reason": "issue:42898 is the closest umbrella regression report because it captures a broad Transformers v5 behavior change, but the cluster overall mixes many different failure modes and subsystems.", "canonical_pr_reason": null, - "best_issue_reason": null, + "best_issue_reason": "issue:42898 is the best single issue candidate as the broadest v5-regression style report in the set, though it is not a true duplicate representative for most of the other items.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:44496", - "right": "issue:45561", + "left": "issue:42994", + "right": "issue:44164", "accept": false, - "reason": "Unrecognized model/config loading vs a pytest-xdist file race; unrelated failure modes." + "reason": "Both involve save/pretrained, but one is quantized model saving and the other is extra_state handling; different bugs and code paths." }, { - "left": "issue:42994", - "right": "issue:44164", + "left": "issue:44496", + "right": "issue:45561", "accept": false, - "reason": "Quantized model save failure vs extra_state handling in save/from_pretrained; different code paths." + "reason": "Unrelated: model config loading failure vs pytest-xdist race in testing utils." }, { "left": "issue:43526", "right": "issue:45561", "accept": false, - "reason": "BeitImageProcessorFast label reduction bug vs testing-utils race; unrelated." + "reason": "Different subsystems: BeitImageProcessorFast label reduction vs a test capture-file race." }, { "left": "issue:44038", "right": "issue:45310", "accept": false, - "reason": "Different Qwen MoE model-loading regressions in different versions/models; not the same concrete bug." + "reason": "Both mention Qwen/MoE and v5, but the specific models and failure modes differ; not clearly the same bug." }, { "left": "issue:43232", "right": "issue:44792", "accept": false, - "reason": "Generation kwargs sync_gpus bug vs janus image-generation test failure; different path." + "reason": "Generation kwargs/sync_gpus bug vs janus image-generation test failure; no shared concrete code path." }, { "left": "issue:43377", "right": "issue:44242", "accept": false, - "reason": "MIMI padding-mask discrepancy vs MoE load-balancing loss omission; unrelated." + "reason": "MIMI padding-mask behavior and load-balancing-loss gating are distinct issues." }, { - "left": "issue:43673", - "right": "issue:44188", + "left": "issue:30990", + "right": "issue:35141", "accept": false, - "reason": "Missing generation cache in chunked_prefill vs divergent attention kernels under torch.compile; different issues." + "reason": "Sentence-Transformers loading hang and embedding reinit on resize_token_embeddings are unrelated." }, { - "left": "issue:30990", - "right": "issue:35141", + "left": "issue:43673", + "right": "issue:44188", "accept": false, - "reason": "Sentence-Transformers loading stall vs embedding reinit after resize_token_embeddings; unrelated." + "reason": "Cache missing in chunked prefill and attention-kernel divergence under torch.compile are separate generation bugs." }, { "left": "issue:36010", "right": "issue:41628", "accept": false, - "reason": "Two separate import errors for different symbols; same general area but not the same bug." + "reason": "Both are import errors, but for different symbols and likely different export regressions." }, { "left": "issue:42898", "right": "issue:44568", "accept": false, - "reason": "Tokenization-space cleanup regression vs BOS/EOS not added for a specific tokenizer; different tokenizer behaviors." + "reason": "Tokenizer cleanup semantics and BOS/EOS insertion behavior are different tokenizer regressions." }, { "left": "issue:43479", "right": "issue:44079", "accept": false, - "reason": "Multimodal config default-init bug vs ModelOutput key assignment bug; unrelated." + "reason": "Config default initialization and ModelOutput key assignment are unrelated." }, { "left": "issue:43479", "right": "issue:44869", "accept": false, - "reason": "Config initialization issue vs Whisper timestamp decode crash; different subsystems." + "reason": "Phi4 multimodal config init and Whisper timestamp decoding crash are different bugs." }, { "left": "issue:44964", "right": "issue:45137", "accept": false, - "reason": "Phi-4 multimodal model loading vs DeepSpeed ZeRO3 deque error; unrelated." + "reason": "Model loading incompatibility vs DeepSpeed ZeRO3 deque underflow; no overlap in cause." }, { "left": "issue:43976", "right": "issue:44933", "accept": false, - "reason": "Python-version packaging incompatibility vs missing import from image_utils; not the same bug." + "reason": "Python version support issue and nonexistent image_utils import are unrelated." }, { "left": "issue:42175", "right": "issue:44933", "accept": false, - "reason": "Backend dependency packaging issue vs nonexistent image_utils import; unrelated." + "reason": "Missing TensorFlow extra and missing image_utils import are different problems." }, { "left": "issue:43784", "right": "issue:43824", "accept": false, - "reason": "Sentence-transformers import NameError vs missing Qwen2.5-VL class import; different import failures." + "reason": "Both are import errors, but for different packages/symbols and different root causes." }, { "left": "issue:44623", "right": "issue:45310", "accept": false, - "reason": "Processor save_pretrained missing files vs Qwen3.5 MoE from_pretrained error; different save/load paths." + "reason": "processor.save_pretrained missing files and Qwen3.5 MoE from_pretrained error are different save/load failures." }, { "left": "issue:43278", "right": "issue:45137", "accept": false, - "reason": "BF16-to-FP32 dtype drift in evaluation vs ZeRO3 deque pop error; unrelated." + "reason": "Embedding dtype drift in eval and a DeepSpeed deque error are unrelated." }, { "left": "issue:44188", "right": "issue:44373", "accept": false, - "reason": "Attention-kernel divergence vs wrong docstring; completely different severity and code." + "reason": "Attention-kernel branching and a wrong docstring are not the same issue." }, { "left": "issue:44060", "right": "issue:45125", "accept": false, - "reason": "Wrong tied-weights warning in Qwen3-Next vs missing _tp_plan in Qwen3_5Moe; different model bugs." + "reason": "Incorrect tied-weights warning and missing _tp_plan for tensor parallelism are different model-architecture issues." }, { "left": "issue:44734", "right": "issue:45103", "accept": false, - "reason": "Serving KV-cache indexing crash vs auto_docstring annotation handling crash; unrelated." + "reason": "Serve KV-cache indexing crash and auto_docstring annotations crash affect different code paths." }, { "left": "issue:42898", "right": "issue:45356", "accept": false, - "reason": "clean_up_tokenization_spaces behavior change vs Kimi tokenizer codec/warning regression; different tokenizer issues." + "reason": "Tokenizer cleanup behavior change and Kimi-K2.5 codec handling regression are different tokenizer bugs." }, { "left": "issue:43723", "right": "issue:44568", "accept": false, - "reason": "Tokenizer loading failure vs add_special_tokens BOS/EOS regression; different failure modes." + "reason": "Tokenizer loading failure and add_special_tokens BOS/EOS behavior are not the same bug." }, { "left": "issue:42898", "right": "issue:44779", "accept": false, - "reason": "Whitespace cleanup regression vs Deepseek tokenizer incorrect results; not the same bug." + "reason": "Both are tokenizer regressions, but the observed behavior and likely code paths differ substantially." }, { "left": "issue:45137", "right": "issue:45276", "accept": false, - "reason": "DeepSpeed ZeRO3 deque error vs Gemma4 resize_token_embeddings not propagating; unrelated." + "reason": "DeepSpeed deque error and Gemma4 resize_token_embeddings propagation bug are unrelated." }, { "left": "issue:43874", "right": "issue:45092", "accept": false, - "reason": "Missing image-patch-count method vs remote-code checkpoint/meta-init incompatibility; different model loading problems." + "reason": "Missing image-patch helper and old InternVL2 meta-init incompatibility are distinct multimodal issues." }, { "left": "issue:44077", "right": "issue:44242", "accept": false, - "reason": "patchtsmixer post_init policy issue vs load-balancing-loss omission; unrelated." + "reason": "Optional post_init policy and missing load-balancing loss are different model implementation bugs." }, { "left": "issue:42898", "right": "issue:44448", "accept": false, - "reason": "Tokenizer cleanup regression vs Pegasus v4/v5 output difference; different behaviors." + "reason": "Both are v5 regressions, but Pegasus output mismatch is broader generation behavior, not the specific clean_up_tokenization_spaces change." }, { "left": "issue:41628", "right": "issue:45459", "accept": false, - "reason": "Missing class import vs tokenizer error masking when protobuf is absent; different import/error-handling bugs." + "reason": "AutoImageProcessor import failure and hidden tokenizer errors without protobuf are unrelated." }, { "left": "issue:44222", "right": "issue:44297", "accept": false, - "reason": "FP8 save_pretrained bug vs tokenizer_class metadata mismatch; different save_pretrained problems." + "reason": "Both touch save_pretrained, but one is FP8 MoE model saving and the other is tokenizer metadata mismatch." }, { "left": "issue:42898", "right": "issue:43295", "accept": false, - "reason": "Tokenizer cleanup regression vs multimodal processor.tokenizer regression; unrelated. " + "reason": "Tokenizer cleanup change and processor/tokenizer image-passing regression are different v5 issues." } ] }, @@ -10641,8 +10637,8 @@ "issue:44514|issue:45103", "issue:43479|issue:44913", "issue:44373|issue:44485", - "issue:44261|issue:45092", "issue:43723|issue:44871", + "issue:44261|issue:45092", "issue:43723|issue:44991", "issue:43723|issue:44442", "issue:45092|issue:45357", @@ -10653,20 +10649,20 @@ "issue:42831|issue:44038", "issue:43576|issue:43784", "issue:44263|issue:44373", - "issue:43122|issue:44521", "issue:42757|issue:45278", + "issue:43122|issue:44521", "issue:42886|issue:43066", "issue:44117|issue:44704", "issue:43122|issue:44514", "issue:30990|issue:33357", "issue:44704|issue:45290" ], - "prepared_review_unit_hash": "5d5c935709234d60e01e0000bf5f940119d8cd4c0821294f8fd25016744aa7c6", + "prepared_review_unit_hash": "f3a1ffd23eae00a1c85c5e5a9f214c8145564cfa8d797e2411e87c15fcde902a", "component_budget": { "node_count": 367, "item_count": 367, "soft_pair_count": 1370, - "serialized_chars": 706987, + "serialized_chars": 706986, "estimated_input_tokens": 176747, "estimated_eval_tokens": 353750 }, @@ -10674,183 +10670,183 @@ "node_count": 48, "item_count": 48, "soft_pair_count": 36, - "serialized_chars": 38118, - "estimated_input_tokens": 9530, - "estimated_eval_tokens": 19316 + "serialized_chars": 38110, + "estimated_input_tokens": 9528, + "estimated_eval_tokens": 19312 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is heterogeneous: most items are distinct Transformers v5 regressions across tokenizer loading, imports, processor/chat-template behavior, config persistence, and save_pretrained paths. I did not find any soft-edge pair that clearly represents the same underlying bug/change, so all soft pairs are rejected.", - "confidence": 0.92, - "canonical_issue_reason": "Issue 45278 is the broadest umbrella in the set: it captures multiple import failures after upgrading to Transformers 5.5.0, which best matches the dominant import/load-regression theme here.", + "summary": "This cluster is heterogeneous: a few issues are about v5 import/tokenizer regressions, but many others are unrelated (FP8, multimodal processors, chat templates, config serialization, etc.). I don\u2019t see any soft pair that clearly describes the same underlying bug/change.", + "confidence": 0.91, + "canonical_issue_reason": "Issue 45278 is the broadest umbrella in the set, covering the general v5 import-regression theme better than the narrower single-symbol reports.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 45278 is the best representative because it is broad, recent, and already frames the recurring v5 import problems seen in nearby reports; however, it should be treated as an umbrella report, not a duplicate target for the other specific bugs.", + "best_issue_reason": "Issue 45278 is the most representative cluster anchor because it is the most general report of the recurring v5 breakage pattern.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43479", "right": "issue:43644", "accept": false, - "reason": "Different bugs: config default initialization vs non-persistent buffer corruption." + "reason": "Different bugs: default config initialization for Phi4MultimodalConfig vs junk-filled non-persistent buffers in v5.0.0." }, { "left": "issue:44442", "right": "issue:45459", "accept": false, - "reason": "Tokenizer loading failure vs protobuf-error masking in error handling." + "reason": "Both involve tokenizer loading, but one is a missing tokenizer class while the other hides protobuf-related decode errors; not the same bug." }, { "left": "issue:44222", "right": "issue:44623", "accept": false, - "reason": "Both involve save_pretrained, but one is FP8 MoE saving and the other is missing processor files." + "reason": "FP8 moe save_pretrained failure and processor.save_pretrained missing files are separate save paths with different failure modes." }, { "left": "issue:42831", "right": "issue:44623", "accept": false, - "reason": "Accuracy regression in FineGrainedFP8 is unrelated to processor.save_pretrained file omission." + "reason": "Accuracy regression in FineGrainedFP8 is unrelated to processor save_pretrained file omissions." }, { "left": "issue:44206", "right": "issue:45103", "accept": false, - "reason": "Feature-extractor argument crash and auto_docstring annotation crash are unrelated code paths." + "reason": "Unsupported center arg in LasrFeatureExtractor and auto_docstring AttributeError are unrelated code paths." }, { "left": "issue:22355", "right": "issue:28282", "accept": false, - "reason": "Missing transformers.onnx module and missing PyTorch dependency are different import problems." + "reason": "Both are import errors, but one is missing transformers.onnx and the other is missing PyTorch; different root causes." }, { "left": "issue:44297", "right": "issue:45310", "accept": false, - "reason": "Tokenizer metadata mismatch and Qwen3.5 MoE from_pretrained failure are different bugs." + "reason": "Tokenizer config mismatch on save is not the same as Qwen3.5 MoE from_pretrained failure." }, { "left": "issue:41628", "right": "issue:42757", "accept": false, - "reason": "Different missing-symbol import errors in different packages/modules." + "reason": "Different missing imports from different packages (AutoImageProcessor vs is_offline_mode)." }, { "left": "issue:44483", "right": "issue:45406", "accept": false, - "reason": "Server request handling crash is unrelated to Gemma4Processor missing _tokenizer." + "reason": "Chat/completions request rejection and Gemma4Processor missing _tokenizer are not the same failure." }, { "left": "issue:44222", "right": "issue:44484", "accept": false, - "reason": "save_pretrained failure and a max_shard_size question are not the same issue." + "reason": "Both mention save_pretrained, but one is FP8/moe serialization and the other is shard-size default behavior." }, { "left": "issue:44263", "right": "issue:44360", "accept": false, - "reason": "Same DSA indexer area, but torch.split handling and missing ReLU are distinct defects." + "reason": "Same subsystem name, but one is about torch.split outputs and the other is about a missing ReLU in the DSA indexer." }, { "left": "issue:43723", "right": "issue:44977", "accept": false, - "reason": "AutoTokenizer load regression and flash-attention generation failure are different code paths." + "reason": "Tokenizer-loading regression vs flash-attention generation bug; different code paths." }, { "left": "issue:44263", "right": "issue:45092", "accept": false, - "reason": "DSA indexer bug is unrelated to InternVL2 meta-initialization incompatibility." + "reason": "GLM DSA indexing bug is unrelated to remote-code checkpoint meta-initialization incompatibility." }, { "left": "issue:44295", "right": "issue:44623", "accept": false, - "reason": "position_ids buffer read issue is unrelated to processor save_pretrained missing files." + "reason": "Position_ids buffer read error is unrelated to missing processor save files." }, { "left": "issue:44514", "right": "issue:45103", "accept": false, - "reason": "Chat-template crash with tool calls is unrelated to auto_docstring annotation handling." + "reason": "Qwen2_5_VLProcessor batched apply_chat_template crash is unrelated to auto_docstring annotation handling." }, { "left": "issue:43479", "right": "issue:44913", "accept": false, - "reason": "Config constructor defaulting and rotary_pct persistence are different configuration bugs." + "reason": "Both are config-related, but they affect different models and different config fields." }, { "left": "issue:44373", "right": "issue:44485", "accept": false, - "reason": "Wrong docstring for position_ids is not the GLM-5 RoPE implementation bug." + "reason": "Docstring correction for position_ids is not the same as a RoPE implementation bug." }, { - "left": "issue:44261", - "right": "issue:45092", + "left": "issue:43723", + "right": "issue:44871", "accept": false, - "reason": "MLA RMS norm epsilon precision issue is unrelated to remote-code meta-init incompatibility." + "reason": "Both concern tokenizer/model loading, but they describe different regressions with different symptoms and targets." }, { - "left": "issue:43723", - "right": "issue:44871", + "left": "issue:44261", + "right": "issue:45092", "accept": false, - "reason": "Tokenizer loading failure and eos_token_id inconsistency are different symptoms and fixes." + "reason": "MLA rms_norm_eps precision issue is unrelated to remote-code meta initialization compatibility." }, { "left": "issue:43723", "right": "issue:44991", "accept": false, - "reason": "Both are tokenizer-related, but they concern different models and different failure modes." + "reason": "Generic AutoTokenizer load issue and a specific EMBEDDIA/est-roberta tokenizer failure are not clearly the same bug." }, { "left": "issue:43723", "right": "issue:44442", "accept": false, - "reason": "Different tokenizer load failures for different models; not the same concrete bug." + "reason": "Both are tokenizer loading problems, but they point to different tokenizer classes and likely different causes." }, { "left": "issue:45092", "right": "issue:45357", "accept": false, - "reason": "Remote-code meta-init incompatibility is unrelated to incorrect visual encoder keys on save_pretrained." + "reason": "Loading incompatibility for old InternVL2 checkpoints is unrelated to Qwen3.5 save_pretrained key serialization." }, { "left": "issue:30990", "right": "issue:38175", "accept": false, - "reason": "Sentence-Transformers loading stall and zero-probability Siglip2 outputs are unrelated." + "reason": "Sentence Transformers loading hang and zero-probability SigLIP output are different issues." }, { "left": "issue:42886", "right": "issue:42994", "accept": false, - "reason": "Offline tokenizer cache loading and quantized model saving are different problems." + "reason": "Offline tokenizer cache loading and quantized model saving failures are unrelated." }, { "left": "issue:42831", "right": "issue:45310", "accept": false, - "reason": "FP8 accuracy regression and Qwen3.5 MoE from_pretrained error are unrelated." + "reason": "FP8 accuracy regression and Qwen3.5 MoE from_pretrained error are different bugs." }, { "left": "issue:41628", "right": "issue:43824", "accept": false, - "reason": "Different missing imports for different symbols; same broad theme, not the same bug." + "reason": "Different missing imports from different symbols and model classes." }, { "left": "issue:42831", "right": "issue:44038", "accept": false, - "reason": "FP8 accuracy issue is unrelated to the Qwen3.5-VL-Moe regression." + "reason": "FineGrainedFP8 accuracy problems are unrelated to Qwen3-VL-Moe loading on v5." }, { "left": "issue:43576", @@ -10862,49 +10858,49 @@ "left": "issue:44263", "right": "issue:44373", "accept": false, - "reason": "One is a runtime DSA-indexer bug, the other is just a wrong docstring." + "reason": "Indexing behavior and docstring text are not the same underlying bug." }, { - "left": "issue:43122", - "right": "issue:44521", + "left": "issue:42757", + "right": "issue:45278", "accept": false, - "reason": "Tokenization behavior change and assistant_masks bug are different multimodal/template issues." + "reason": "A specific missing hub symbol is not enough to call it the same bug as a broad 'many import errors' report." }, { - "left": "issue:42757", - "right": "issue:45278", + "left": "issue:43122", + "right": "issue:44521", "accept": false, - "reason": "45278 is a broad umbrella for import failures, but 42757 is one specific missing symbol; not the same concrete bug." + "reason": "Different tokenization problems: reload changes tokenization vs assistant mask handling in multimodal chat templates." }, { "left": "issue:42886", "right": "issue:43066", "accept": false, - "reason": "Offline cache loading issue and wrong tokenizer decoder type are different tokenizer bugs." + "reason": "Offline cache loading failure and tokenizer decoder type mismatch are unrelated." }, { "left": "issue:44117", "right": "issue:44704", "accept": false, - "reason": "Tokenizer mapping None and AutoProcessor kwargs propagation are unrelated failures." + "reason": "TOKENIZER_MAPPING_NAMES returning None and AutoProcessor kwargs forwarding are different failure points." }, { "left": "issue:43122", "right": "issue:44514", "accept": false, - "reason": "Different apply_chat_template-related regressions with different triggers and outputs." + "reason": "Tokenization changes across versions and batched chat-template crashes are different bugs." }, { "left": "issue:30990", "right": "issue:33357", "accept": false, - "reason": "Loading hang in sentence-transformers and MacOS bus error in CLIP are unrelated." + "reason": "Loading stuck vs MacOS bus error on a different model/version; no same code-path evidence." }, { "left": "issue:44704", "right": "issue:45290", "accept": false, - "reason": "cached_file kwargs forwarding and tool-call assistant content crash are different code-path bugs." + "reason": "AutoProcessor cached_file kwargs issue is unrelated to apply_chat_template crashing on tool-call messages." } ] }, @@ -10989,8 +10985,8 @@ "issue:44625|issue:44829", "issue:41628|issue:45278", "issue:44360|issue:44485", - "issue:29127|issue:30990", "issue:43116|issue:45200", + "issue:29127|issue:30990", "issue:39401|issue:44521", "issue:44261|issue:44373", "issue:43122|issue:44779", @@ -11002,20 +10998,20 @@ "issue:42907|issue:43452", "issue:44038|issue:44222", "issue:43992|issue:44704", - "issue:43122|issue:44977", "issue:45216|issue:45478", - "issue:43278|issue:44945", + "issue:43122|issue:44977", "issue:43824|issue:44351", + "issue:43278|issue:44945", "issue:42907|issue:44222", "issue:42757|issue:44351", "issue:42222|issue:43441" ], - "prepared_review_unit_hash": "c23c67ee45c4b22730e405650ed3e730f4c060224d321b4a94cde1d215849fda", + "prepared_review_unit_hash": "cd100f53f5dddd64782cb1f70e28d97e95d7d381ca0536995a4b498dc79e54ef", "component_budget": { "node_count": 367, "item_count": 367, "soft_pair_count": 1370, - "serialized_chars": 706987, + "serialized_chars": 706986, "estimated_input_tokens": 176747, "estimated_eval_tokens": 353750 }, @@ -11023,213 +11019,213 @@ "node_count": 47, "item_count": 47, "soft_pair_count": 32, - "serialized_chars": 36179, - "estimated_input_tokens": 9045, - "estimated_eval_tokens": 18346 + "serialized_chars": 36173, + "estimated_input_tokens": 9044, + "estimated_eval_tokens": 18344 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Mostly heterogeneous cluster with a few superficial similarity themes (import errors, tokenizer regressions, FP8/MoE, SAM3), but no soft pair appears to be the same concrete bug or change.", - "confidence": 0.29, - "canonical_issue_reason": "issue:45278 is the broadest umbrella in the set, covering the recurring import-regression theme seen in several nearby items; however, the overall set is still too mixed for a strong canonical duplicate.", + "summary": "This cluster is mostly heterogeneous: import regressions, tokenizer issues, generation bugs, multimodal/model loading problems, and FP8/parallelism issues are all mixed together. I did not find any high-confidence duplicate pair; the closest look-alike is a tokenizer regression pair, but the titles are still too broad to merge confidently.", + "confidence": 0.43, + "canonical_issue_reason": "issue:43824 is the most central concrete regression in the set, with multiple inbound references and a clear import-failure symptom; the rest of the items are too diverse to form one true duplicate cluster.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45278 is the least misleading global representative because it captures the widest shared symptom class among the related-looking items, though it does not actually unify the whole list.", + "best_issue_reason": "issue:43824 is the best representative issue here because it is specific, user-facing, and has the strongest linkage among the listed issues.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:42371", "right": "issue:44263", "accept": false, - "reason": "Different bugs: TF32 settings guidance vs a torch.split behavior issue in GlmMoeDsaIndexer." + "reason": "TF32 behavior guidance vs a torch.split indexing bug; unrelated symptoms and code paths." }, { "left": "issue:44295", "right": "issue:44297", "accept": false, - "reason": "Unrelated failures: position_ids buffer access vs tokenizer_class mismatch on save_pretrained." + "reason": "Buffer access error for position_ids vs tokenizer_class mismatch on save; different subsystems." }, { "left": "issue:42907", "right": "issue:44945", "accept": false, - "reason": "Different code paths: save/dequantization problem vs incorrect output under pipeline parallelism." + "reason": "Dequantized model save failure vs incorrect output under pipeline parallelism; not the same bug." }, { "left": "issue:43976", "right": "issue:44945", "accept": false, - "reason": "Version/Python compatibility issue vs pipeline-parallel output bug; no shared concrete defect." + "reason": "Python version compatibility issue vs pipeline-parallel output bug; different failure modes." }, { "left": "issue:43232", "right": "issue:45245", "accept": false, - "reason": "Generation state update bug vs category-count runtime error; not the same issue." + "reason": "Generation kwargs/cache bug vs category-cardinality runtime error; unrelated." }, { "left": "issue:43232", "right": "issue:44743", "accept": false, - "reason": "Both touch generation state, but one is sync_gpus bookkeeping and the other is recurrent-state reset in Qwen3.5; too different." + "reason": "Both touch generation state, but one is sync_gpus/update_model_kwargs and the other is recurrent-state reset in modular_qwen3_5; too distinct." }, { "left": "issue:39401", "right": "issue:44488", "accept": false, - "reason": "Tokenizer offset_mapping bug vs model-loading regression; no shared underlying change." + "reason": "Qwen3 tokenizer offset_mapping regression vs model load failure for a different model; not the same issue." }, { "left": "issue:42907", "right": "issue:43937", "accept": false, - "reason": "Different model-specific failures: dequantized save vs invalid GenerationConfig for GLM-5." + "reason": "Dequantized save failure vs invalid GenerationConfig for GLM-5; different models and code paths." }, { "left": "issue:43408", "right": "issue:43475", "accept": false, - "reason": "Same SAM3 area, but one is a model-type warning and the other is a missing attribute in the vision encoder output." + "reason": "Model-type warning during instantiation vs missing attribute on Sam3VisionEncoderOutput; related family, but different concrete bugs." }, { "left": "issue:44625", "right": "issue:44829", "accept": false, - "reason": "Different problems: num_labels propagation vs flash_attention_3 training degeneration." + "reason": "num_labels propagation bug vs flash_attention_3 degenerate training; unrelated." }, { "left": "issue:41628", "right": "issue:45278", "accept": false, - "reason": "Both are import-related, but one is a specific missing symbol and the other is a broad post-upgrade import regression; not the same bug." + "reason": "Specific AutoImageProcessor import error vs broad post-upgrade import errors; too generic to treat as the same bug." }, { "left": "issue:44360", "right": "issue:44485", "accept": false, - "reason": "DSA ReLU implementation discussion vs GLM-5 RoPE implementation; unrelated." + "reason": "DSA indexer ReLU discussion vs GLM-5 RoPE implementation; unrelated." }, { - "left": "issue:29127", - "right": "issue:30990", + "left": "issue:43116", + "right": "issue:45200", "accept": false, - "reason": "LayoutLMv3 error-message clarity issue vs Sentence Transformers loading hang; unrelated." + "reason": "Example-script multi-label empty results vs mm_token_type_ids defaulting; different model/task paths." }, { - "left": "issue:43116", - "right": "issue:45200", + "left": "issue:29127", + "right": "issue:30990", "accept": false, - "reason": "Example-script multilabel classification bug vs Gemma 4 mm_token_type_ids defaulting issue." + "reason": "LayoutLMv3 error-message clarity vs Sentence Transformers loading hang; unrelated." }, { "left": "issue:39401", "right": "issue:44521", "accept": false, - "reason": "Different multimodal/tokenizer bugs: offset mapping vs assistant_masks all-zero." + "reason": "Tokenizer offset_mapping regression vs multimodal assistant_mask bug; different functionality." }, { "left": "issue:44261", "right": "issue:44373", "accept": false, - "reason": "Precision/config bug vs documentation typo; not the same underlying issue." + "reason": "LayerNorm precision/config issue vs docstring mismatch; not duplicates." }, { "left": "issue:43122", "right": "issue:44779", "accept": false, - "reason": "Both are tokenizer regressions, but for different models and likely different code paths." + "reason": "Both mention tokenizer regressions in v5, but one is broad and the other is Deepseek-specific; not enough evidence they are the same underlying bug." }, { "left": "issue:44038", "right": "issue:44297", "accept": false, - "reason": "Qwen3-VL-Moe compatibility issue vs tokenizer save_pretrained metadata mismatch." + "reason": "Qwen3-VL-Moe bug vs tokenizer save metadata mismatch; different symptoms." }, { "left": "issue:42371", "right": "issue:42831", "accept": false, - "reason": "TF32 behavior settings vs FineGrainedFP8 accuracy issue; unrelated." + "reason": "TF32 configuration warning vs FineGrainedFP8 accuracy issue; unrelated." }, { "left": "issue:44038", "right": "issue:44295", "accept": false, - "reason": "Model/version compatibility issue vs position_ids buffer access bug." + "reason": "Qwen3-VL-Moe bug vs position_ids buffer read error; different code paths." }, { "left": "issue:42907", "right": "issue:44964", "accept": false, - "reason": "Different model-loading/dequantization bug vs Phi-4 multimodal load failure." + "reason": "Dequantized save failure vs Phi-4 multimodal load failure; not the same bug." }, { "left": "issue:36296", "right": "issue:43116", "accept": false, - "reason": "Tensor-parallel training bug vs multilabel example-script bug; no concrete overlap." + "reason": "Tensor parallel training bug vs classification example returning empty results; unrelated." }, { "left": "issue:42907", "right": "issue:43452", "accept": false, - "reason": "Dequantized save failure vs gguf_file loading breakage; different paths." + "reason": "Saving dequantized Ministral/Devstral models vs gguf_file handling for from_pretrained; different failure modes." }, { "left": "issue:44038", "right": "issue:44222", "accept": false, - "reason": "Same broad model family, but one is a load/compatibility bug and the other is an FP8 save_pretrained issue; too distinct." + "reason": "Qwen3-VL-Moe bug vs FP8 save_pretrained issue; too broad and not concretely the same bug." }, { "left": "issue:43992", "right": "issue:44704", "accept": false, - "reason": "Missing weight during from_pretrained vs AutoProcessor kwargs forwarding; unrelated." - }, - { - "left": "issue:43122", - "right": "issue:44977", - "accept": false, - "reason": "Tokenizer regression reports for different models; no evidence they are the same defect." + "reason": "Missing embed_tokens.weight in UMT5Encoder loading vs AutoProcessor kwargs not passed to cached_file; unrelated." }, { "left": "issue:45216", "right": "issue:45478", "accept": false, - "reason": "Both mention 5.4.0 and Qwen3.5, but one is save_pretrained checkpoint correctness and the other is from_pretrained/MoE loading; not clearly the same bug." + "reason": "Qwen3.5 save_pretrained regression vs Qwen3.5 Moe from_pretrained error; same family but different operations and likely different fixes." }, { - "left": "issue:43278", - "right": "issue:44945", + "left": "issue:43122", + "right": "issue:44977", "accept": false, - "reason": "dtype mismatch during eval vs incorrect output under pipeline parallelism; different issues." + "reason": "Tokenizer regression vs flash-attention generation bug; not the same code path." }, { "left": "issue:43824", "right": "issue:44351", "accept": false, - "reason": "Both are import errors, but for different symbols and no shared concrete failing code path is shown." + "reason": "Missing Qwen2_5_VLForConditionalGeneration import vs missing HybridCache import; both import errors, but distinct symbols and likely distinct causes." + }, + { + "left": "issue:43278", + "right": "issue:44945", + "accept": false, + "reason": "Embedding dtype mismatch between train/eval vs incorrect output under pipeline parallelism; different problems." }, { "left": "issue:42907", "right": "issue:44222", "accept": false, - "reason": "Both mention FP8/MoE-adjacent behavior, but one is a save/dequantization problem and the other is a save_pretrained moe issue; not enough to merge." + "reason": "Dequantized save failure vs FP8 save_pretrained moe; both involve saving, but not the same concrete bug." }, { "left": "issue:42757", "right": "issue:44351", "accept": false, - "reason": "Missing import from huggingface_hub vs missing import from transformers; superficially similar but different packages and defects." + "reason": "Missing is_offline_mode from huggingface_hub vs missing HybridCache from transformers; both imports, but different packages and errors." }, { "left": "issue:42222", "right": "issue:43441", "accept": false, - "reason": "Different model families and failures: vitpose breakage vs Ministral-3 FlashAttention failure." + "reason": "Broken vitpose models vs Ministral-3 FlashAttention failure; unrelated models and issues." } ] }, @@ -11318,8 +11314,8 @@ "issue:43976|issue:45446", "issue:43937|issue:45563", "issue:39401|issue:44779", - "issue:43931|issue:44945", "issue:33453|issue:39401", + "issue:43931|issue:44945", "issue:45083|issue:45446", "issue:45341|issue:45561", "issue:43937|issue:44977", @@ -11333,12 +11329,12 @@ "issue:39401|issue:43421", "issue:38617|issue:41628" ], - "prepared_review_unit_hash": "f958c0f1a5e16e1809d47d1b5db6ba026a3a0f527ac614d7747c2180cd2662b8", + "prepared_review_unit_hash": "168ecd36525d07682a988987201d2484fd63639928bfab1326aa9931f811aab2", "component_budget": { "node_count": 367, "item_count": 367, "soft_pair_count": 1370, - "serialized_chars": 706987, + "serialized_chars": 706986, "estimated_input_tokens": 176747, "estimated_eval_tokens": 353750 }, @@ -11346,207 +11342,195 @@ "node_count": 47, "item_count": 47, "soft_pair_count": 30, - "serialized_chars": 35832, - "estimated_input_tokens": 8958, - "estimated_eval_tokens": 18172 + "serialized_chars": 35833, + "estimated_input_tokens": 8959, + "estimated_eval_tokens": 18174 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a loose set of Transformers regression reports with only superficial title similarity. The items span unrelated import errors, tokenizer bugs, model-specific failures, trainer/runtime issues, and docs/examples problems; none of the soft pairs look like the same underlying change or bug.", - "confidence": 0.93, - "canonical_issue_reason": "issue:41628 is the cleanest representative of the cluster\u2019s main theme: a concrete, user-facing import regression in core Transformers with a precise missing public symbol.", + "summary": "These items are mostly unrelated regression reports across different areas (imports, tokenizers, generation, training, multimodal, testing). I do not see any soft pair that clearly describes the same underlying bug/change, so I would not merge any of the suggested edges. If one issue must serve as an umbrella, 45278 is the broadest open import-regression report.", + "confidence": 0.77, + "canonical_issue_reason": "issue:45278 is the broadest open report in the set and the only plausible umbrella-style tracker ('many import errors after update from 4.57.0 to 5.5.0').", "canonical_pr_reason": null, - "best_issue_reason": "issue:41628 is the strongest standalone issue here because it is specific, reproducible, and clearly actionable, unlike the broader or more ambiguous reports in the set.", + "best_issue_reason": "issue:45278 is the most general/open issue and the closest thing to a cluster sink, even though the rest of the set is too heterogeneous for true duplication.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43784", "right": "issue:45446", "accept": false, - "reason": "Different failures: sentence-transformers import NameError vs a flex_attention PyTorch version-check bug." + "reason": "Different failures: import-time NameError in sentence-transformers integration vs a PyTorch version-check bug in flex_attention." }, { "left": "issue:42673", "right": "issue:45072", "accept": false, - "reason": "VRAM leak in Qwen3ForCausalLM is unrelated to dtype mismatch issues in SwitchTransformers/TimmWrapperModel." + "reason": "Unrelated problems: VRAM leak in Qwen3ForCausalLM thread usage vs dtype mismatches in inference tests." }, { "left": "issue:43064", "right": "issue:43673", "accept": false, - "reason": "Different code paths: FSDP2/PEFT optimizer state corruption vs GenerationMixin cache/chunked_prefill behavior." + "reason": "Different code paths: FSDP2/PEFT optimizer-state issue vs GenerationMixin cache handling during chunked prefill." }, { "left": "issue:38617", "right": "issue:45278", "accept": false, - "reason": "One is a specific missing import; the other is a broad report of many import errors after upgrade." + "reason": "45278 is a broad import-error report, but 38617 is one specific missing symbol; not the same concrete bug." }, { "left": "issue:43550", "right": "issue:43784", "accept": false, - "reason": "torch.compile/SDPA model failure is unrelated to an import-time NameError in sentence-transformers." + "reason": "Different subsystems and symptoms: torch.compile/SDPA failure for Bamba vs sentence-transformers import NameError." }, { "left": "issue:22355", "right": "issue:29127", "accept": false, - "reason": "No shared underlying bug: ONNX module import failure vs LayoutLMv3 error-message clarity." + "reason": "No shared underlying bug: ONNX module import failure vs LayoutLMv3 error-message clarity issue." }, { "left": "issue:43937", "right": "issue:44485", "accept": false, - "reason": "GenerationConfig validation error is distinct from GLM-5 RoPE implementation discussion." + "reason": "Both mention GLM-5, but one is invalid GenerationConfig handling and the other is a RoPE implementation discussion." }, { "left": "issue:43116", "right": "issue:43493", "accept": false, - "reason": "Example-script multi-label output bug is not the same as SigLIP2 HF/JAX implementation discrepancy." + "reason": "Different artifacts: example script multi-label classification bug vs SigLIP2 model discrepancy with JAX." }, { "left": "issue:42371", "right": "issue:43874", "accept": false, - "reason": "TF32 API guidance issue is unrelated to a missing image-processor method." + "reason": "TF32-behavior API guidance and a missing image-patch helper in GLM46V are unrelated." }, { "left": "issue:44075", "right": "issue:44936", "accept": false, - "reason": "Optimizer argument handling and trainer evaluate-after-train behavior are different bugs." + "reason": "Optimizer argument handling is unrelated to trainer.evaluate() failing after training." }, { "left": "issue:44222", "right": "issue:45478", "accept": false, - "reason": "FP8 save_pretrained issue is not the same as Qwen3.5 Moe from_pretrained loading error." + "reason": "Both touch MoE/Qwen3.5, but one is FP8 save_pretrained behavior and the other is a from_pretrained load error." }, { "left": "issue:44038", "right": "issue:45478", "accept": false, - "reason": "Different model families and symptoms: Qwen3-VL-Moe bug vs Qwen3.5 Moe loading error." + "reason": "Different Qwen3-VL/MoE reports with different failure modes; not the same concrete bug." }, { "left": "issue:43116", "right": "issue:43825", "accept": false, - "reason": "Empty outputs in a classification example are unrelated to pipeline() translation-task messaging." + "reason": "Example-script classification bug is unrelated to the pipeline() translation-task warning." }, { "left": "issue:43976", "right": "issue:45446", "accept": false, - "reason": "Python-version compatibility regression is not the same as an AuxRequest import/version-check bug." + "reason": "Python-version compatibility regression and a flex_attention import/version check are not the same issue." }, { "left": "issue:43937", "right": "issue:45563", "accept": false, - "reason": "GLM-5 GenerationConfig validation is unrelated to a stale num_return_sequences warning." + "reason": "GenerationConfig invalidity is unrelated to a stale warning in paged generate()." }, { "left": "issue:39401", "right": "issue:44779", "accept": false, - "reason": "Both are tokenizer regressions, but they affect different models and different behavior (offset mapping vs tokenization output)." + "reason": "Both are tokenizer regressions, but they affect different models and different observable failures." }, { - "left": "issue:43931", - "right": "issue:44945", + "left": "issue:33453", + "right": "issue:39401", "accept": false, - "reason": "Qwen3-VL weight-shape mismatch is not the same as pipeline-parallelism output corruption." + "reason": "Generic tokenizer-loading regression and Qwen3 offset_mapping bug are too different to treat as one fix." }, { - "left": "issue:33453", - "right": "issue:39401", + "left": "issue:43931", + "right": "issue:44945", "accept": false, - "reason": "Generic tokenizer-loading regression and Qwen3 offset_mapping bug are not the same concrete issue." + "reason": "Weight-shape mismatch during model loading is unrelated to incorrect output under pipeline parallelism." }, { "left": "issue:45083", "right": "issue:45446", "accept": false, - "reason": "Qwen3 Omni feature-length helper behavior is unrelated to flex_attention import/version checking." + "reason": "Helper-function behavior in qwen3_omni_moe and a PyTorch version-check bug in flex_attention are different code paths." }, { "left": "issue:45341", "right": "issue:45561", "accept": false, - "reason": "A small testing_utils bug is not the same as a pytest-xdist race on captured_info.txt." + "reason": "A generic testing_utils bug and a pytest-xdist race on captured_info.txt are not the same underlying defect." }, { "left": "issue:43937", "right": "issue:44977", "accept": false, - "reason": "GLM-5 generation config validation and Qwen3.5 flash-attention generation failure are different bugs." + "reason": "Different generation bugs: invalid GenerationConfig vs flash-attention-related generation failure in Qwen3.5." }, { "left": "issue:43827", "right": "issue:45446", "accept": false, - "reason": "Docs still referencing pipeline() is unrelated to a flex_attention import/version issue." + "reason": "Docs still referencing pipeline() after v5 removals is unrelated to the flex_attention version-check issue." }, { "left": "issue:42673", "right": "issue:43906", "accept": false, - "reason": "VRAM leak report is not the same as an isolated reproduction of issue #38071." + "reason": "The second item is only an isolated reproduction of another issue; there is no evidence it matches the VRAM leak." }, { "left": "issue:44291", "right": "issue:45071", "accept": false, - "reason": "Unexpected _is_hf_initialized argument during init_empty_weights is distinct from PretrainedConfig type-checking breakage." + "reason": "Unexpected _is_hf_initialized argument in init_empty_weights and PreTrainedConfig type-checking breakage are distinct loading/type-system regressions." }, { "left": "issue:45405", "right": "issue:45446", "accept": false, - "reason": "Release/version pinning for PEFT is unrelated to the AuxRequest import/version-check bug." + "reason": "PEFT version pinning problem is unrelated to a PyTorch version check for flex_attention." }, { "left": "issue:43450", "right": "issue:43986", "accept": false, - "reason": "Batched video-shape bug is different from crashing because torchvision is missing." + "reason": "Batched video-processor shape bug and AutoProcessor crash without torchvision are different failures." }, { "left": "issue:44829", "right": "issue:45216", "accept": false, - "reason": "flash_attention_3 training degeneration is unrelated to Qwen3.5 save_pretrained checkpoint corruption." + "reason": "Flash-attention-3 training degeneration and save_pretrained checkpoint correctness are not the same bug." }, { "left": "issue:44484", "right": "issue:45310", "accept": false, - "reason": "A save_pretrained shard-size question is not the same as Qwen3.5 Moe from_pretrained failure." - }, - { - "left": "issue:39401", - "right": "issue:43421", - "accept": false, - "reason": "Tokenizer offset mapping regression is unrelated to runtime post-processor updates for special tokens." - }, - { - "left": "issue:38617", - "right": "issue:41628", - "accept": false, - "reason": "Both are import errors, but they expose different missing symbols and likely different code paths." + "reason": "A question about max_shard_size and a Qwen3.5 MoE from_pretrained error are unrelated." }, { "left": "issue:38617", "right": "issue:41628", "accept": false, - "reason": "Same as above: separate import regressions, not the same underlying fix." + "reason": "Both are import errors, but one is a missing internal symbol and the other is a missing top-level export; no same fix is evident." } ] }, @@ -11622,15 +11606,15 @@ "soft_pairs": [ "issue:44117|issue:45081", "issue:44410|issue:45310", - "issue:43408|issue:43742", "issue:42673|issue:43582", - "issue:42222|issue:45542", + "issue:43408|issue:43742", "issue:42673|issue:43577", + "issue:42222|issue:45542", "issue:42371|issue:44261", "issue:43937|issue:44496", "issue:38617|issue:44220", - "issue:44483|issue:45464", "issue:42175|issue:42222", + "issue:44483|issue:45464", "issue:42617|issue:45341", "issue:44933|issue:45278", "issue:44484|issue:45478", @@ -11656,12 +11640,12 @@ "issue:43408|issue:44368", "issue:43874|issue:44263" ], - "prepared_review_unit_hash": "96f35fa3897a5de507614e0f80fb7ba85af297d63d0770f7725d18450d2dadb2", + "prepared_review_unit_hash": "c861c0621562764fca53427cb4152bc0b96a8fcb801cd1d34d0077ffe49bfed8", "component_budget": { "node_count": 367, "item_count": 367, "soft_pair_count": 1370, - "serialized_chars": 706987, + "serialized_chars": 706986, "estimated_input_tokens": 176747, "estimated_eval_tokens": 353750 }, @@ -11669,231 +11653,231 @@ "node_count": 48, "item_count": 48, "soft_pair_count": 35, - "serialized_chars": 38016, - "estimated_input_tokens": 9504, - "estimated_eval_tokens": 19264 + "serialized_chars": 38012, + "estimated_input_tokens": 9503, + "estimated_eval_tokens": 19262 }, "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 and model-specific bugs; none of the soft pairs look like clear duplicates of the same concrete fix.", - "confidence": 0.64, - "canonical_issue_reason": "issue:45278 is the broadest umbrella report in the set, covering multiple import errors after upgrading from 4.57.0 to 5.5.0, which makes it the best representative of the cluster's API-breakage theme.", + "summary": "These items are mostly unrelated single-issue reports; none of the soft pairs look like the same underlying bug/change closely enough to merge. The closest thematic overlap is around import/regression reports, but even there the concrete failure modes differ.", + "confidence": 0.88, + "canonical_issue_reason": "If one issue must anchor the cluster, issue 45278 is the broadest import-regression report ('Many import errors after update from 4.57.0 to 5.5.0'), but it is still not a true duplicate umbrella for the others.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45278 is the most useful cluster anchor because it is user-facing, version-specific, and broad enough to subsume related import-regression reports.", + "best_issue_reason": "Issue 45278 is the best global representative because it is the most general and user-facing import regression in the set, making it the most plausible triage hub among the issue-only items.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44117", "right": "issue:45081", "accept": false, - "reason": "Different failures: tokenizer mapping None-handling vs a Mistral regex patch crash in tokenizer internals." + "reason": "Both are tokenizer-loading failures, but one is a mapping/assumption bug and the other is a Mistral regex patch crash; different code paths." }, { "left": "issue:44410", "right": "issue:45310", "accept": false, - "reason": "Different model families and symptoms: Qwen3Next missing projections vs Qwen3.5 MoE from_pretrained error." + "reason": "Both involve Qwen-family models, but the reported failures are different model/config loading problems, not the same concrete bug." + }, + { + "left": "issue:42673", + "right": "issue:43582", + "accept": false, + "reason": "VRAM leakage in Qwen3ForCausalLM is unrelated to an Apple Silicon TypeError in caching_allocator_warmup." }, { "left": "issue:43408", "right": "issue:43742", "accept": false, - "reason": "Unrelated models and code paths: SAM3 tracker warning vs MobileLLM key error." + "reason": "A SAM tracker/model-type warning is not the same as a MobileLLM key error during loading." }, { "left": "issue:42673", - "right": "issue:43582", + "right": "issue:43577", "accept": false, - "reason": "VRAM leak in Qwen3ForCausalLM threads is not the same as an Apple Silicon TypeError in allocator warmup." + "reason": "Threaded VRAM leakage and Blip2 dtype propagation are different runtime issues." }, { "left": "issue:42222", "right": "issue:45542", "accept": false, - "reason": "VitPose model breakage is unrelated to the TensorBoard-without-TensorFlow backend error." - }, - { - "left": "issue:42673", - "right": "issue:43577", - "accept": false, - "reason": "Different issue class: multi-thread VRAM leak vs Blip2 dtype propagation bug." + "reason": "VitPose model breakage and missing TensorFlow backend handling are unrelated failures." }, { "left": "issue:42371", "right": "issue:44261", "accept": false, - "reason": "TF32 settings warning is unrelated to MLA layernorm epsilon precision behavior." + "reason": "TF32 API warnings and MLA rms_norm_eps precision drift are different configuration/precision issues." }, { "left": "issue:43937", "right": "issue:44496", "accept": false, - "reason": "GLM-5 generation config validation differs from an unrecognized-model/config.json loading error." + "reason": "One is invalid GenerationConfig handling; the other is missing model_type during model resolution." }, { "left": "issue:38617", "right": "issue:44220", "accept": false, - "reason": "ImportError for configuration_utils is unrelated to an audio fbank feature extraction bug." + "reason": "ImportError for a config utility and a speech feature extraction bug do not share a code path." }, { - "left": "issue:44483", - "right": "issue:45464", + "left": "issue:42175", + "right": "issue:42222", "accept": false, - "reason": "Both involve chat APIs, but the concrete failures differ: generic v5.3 request rejection vs Qwen3.5 streaming inference failure." + "reason": "Backend dependency installation and VitPose model breakage are not the same bug." }, { - "left": "issue:42175", - "right": "issue:42222", + "left": "issue:44483", + "right": "issue:45464", "accept": false, - "reason": "TensorFlow dependency packaging issue is not the same as broken VitPose models." + "reason": "Both mention chat/completions, but one is a general API acceptance problem and the other is a streaming inference failure on a specific model." }, { "left": "issue:42617", "right": "issue:45341", "accept": false, - "reason": "3d_parallel.py runtime failure is unrelated to a testing_utils bug." + "reason": "A script execution failure and a testing_utils bug are unrelated." }, { "left": "issue:44933", "right": "issue:45278", "accept": false, - "reason": "45278 is a broad import-regression report, but 44933 is a specific missing import from image_utils; not enough evidence of the same underlying bug." + "reason": "The first is a specific missing import from image_utils; the second is a broad post-upgrade import-regression report, so they are not clearly the same bug." }, { "left": "issue:44484", "right": "issue:45478", "accept": false, - "reason": "save_pretrained shard-size behavior is unrelated to Qwen3.5 MoE loading failure." + "reason": "Sharding-size behavior in save_pretrained and a Qwen3.5-Moe from_pretrained error are unrelated." }, { "left": "issue:43066", "right": "issue:43122", "accept": false, - "reason": "Both are tokenizer regressions, but they describe different concrete problems: decoder type vs tokenization output drift." + "reason": "Tokenizer decoder type mismatch and changed tokenization output are related in theme but not the same concrete failure." }, { "left": "issue:43673", "right": "issue:45341", "accept": false, - "reason": "Generation cache regression is unrelated to a testing_utils bug." + "reason": "Generation cache handling in v5.0.0 and a testing utility bug are unrelated." }, { "left": "issue:43097", "right": "issue:44493", "accept": false, - "reason": "Removal of tie_embeddings_and_encoder_decoder is not the same as unexpected position-id keys." + "reason": "Removing tie_embeddings_and_encoder_decoder is a config/API change, while unexpected position-id keys are a separate model loading warning." }, { "left": "issue:44410", "right": "issue:45478", "accept": false, - "reason": "Different Qwen models and different failure modes; no clear shared code-path." + "reason": "Both are Qwen-related, but they concern different model families and different load-time failures." }, { "left": "issue:42222", "right": "issue:45459", "accept": false, - "reason": "VitPose model breakage and protobuf-related tokenizer error are unrelated." + "reason": "VitPose model breakage and tokenizer protobuf error masking are unrelated." }, { "left": "issue:43874", "right": "issue:44261", "accept": false, - "reason": "Missing image patch-count method is unrelated to MLA RMS norm epsilon precision behavior." + "reason": "An image processor method omission is not the same as an MLA normalization-precision bug." }, { "left": "issue:43441", "right": "issue:44829", "accept": false, - "reason": "FlashAttention failures in different models/training contexts are not the same concrete bug." + "reason": "FlashAttention failure on Ministral-3 and degenerate training with flash_attention_3 are distinct code-path problems." }, { "left": "issue:44263", "right": "issue:44485", "accept": false, - "reason": "torch.split return handling in GlmMoeDsaIndexer is not the same as GLM-5 RoPE implementation concerns." + "reason": "A torch.split return-value issue is not the same as a GLM-5 RoPE implementation discussion." }, { "left": "issue:43012", "right": "issue:43408", "accept": false, - "reason": "bfloat16 precision warning is unrelated to the SAM3 tracker warning." + "reason": "A bfloat16 compilation warning and a SAM model-type warning are unrelated." }, { "left": "issue:43986", "right": "issue:44484", "accept": false, - "reason": "AutoProcessor video/torchvision crash is unrelated to max_shard_size semantics." + "reason": "AutoProcessor without torchvision and save_pretrained shard sizing are different areas entirely." }, { "left": "issue:41628", "right": "issue:45071", "accept": false, - "reason": "AutoImageProcessor import failure and PretrainedConfig type-checking regression are different breakages." + "reason": "An AutoImageProcessor import error and a PreTrainedConfig type-checking regression do not share the same underlying bug." }, { "left": "issue:45071", "right": "issue:45542", "accept": false, - "reason": "PretrainedConfig typing regression is unrelated to the TensorBoard/TensorFlow backend issue." + "reason": "PreTrainedConfig type checking and missing TensorFlow backend detection are unrelated." }, { "left": "issue:45003", "right": "issue:45341", "accept": false, - "reason": "sys.modules access bug in modeling_utils is unrelated to testing_utils." + "reason": "Unsafe sys.modules access in modeling_utils and a testing_utils bug are not duplicates." }, { "left": "issue:43931", "right": "issue:44410", "accept": false, - "reason": "Different model-loading errors for different model families; not the same bug." + "reason": "Qwen3-VL weight-shape mismatch and missing projections in qwen3next are different model-loading failures." }, { "left": "issue:44117", "right": "issue:44987", "accept": false, - "reason": "Tokenizer mapping None-handling may be adjacent, but 44987 is a broad model-loading failure without enough concrete overlap." + "reason": "A tokenizer mapping None/assumption bug is not the same as a package load failure for physical-intelligence/fast." }, { "left": "issue:43097", "right": "issue:43452", "accept": false, - "reason": "Embedding-tie removal and gguf_file loading breakage are unrelated." + "reason": "Removing a config attribute and gguf_file handling breakage are unrelated." }, { "left": "issue:43986", "right": "issue:45341", "accept": false, - "reason": "Different files and symptoms: AutoProcessor/video dependency crash vs testing_utils bug." + "reason": "Video-model loading without torchvision and a testing utility bug are unrelated." }, { "left": "issue:43931", "right": "issue:44038", "accept": false, - "reason": "Both concern Qwen3-VL, but one is weight-shape mismatch and the other is a broader transformers 5.0 bug report; not clearly the same fix." + "reason": "A Qwen3-VL weight-shape mismatch and a Qwen3-VL-Moe bug are different failures." }, { "left": "issue:43335", "right": "issue:45071", "accept": false, - "reason": "SwitchTransformers sparse-layer construction is unrelated to PretrainedConfig type checking." + "reason": "A SwitchTransformers sparse-layer construction bug is unrelated to PreTrainedConfig type checking." }, { "left": "issue:43408", "right": "issue:44368", "accept": false, - "reason": "SAM3 tracker/model-type warning is unrelated to the tie_word_embeddings warning in Qwen3.5 fine-tuning." + "reason": "A sam3 tracker/model-type warning and a tie_word_embeddings warning in Qwen3.5 are not the same issue." }, { "left": "issue:43874", "right": "issue:44263", "accept": false, - "reason": "Missing image patch-count API and torch.split return handling are separate model bugs." + "reason": "An image processor helper method missing and a torch.split return-value bug are unrelated." } ] }, @@ -11975,18 +11959,18 @@ "issue:38617|issue:42757", "issue:44261|issue:44485", "issue:44410|issue:45375", - "issue:44038|issue:44410", "issue:38617|issue:44351", - "issue:44704|issue:45103", + "issue:44038|issue:44410", "issue:44062|issue:44117", + "issue:44704|issue:45103", "issue:43986|issue:44117", "issue:43502|issue:43992", "issue:43097|issue:43873", "issue:43992|issue:44484", "issue:43824|issue:43986", "issue:43066|issue:43906", - "issue:43352|issue:44380", "issue:43825|issue:44655", + "issue:43352|issue:44380", "issue:38617|issue:43097", "issue:43097|issue:43901", "issue:43012|issue:43550", @@ -12002,12 +11986,12 @@ "issue:44483|issue:44514", "issue:43618|issue:44380" ], - "prepared_review_unit_hash": "07284759cd556d57c9b9eacff4acef514bf2160d5e0752d5cb9ff4acc0b10fac", + "prepared_review_unit_hash": "ee2bd8d3a5159d2ec7e9ddbc220dfa194d25d77ef5cc67fe0eecb8438538b695", "component_budget": { "node_count": 367, "item_count": 367, "soft_pair_count": 1370, - "serialized_chars": 706987, + "serialized_chars": 706986, "estimated_input_tokens": 176747, "estimated_eval_tokens": 353750 }, @@ -12015,225 +11999,225 @@ "node_count": 48, "item_count": 48, "soft_pair_count": 34, - "serialized_chars": 37827, - "estimated_input_tokens": 9457, - "estimated_eval_tokens": 19170 + "serialized_chars": 37815, + "estimated_input_tokens": 9454, + "estimated_eval_tokens": 19164 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The items are largely heterogeneous. The soft-similarity pairs share broad subsystems or naming patterns, but none look like the same concrete bug/change, so no soft edges should be merged.", - "confidence": 0.57, + "summary": "The cluster is mostly a grab bag of distinct Transformers bugs/regressions (imports, model configs, loaders, training, and docs). The soft edges are all superficial similarities rather than true duplicates.", + "confidence": 0.96, "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "issue:43873 is the strongest standalone representative: it is open, has active discussion, and describes a concrete, user-visible bug in quantization/offloading behavior.", + "best_issue_reason": "issue:43097 is the closest cluster anchor because it is a broad v5 API regression with some inbound references, but it still does not subsume the other reports.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44829", "right": "issue:45245", "accept": false, - "reason": "Different failure modes: flash_attention_3 training collapse vs category-count runtime limit." + "reason": "Different failures: flash-attention training degeneration vs category-cardinality runtime error." }, { "left": "issue:43976", "right": "issue:45071", "accept": false, - "reason": "Python/version compatibility break vs PretrainedConfig type-checking regression; unrelated causes." + "reason": "Python version compatibility problem vs PretrainedConfig type-checking regression." }, { "left": "issue:43986", "right": "issue:45372", "accept": false, - "reason": "Different import-time crashes: missing torchvision vs missing ReasoningEffort dependency." + "reason": "Different loading paths and missing dependencies: torchvision vs mistral_common import." }, { "left": "issue:43531", "right": "issue:45341", "accept": false, - "reason": "Qwen3-MoE sliding_window bug vs a testing_utils issue; no shared code path." + "reason": "Unrelated bugs in Qwen3-MoE sliding windows and testing utilities." }, { "left": "issue:43064", "right": "issue:44315", "accept": false, - "reason": "FSDP2/PEFT optimizer-state corruption vs Liger Kernel not applied with model_init; distinct training paths." + "reason": "Different training issues: FSDP2 optimizer state corruption vs Liger kernel not applied." }, { "left": "issue:38617", "right": "issue:42757", "accept": false, - "reason": "Both are import errors, but for different symbols/modules and different underlying breakages." + "reason": "Both are import errors, but for different symbols and different modules." }, { "left": "issue:44261", "right": "issue:44485", "accept": false, - "reason": "MLA norm epsilon precision bug vs GLM-5 RoPE implementation question; not the same defect." + "reason": "Different model-specific bugs: RMS norm precision vs GLM-5 RoPE implementation." }, { "left": "issue:44410", "right": "issue:45375", "accept": false, - "reason": "Both mention Qwen models, but one is missing attention projections and the other is a dropped config field." - }, - { - "left": "issue:44038", - "right": "issue:44410", - "accept": false, - "reason": "Different Qwen model bugs: version-specific VL-MoE failure vs missing projections in qwen3next." + "reason": "Both involve Qwen-related configs, but the missing projections and missing config field are separate bugs." }, { "left": "issue:38617", "right": "issue:44351", "accept": false, - "reason": "Different import failures from transformers: missing layer_type_validation vs missing HybridCache." + "reason": "Separate missing-export import errors with different affected APIs." }, { - "left": "issue:44704", - "right": "issue:45103", + "left": "issue:44038", + "right": "issue:44410", "accept": false, - "reason": "cached_file kwarg forwarding bug vs auto_docstring crash under future annotations; unrelated." + "reason": "Different Qwen model families and different broken code paths." }, { "left": "issue:44062", "right": "issue:44117", "accept": false, - "reason": "Tokenizer AddedToken kwarg conflict vs TOKENIZER_MAPPING_NAMES returning None; different loading bugs." + "reason": "Tokenizer construction error vs tokenizer-mapping None handling; not the same defect." + }, + { + "left": "issue:44704", + "right": "issue:45103", + "accept": false, + "reason": "AutoProcessor forwarding bug vs auto_docstring crash; unrelated subsystems." }, { "left": "issue:43986", "right": "issue:44117", "accept": false, - "reason": "Processor/video dependency crash vs tokenizer mapping assumption bug; not the same issue." + "reason": "Video-model loading dependency issue vs tokenizer mapping assumptions." }, { "left": "issue:43502", "right": "issue:43992", "accept": false, - "reason": "local_files_only network leak vs missing embed_tokens.weight in UMT5Encoder; different pre-trained loading paths." + "reason": "Offline/local-file loading leak vs missing encoder weights on load." }, { "left": "issue:43097", "right": "issue:43873", "accept": false, - "reason": "tie_embeddings deprecation/removal warning issue vs quantization offloading malfunction; unrelated." + "reason": "Broad v5 API removal vs quantization/offloading behavior; different problems." }, { "left": "issue:43992", "right": "issue:44484", "accept": false, - "reason": "Missing weight loading vs shard-size default question; no shared bug." + "reason": "Load-state mismatch vs save_pretrained shard-size default question." }, { "left": "issue:43824", "right": "issue:43986", "accept": false, - "reason": "Two different processor/model import errors with different missing dependencies/symbols." + "reason": "Different import/load failures affecting different processor/model paths." }, { "left": "issue:43066", "right": "issue:43906", "accept": false, - "reason": "Wrong tokenizer decoder type vs an isolated reproduction reference; not enough evidence of same bug." + "reason": "Tokenizer decoder-type regression vs isolated reproduction of another issue." }, { - "left": "issue:43352", - "right": "issue:44380", + "left": "issue:43825", + "right": "issue:44655", "accept": false, - "reason": "Different attention-backend problems in different models; same broad area, not same defect." + "reason": "Pipeline error-message bug vs pipeline save_pretrained failure." }, { - "left": "issue:43825", - "right": "issue:44655", + "left": "issue:43352", + "right": "issue:44380", "accept": false, - "reason": "Pipeline error-message wording vs saving Pipeline objects; different functionality." + "reason": "Different attention-backend issues on different models and with different symptoms." }, { "left": "issue:38617", "right": "issue:43097", "accept": false, - "reason": "ImportError for a removed helper vs tie-embeddings warning/removal; unrelated." + "reason": "ImportError for a symbol vs v5 embedding-link API removal; unrelated." }, { "left": "issue:43097", "right": "issue:43901", "accept": false, - "reason": "Embedding-tie warning removal vs text-classification docs mismatch; different surfaces." + "reason": "Core API removal vs documentation mismatch for return_all_scores." }, { "left": "issue:43012", "right": "issue:43550", "accept": false, - "reason": "bfloat16 compile warning vs Bamba torch.compile SDPA failure; distinct issues." + "reason": "Compilation warning on bfloat16 vs torch.compile SDPA failure; distinct bugs." }, { "left": "issue:43012", "right": "issue:44368", "accept": false, - "reason": "Two different warning conditions with different configs and models; not duplicates." + "reason": "Both are warnings, but about different APIs and causes." }, { "left": "issue:42886", "right": "issue:43381", "accept": false, - "reason": "Offline cache loading bug vs gradient checkpointing eval-mode restriction; unrelated." + "reason": "Offline cache loading bug vs gradient-checkpointing eval-mode limitation." }, { "left": "issue:43208", "right": "issue:43550", "accept": false, - "reason": "xLSTM training blockers vs Bamba compile/SDPA failure; different model-specific bugs." + "reason": "xLSTM training blockers vs Bamba compile/SDPA failure." }, { "left": "issue:42907", "right": "issue:44655", "accept": false, - "reason": "Saving dequantized models vs saving Pipeline objects; same verb, different code paths." + "reason": "Both involve saving, but dequantized model save and pipeline save are different code paths." }, { "left": "issue:43525", "right": "issue:44117", "accept": false, - "reason": "Missing pad_token_id on Llama4Config vs tokenizer mapping None-handling; not the same defect." + "reason": "Missing config attribute vs tokenizer-mapping None handling." }, { "left": "issue:44380", "right": "issue:45479", "accept": false, - "reason": "Attention scaling ignored in backend kernels vs degenerate zero loss for num_labels=1; unrelated training bugs." + "reason": "Attention scaling backend regression vs degenerate classification loss; not the same bug." }, { "left": "issue:43352", "right": "issue:43976", "accept": false, - "reason": "Model backend support error vs Python version compatibility issue; unrelated." + "reason": "Model/backend compatibility issue vs Python version packaging issue." }, { "left": "issue:43064", "right": "issue:44380", "accept": false, - "reason": "FSDP2/PEFT optimizer-state bug vs GPT2 backend scaling bug; different code paths." + "reason": "FSDP optimizer-state corruption vs GPT2 attention scaling ignored in backends." }, { "left": "issue:43208", "right": "issue:43526", "accept": false, - "reason": "xLSTM training blockers vs BeitImageProcessorFast label-reduction bug; unrelated." + "reason": "Different model families and unrelated training vs image-processor bugs." }, { "left": "issue:44483", "right": "issue:44514", "accept": false, - "reason": "API request validation issue vs batched chat-template crash; different endpoints and failures." + "reason": "Chat-completions request rejection vs batched processor crash; different endpoints and causes." }, { "left": "issue:43618", "right": "issue:44380", "accept": false, - "reason": "CLIPOutput attentions regression vs GPT2 backend scaling bug; not the same bug." + "reason": "CLIPOutput attention field regression vs GPT2 backend scaling bug." } ] }, @@ -12322,14 +12306,14 @@ "issue:36683|issue:43475", "issue:43867|issue:45216", "issue:43352|issue:43937", - "issue:44655|issue:45478", "issue:43066|issue:45381", + "issue:44655|issue:45478", "issue:44077|issue:44380", "issue:36683|issue:44077", - "issue:44263|issue:45440", "issue:43064|issue:44492", - "issue:43976|issue:44938", + "issue:44263|issue:45440", "issue:36683|issue:43525", + "issue:43976|issue:44938", "issue:36010|issue:42757", "issue:44188|issue:44380", "issue:43618|issue:45216", @@ -12342,12 +12326,12 @@ "issue:43867|issue:45375", "issue:36683|issue:43064" ], - "prepared_review_unit_hash": "cb52e3775e2fd37e5f1d422249b6bccc7d6db96f5c8942935353fbe313b7c9b3", + "prepared_review_unit_hash": "df347db9ef5a6ccc91ac01a77eca053f9b3401fc984ca19156423449a139f8d0", "component_budget": { "node_count": 367, "item_count": 367, "soft_pair_count": 1370, - "serialized_chars": 706987, + "serialized_chars": 706986, "estimated_input_tokens": 176747, "estimated_eval_tokens": 353750 }, @@ -12355,7 +12339,7 @@ "node_count": 48, "item_count": 48, "soft_pair_count": 34, - "serialized_chars": 37421, + "serialized_chars": 37423, "estimated_input_tokens": 9356, "estimated_eval_tokens": 18968 }, @@ -12364,229 +12348,202 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Mostly a grab-bag of superficially similar Transformers regressions and import/load errors. Only one soft pair looks like a true duplicate: the two identical Qwen3.5 MoE from_pretrained bug reports. The rest share broad themes (imports, model loading, cache, vision, compile, v5 regressions) but appear to be distinct underlying issues.", - "confidence": 0.93, - "canonical_issue_reason": "issue:45310 is the clearest canonical issue candidate: it is an exact duplicate of issue:45478 and describes a concrete, user-facing Qwen3.5 MoE from_pretrained regression.", + "summary": "Mostly a mixed Transformers regression set rather than one tight duplicate cluster. issue:45278 is a reasonable umbrella anchor for the broad import-error reports, the CLIPVisionModel v5 output regressions (43618/43761) look like the clearest duplicate-like pair, and there is also a related Qwen3.5 MoE save/load mini-cluster (45216/45310/45478) that seems thematically close but not provably identical from the snippets alone.", + "confidence": 0.41, + "canonical_issue_reason": "issue:45278 is the broadest open umbrella in the packet and best fits the general post-upgrade import-regression theme without overcommitting to a narrower model-specific bug.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45310 is the best representative issue here because it matches an exact duplicate and names a specific reproducible bug rather than a broad symptom cluster.", + "best_issue_reason": "issue:45278 is the most useful triage anchor because it is open, broad, and central to the import-error family represented in the packet.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43618", "right": "issue:43761", - "accept": false, - "reason": "Both are v5 regression reports, but one is CLIP attentions missing and the other is CLIPVisionModel hidden_states missing; different code paths and models." + "accept": true, + "reason": "Both point to the same CLIPVisionModel v5 forward-path regression where optional outputs are dropped from the returned object." }, { "left": "issue:36683", "right": "issue:43582", "accept": false, - "reason": "Gemma3Config missing vocab_size is unrelated to the Apple Silicon caching_allocator_warmup TypeError." + "reason": "Different failures and subsystems: Gemma3Config vocab_size vs an Apple Silicon allocator TypeError." }, { "left": "issue:44483", "right": "issue:45290", "accept": false, - "reason": "One is a /v1/chat/completions request rejection; the other is an apply_chat_template crash on assistant tool-call messages." + "reason": "One is chat/completions request validation; the other is apply_chat_template handling of tool-call assistant messages." }, { "left": "issue:43012", "right": "issue:44683", "accept": false, - "reason": "A bfloat16 compile warning and a compiled flex_attention failure are different bugs." + "reason": "Both are PyTorch/attention-related, but the reported bugs are different: a bfloat16 warning versus a flex_attention failure on newer torch." }, { "left": "issue:44038", "right": "issue:44230", "accept": false, - "reason": "Qwen3-VL-Moe loading breakage is not the same as an FP8 inference support request." + "reason": "Same model family, but one is a loading regression and the other is fp8 inference support; not the same concrete bug." }, { "left": "issue:42757", "right": "issue:43824", "accept": false, - "reason": "Both are import errors, but for different symbols in different packages/modules." + "reason": "Both are import errors, but for different symbols and likely different breakpoints." }, { "left": "issue:44483", "right": "issue:45381", "accept": false, - "reason": "A chat-completions API rejection is unrelated to the qwen2.5-vl video vision_position_ids bug." + "reason": "Different API surfaces and symptoms: request acceptance vs Qwen2.5-VL video position-id handling." }, { "left": "issue:43299", "right": "issue:43867", "accept": false, - "reason": "Qwen3-VL-Moe loading failure and a sorted-state_dict load error are distinct loading issues." + "reason": "Both are model-loading issues, but one is a Qwen3-VL-Moe regression and the other is a state_dict sorting/load-order problem." }, { "left": "issue:38617", "right": "issue:43502", "accept": false, - "reason": "ImportError for layer_type_validation is unrelated to local_files_only still triggering network requests." + "reason": "Unrelated problems: missing configuration_utils symbol versus local_files_only still triggering network access." }, { "left": "issue:43756", "right": "issue:44485", "accept": false, - "reason": "Smollm3 RoPE layer dropping and GLM-5 RoPE implementation discussion are different model-specific concerns." + "reason": "Different models and different concerns: Smollm3 RoPE-layer dropping versus GLM-5 RoPE discussion/regression." }, { "left": "issue:43901", "right": "issue:44655", "accept": false, - "reason": "Docs mentioning return_all_scores and pipeline save_pretrained failure are not the same bug." + "reason": "Docs/behavior mismatch for return_all_scores is separate from Pipeline.save_pretrained serialization failure." }, { "left": "issue:42994", "right": "issue:44655", "accept": false, - "reason": "Quantized model saving failure is broader and different from inability to save Pipeline objects." + "reason": "Both involve saving, but quantized-model saving and pipeline serialization are different code paths." }, { "left": "issue:36683", "right": "issue:43475", "accept": false, - "reason": "Missing vocab_size on Gemma3Config is a different attribute bug than missing fpn_position_embeddings on Sam3VisionEncoderOutput." + "reason": "Distinct attribute errors on different model configs and different forward paths." }, { "left": "issue:43867", "right": "issue:45216", "accept": false, - "reason": "A sorted state_dict loading error is not the same as an incorrect saved checkpoint from save_pretrained." + "reason": "A state_dict-order load error is not the same as a Qwen3.5 save_pretrained regression." }, { "left": "issue:43352", "right": "issue:43937", "accept": false, - "reason": "Flash Attention 2 unsupported on Nemotron is unrelated to GLM-5 GenerationConfig invalidation." + "reason": "Nemotron FlashAttention support gating and GLM-5 GenerationConfig validation are unrelated." }, { - "left": "issue:44655", - "right": "issue:45478", + "left": "issue:43066", + "right": "issue:45381", "accept": false, - "reason": "Pipeline save_pretrained failure and Qwen3.5 MoE from_pretrained error are different operations and bugs." + "reason": "Wrong tokenizer decoder type and wrong Qwen2.5-VL video position ids are separate bugs." }, { - "left": "issue:43066", - "right": "issue:45381", + "left": "issue:44655", + "right": "issue:45478", "accept": false, - "reason": "Wrong tokenizer decoder type and wrong vision_position_ids affect different subsystems." + "reason": "Pipeline save_pretrained and Qwen3.5 MoE from_pretrained failures do not look like the same underlying issue." }, { "left": "issue:44077", "right": "issue:44380", "accept": false, - "reason": "post_init policy change for patchtsmixer is unrelated to GPT2 attention scaling under SDPA/FlashAttention." + "reason": "PatchTSMixer post_init policy and GPT2 attention scaling under SDPA/FlashAttention are different changes." }, { "left": "issue:36683", "right": "issue:44077", "accept": false, - "reason": "Gemma3Config attribute missing and patchtsmixer post_init validation are different issues." - }, - { - "left": "issue:44263", - "right": "issue:45440", - "accept": false, - "reason": "torch.split return values in GlmMoeDsaIndexer and DeepseekV3MoE divergence are unrelated model internals." + "reason": "Gemma3Config vocab_size and patchtsmixer post_init policy are unrelated." }, { "left": "issue:43064", "right": "issue:44492", "accept": false, - "reason": "Wrong optimizer states on ranks and a typo in cache strategies are unrelated." + "reason": "A distributed optimizer-state bug and a cache-strategy typo are not the same issue." }, { - "left": "issue:43976", - "right": "issue:44938", + "left": "issue:44263", + "right": "issue:45440", "accept": false, - "reason": "Python 3.9/3.10 compatibility complaint and Python 3.14 load failure are different environment regressions." + "reason": "Different model implementation problems: GlmMoeDsaIndexer split behavior versus DeepseekV3MoE divergence." }, { "left": "issue:36683", "right": "issue:43525", "accept": false, - "reason": "Gemma3Config missing vocab_size and Llama4Config missing pad_token_id are separate config attribute bugs." + "reason": "Different config attribute regressions on different models." + }, + { + "left": "issue:43976", + "right": "issue:44938", + "accept": false, + "reason": "Both concern Python compatibility, but they describe different version constraints and failures." }, { "left": "issue:36010", "right": "issue:42757", "accept": false, - "reason": "Different import failures from different symbols and packages; not the same underlying bug." + "reason": "Both are import errors, but they involve different missing symbols and likely unrelated fixes." }, { "left": "issue:44188", "right": "issue:44380", "accept": false, - "reason": "allow_is_bidirectional_skip branching under torch.compile is unrelated to GPT2 attention scaling." + "reason": "Attention/compile behavior is the common area, but the concrete bugs are different." }, { "left": "issue:43618", "right": "issue:45216", "accept": false, - "reason": "Missing CLIP attentions and Qwen3.5 saved-checkpoint corruption are different regressions." + "reason": "CLIPVisionModel output regression and Qwen3.5 save_pretrained regression are unrelated." }, { "left": "issue:43992", "right": "issue:45310", "accept": false, - "reason": "UMT5Encoder missing embed_tokens.weight is unrelated to the Qwen3.5 MoE from_pretrained error." + "reason": "UMT5Encoder missing embed_tokens.weight is not the same bug as Qwen3.5 Moe from_pretrained failure." }, { "left": "issue:36683", "right": "issue:44242", "accept": false, - "reason": "A missing config attribute and load balancing loss not being added are different bugs." + "reason": "Gemma3Config vocab_size and missing load-balancing loss are separate model issues." }, { "left": "issue:43010", "right": "issue:45278", "accept": false, - "reason": "AppleSilicon caching_allocator_warmup TypeError is unrelated to broad import errors after upgrade." + "reason": "Cache update no_grad decoration is a targeted API behavior change, not the broad import-regression cluster." }, { "left": "issue:36683", "right": "issue:43352", "accept": false, - "reason": "Missing Gemma3Config vocab_size is unrelated to Flash Attention 2 support on Nemotron." - }, - { - "left": "issue:43352", - "right": "issue:43931", - "accept": false, - "reason": "Flash Attention unsupported and Qwen3-VL-30B-A3B weight-shape mismatch are different issues." - }, - { - "left": "issue:41669", - "right": "issue:44246", - "accept": false, - "reason": "Import * slowdown and intermittent import transformers slowness are related only at a very broad level, not the same concrete bug." - }, - { - "left": "issue:43867", - "right": "issue:45375", - "accept": false, - "reason": "A sorted state_dict loading error is unrelated to missing deepstack_visual_indexes being dropped by strict config handling." - }, - { - "left": "issue:36683", - "right": "issue:43064", - "accept": false, - "reason": "Gemma3Config missing vocab_size and FSDP2/PEFT optimizer-state mismatch are unrelated." - }, - { - "left": "issue:45310", - "right": "issue:45478", - "accept": true, - "reason": "Exact same Qwen3.5 MoE from_pretrained regression title/topic; these are duplicate reports of the same bug." + "reason": "Unrelated config attribute error versus FlashAttention support gating." } ] }, - "evaluator_result": null, - "evaluator_used": false, - "retried": false, - "accepted_nontrivial_soft_edge": false, + "evaluator_result": { + "accept": false, + "feedback": "The broad anchor choice (issue:45278) is grounded, and the CLIPVisionModel pair looks plausibly duplicate. However, the summary is a bit too strong: the packet also contains a closely related Qwen3.5 MoE save/load cluster (e.g. 45216, 45310, 45478) that makes \"the only plausible duplicate pair\" overconfident. Several soft-edge rejections also read as more definitive than the packet supports, given that we only have titles/body snippets and no diff evidence. Please keep the cluster characterization more conservative." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null }, @@ -12664,24 +12621,24 @@ "issue:36296|issue:43575", "issue:43867|issue:44484", "issue:43618|issue:44857", - "issue:28282|issue:33453", "issue:42890|issue:44964", "issue:43618|issue:44977", + "issue:28282|issue:33453", "issue:42890|issue:44112", "issue:43010|issue:44075", "issue:43723|issue:44556", "issue:42890|issue:45341", "issue:43502|issue:44062", - "issue:44387|issue:44556", "issue:44222|issue:45125", + "issue:44387|issue:44556", "issue:43010|issue:43065", "issue:43010|issue:43232", "issue:45125|issue:45478", "issue:42548|issue:44351", "issue:45125|issue:45375", "issue:44355|issue:44749", - "issue:30990|issue:31795", "issue:43502|issue:44485", + "issue:30990|issue:31795", "issue:44075|issue:44871", "issue:44315|issue:44360", "issue:43010|issue:44368", @@ -12689,12 +12646,12 @@ "issue:43502|issue:44222", "issue:43299|issue:44556" ], - "prepared_review_unit_hash": "ad6be4b0290bf84acaffad84453ed96e991b2a5c8aa1551741c554dcfdd7b75f", + "prepared_review_unit_hash": "d8e2723ad30220130109499c7e2ef6790a42457d573c6bcd29b0f0e53e069fb1", "component_budget": { "node_count": 367, "item_count": 367, "soft_pair_count": 1370, - "serialized_chars": 706987, + "serialized_chars": 706986, "estimated_input_tokens": 176747, "estimated_eval_tokens": 353750 }, @@ -12702,7 +12659,7 @@ "node_count": 48, "item_count": 48, "soft_pair_count": 35, - "serialized_chars": 37642, + "serialized_chars": 37644, "estimated_input_tokens": 9411, "estimated_eval_tokens": 19078 }, @@ -12711,222 +12668,222 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly heterogeneous: the soft links are similarity-only and do not describe the same underlying bug or change. I would reject all soft edges. The most representative issue is the broad v5 loading-regression report #43723, but there is no true canonical duplicate set here.", - "confidence": 0.71, - "canonical_issue_reason": "#43723 is the broadest, most central loading-regression report in the set; it best matches the common v5 import/load-failure theme, even though the cluster is not a real duplicate group.", + "summary": "The cluster is broad and mostly not duplicate-worthy: the soft pairs are only superficially similar (same subsystem or same release) but do not look like the same underlying bug or change. No PRs are present.", + "confidence": 0.46, + "canonical_issue_reason": "Issue 43723 is the clearest, most generic loading-regression report in the set, with an explicit API target (`AutoTokenizer.from_pretrained`) and a theme that matches several nearby loading-related reports.", "canonical_pr_reason": null, - "best_issue_reason": "#43723 is the best single issue to stand in for the cluster because it is general and user-facing, unlike the many highly specific model/test/version edge cases here.", + "best_issue_reason": "43723 is the best standalone representative of the cluster\u2019s loading/import regression subset: concrete repro, explicit failing path, and broader applicability than the model-specific reports.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:42890", "right": "issue:45561", "accept": false, - "reason": "Both are test-related, but one is a missing set_seed integration issue and the other is a captured_info.txt xdist race; different bugs." + "reason": "Different test failures: missing `set_seed()` flakiness vs xdist race on `captured_info.txt`." }, { "left": "issue:43992", "right": "issue:45478", "accept": false, - "reason": "Both are model-loading failures, but for different models and different failure modes; not the same code-path bug." + "reason": "Different model-loading bugs with different causes; shared `from_pretrained` wording is not enough." }, { "left": "issue:44938", "right": "issue:45071", "accept": false, - "reason": "One is Python 3.14 load failure, the other is PretrainedConfig type checking in v5.4; unrelated regressions." + "reason": "Both are v5 regressions, but one is a Python 3.14 import/load failure and the other is `PretrainedConfig` type checking." }, { "left": "issue:43867", "right": "issue:45357", "accept": false, - "reason": "A state_dict sorting load error is not the same as incorrect visual encoder keys being saved; save/load sides differ." + "reason": "One is a state-dict load ordering issue; the other is incorrect keys saved for Qwen3.5 visual encoder." }, { "left": "issue:43010", "right": "issue:43976", "accept": false, - "reason": "Torch no_grad on cache update and Python version compatibility are unrelated." + "reason": "`@torch.no_grad` on cache updates is unrelated to Python-version compatibility." }, { "left": "issue:44938", "right": "issue:45381", "accept": false, - "reason": "Generic Python 3.14 loading failure and Qwen2.5-VL vision position IDs are different bugs." + "reason": "Python 3.14 loading failure vs Qwen2.5-VL video `vision_position_ids` bug; different code paths." }, { "left": "issue:44360", "right": "issue:45440", "accept": false, - "reason": "Both mention implementation differences, but one is a DSA indexer ReLU complaint and the other is DeepseekV3MoE divergence; not the same defect." + "reason": "DSA indexer activation bug and DeepSeekV3 implementation divergence are unrelated." }, { "left": "issue:43618", "right": "issue:44077", "accept": false, - "reason": "CLIP attentions assignment and patchtsmixer post_init validation are separate model issues." + "reason": "Different issues: missing CLIP attentions assignment vs PatchTSMixer `post_init` API constraint." }, { "left": "issue:36296", "right": "issue:43575", "accept": false, - "reason": "Tensor parallel training bug vs OOM loading a specific large Qwen2.57B model are different concrete problems." + "reason": "Both mention tensor parallelism, but one is a training bug and the other is an OOM while loading a specific model." }, { "left": "issue:43867", "right": "issue:44484", "accept": false, - "reason": "A load-model state_dict issue is unrelated to the default max_shard_size question in save_pretrained()." + "reason": "State-dict loading error vs `save_pretrained` shard-size default; unrelated." }, { "left": "issue:43618", "right": "issue:44857", "accept": false, - "reason": "Different models and different failures: CLIPOutput attentions vs float16 AMP crash." - }, - { - "left": "issue:28282", - "right": "issue:33453", - "accept": false, - "reason": "Missing PyTorch import support and tokenizer-loading regression are not the same bug." + "reason": "CLIP output regression vs AMP float16 crash in a different loss module." }, { "left": "issue:42890", "right": "issue:44964", "accept": false, - "reason": "CI/test seed flakiness does not match a Phi-4 multimodal model loading failure." + "reason": "Test-seed flakiness is unrelated to loading `Phi-4-multimodal-instruct`." }, { "left": "issue:43618", "right": "issue:44977", "accept": false, - "reason": "CLIP attentions field issue and Qwen3.5 flash-attention generation bug are unrelated." + "reason": "Different bugs: CLIP attentions regression vs Qwen3.5 flash-attention generation problem." + }, + { + "left": "issue:28282", + "right": "issue:33453", + "accept": false, + "reason": "PyTorch import requirement error vs tokenizer-loading regression; no shared root cause." }, { "left": "issue:42890", "right": "issue:44112", "accept": false, - "reason": "A flaky SamHQ integration test is not the same as a GraniteSpeech stale device override test." + "reason": "Unrelated CI/test flake and stale device-override test failure." }, { "left": "issue:43010", "right": "issue:44075", "accept": false, - "reason": "Decorator/no_grad cache behavior and SGD argument handling are separate optimizer/model issues." + "reason": "Cache update decoration issue vs SGD optimizer args not being used." }, { "left": "issue:43723", "right": "issue:44556", "accept": false, - "reason": "Both are loading-related, but one is tokenizer loading in v5 and the other is checkpoint reload across versions; too different to merge." + "reason": "Both involve loading, but one is tokenizer loading in v5 and the other is checkpoint reload compatibility after upgrading versions." }, { "left": "issue:42890", "right": "issue:45341", "accept": false, - "reason": "SamHQ test seeding and a bug in testing_utils.py are not the same specific failure." + "reason": "Different test utilities issues with no common underlying bug." }, { "left": "issue:43502", "right": "issue:44062", "accept": false, - "reason": "local_files_only network leakage and AddedToken() keyword duplication are unrelated tokenizer/config bugs." + "reason": "Offline-loading network leak vs `AddedToken(..., special=...)` constructor error are unrelated." }, { - "left": "issue:44387", - "right": "issue:44556", + "left": "issue:44222", + "right": "issue:45125", "accept": false, - "reason": "CUDA memory growth under int4 quantization is not the same as checkpoint reload failure after upgrade." + "reason": "FP8 `save_pretrained` bug vs missing tensor-parallel plan on a Qwen MoE model." }, { - "left": "issue:44222", - "right": "issue:45125", + "left": "issue:44387", + "right": "issue:44556", "accept": false, - "reason": "FP8 save_pretrained behavior and missing tensor-parallel plan are different areas and different bugs." + "reason": "Int4 CUDA memory/OOM regression vs inability to reload checkpoints after version upgrade." }, { "left": "issue:43010", "right": "issue:43065", "accept": false, - "reason": "Torch no_grad on cache update has nothing to do with the dummy Conv2d in Sam3PixelDecoder." + "reason": "`no_grad` on cache updates is unrelated to the dummy Conv2d in Sam3PixelDecoder." }, { "left": "issue:43010", "right": "issue:43232", "accept": false, - "reason": "Cache update decoration and generation kwargs syncing after sync_gpus are different generation-path bugs." + "reason": "Different generation-path bugs: cache update decoration vs `_update_model_kwargs_for_generation` after `sync_gpus`." }, { "left": "issue:45125", "right": "issue:45478", "accept": false, - "reason": "Missing _tp_plan and a from_pretrained error for Qwen3.5 MoE are related only at a high level, not the same concrete bug." + "reason": "Related model family, but different failures: missing `_tp_plan` vs `from_pretrained` loading error in v5.4+." }, { "left": "issue:42548", "right": "issue:44351", "accept": false, - "reason": "Importing PreTrainedModel and importing HybridCache are separate missing-export problems." + "reason": "Missing `PreTrainedModel` import and missing `HybridCache` import are distinct import regressions." }, { "left": "issue:45125", "right": "issue:45375", "accept": false, - "reason": "Tensor-parallel plan missing and a dropped vision config field are unrelated." + "reason": "Tensor-parallel plan issue vs missing `deepstack_visual_indexes` config field; different root causes." }, { "left": "issue:44355", "right": "issue:44749", "accept": false, - "reason": "Compiled Python file errors and a v4.57\u2192v5.3 data slowdown issue are not the same bug." - }, - { - "left": "issue:30990", - "right": "issue:31795", - "accept": false, - "reason": "A loading hang and a confusing docstring are clearly different." + "reason": "Compiled Python file errors vs a performance regression after upgrading from 4.57.3 to 5.3.0." }, { "left": "issue:43502", "right": "issue:44485", "accept": false, - "reason": "Local-files-only network requests and GLM-5 RoPE implementation are unrelated." + "reason": "Offline-loading bug vs GLM-5 RoPE implementation discussion." + }, + { + "left": "issue:30990", + "right": "issue:31795", + "accept": false, + "reason": "Sentence Transformers loading hang and forward-doc confusion are unrelated." }, { "left": "issue:44075", "right": "issue:44871", "accept": false, - "reason": "Unused SGD args and inconsistent eos_token_id config are different features." + "reason": "Unused SGD args vs Gemma-3 `eos_token_id` configuration mismatch." }, { "left": "issue:44315", "right": "issue:44360", "accept": false, - "reason": "Liger kernel application in model_init and a DSA ReLU discussion are unrelated." + "reason": "Model-init Liger-kernel application bug vs DSA indexer activation issue." }, { "left": "issue:43010", "right": "issue:44368", "accept": false, - "reason": "Cache/no_grad decoration and a tie_word_embeddings warning are not the same issue." + "reason": "`no_grad` cache-update bug vs warning about `tie_word_embeddings` in Qwen3.5 LoRA fine-tuning." }, { "left": "issue:42890", "right": "issue:44373", "accept": false, - "reason": "A flaky integration test and a wrong docstring for position_ids are unrelated." + "reason": "Test flake vs wrong docstring for `position_ids`; different severity and code path." }, { "left": "issue:43502", "right": "issue:44222", "accept": false, - "reason": "Network requests despite local_files_only and FP8 save_pretrained moe are different bugs." + "reason": "Local-files-only network bug vs FP8 MoE save/load issue; not the same change." }, { "left": "issue:43299", "right": "issue:44556", "accept": false, - "reason": "Qwen3VL MoE loading breakage and checkpoint reload failure across versions are distinct regressions." + "reason": "Qwen3VL MoE loading breakage and post-upgrade checkpoint reload failure are different bugs." } ] }, @@ -13037,12 +12994,12 @@ "issue:44704|issue:44908", "issue:44246|issue:44273" ], - "prepared_review_unit_hash": "2ff7506432a97cdfed435d4b63f11c244730302d013dc2f31dabfdb2ff54e98a", + "prepared_review_unit_hash": "3a47f5ff87530614cf0d6f95de2ca3042027c662bc4b1e16b1fd8cf5adc3497c", "component_budget": { "node_count": 367, "item_count": 367, "soft_pair_count": 1370, - "serialized_chars": 706987, + "serialized_chars": 706986, "estimated_input_tokens": 176747, "estimated_eval_tokens": 353750 }, @@ -13050,249 +13007,249 @@ "node_count": 47, "item_count": 47, "soft_pair_count": 37, - "serialized_chars": 37406, - "estimated_input_tokens": 9352, - "estimated_eval_tokens": 18960 + "serialized_chars": 37411, + "estimated_input_tokens": 9353, + "estimated_eval_tokens": 18962 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This set is mostly a loose similarity cluster of unrelated Transformers issues. Only the lazy-loading/sys.modules pair looks like a plausible duplicate; the rest are adjacent but distinct regressions, import errors, docs, or model-specific bugs.", - "confidence": 0.57, - "canonical_issue_reason": "issue:43576 is the most central and broadly connected issue in the set, with several soft similarity links to other v5 regression/import-related reports; it is concrete enough to serve as the cluster anchor.", + "summary": "This cluster is mostly a grab-bag of Transformers v5-era regressions and import/loading complaints. Most pairs are only loosely related by subsystem or release, not by the same concrete bug. The one clear duplicate-like edge is the lazy-loading failure versus unsafe `sys.modules` access in `modeling_utils`.", + "confidence": 0.78, + "canonical_issue_reason": "issue:45003 is the most code-path-specific report in the set: it points to an unsafe `sys.modules[]` access in `modeling_utils`, which plausibly explains the broader lazy-loading/import failures around it.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43576 is the best representative issue overall because it is specific, user-facing, and sits closest to the repeated v5 breakage/import-regression theme in this cluster.", + "best_issue_reason": "issue:45003 is the best cluster anchor because it names a concrete mechanism rather than a symptom-only report.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43502", "right": "issue:44961", "accept": false, - "reason": "Different problems: local_files_only network access vs an unrelated placeholder issue." + "reason": "No shared bug; local_files_only networking behavior is unrelated to the placeholder/nonsense racoon issue." }, { "left": "issue:44222", "right": "issue:44230", "accept": false, - "reason": "Both mention FP8/moe, but one is save_pretrained behavior and the other is inference support; not the same bug." + "reason": "Both mention FP8/MoE, but one is save_pretrained and the other is inference support; different code paths." }, { "left": "issue:44355", "right": "issue:44655", "accept": false, - "reason": "Compiled Python file errors are unrelated to pipeline save_pretrained missing files." + "reason": "Compiled Python file errors are unrelated to Pipeline save_pretrained missing files." }, { "left": "issue:44355", "right": "issue:45071", "accept": false, - "reason": "Different failure modes: compiled Python file handling vs PretrainedConfig type checking." + "reason": "No overlap beyond general v5 breakage; compiled-file errors and PretrainedConfig type-checking are different bugs." }, { "left": "issue:42831", "right": "issue:43874", "accept": false, - "reason": "Accuracy drift in FineGrainedFP8 is unrelated to a missing image-patch method AttributeError." + "reason": "Different models and failure modes: FP8 accuracy vs a missing image-patch helper AttributeError." }, { "left": "issue:33453", "right": "issue:34689", "accept": false, - "reason": "Both are loading regressions, but tokenizer loading and Llama 3.2 Vision model loading are different code paths." + "reason": "Both are loading regressions, but they affect different loaders/models and do not look like the same bug." }, { "left": "issue:42757", "right": "issue:43576", "accept": false, - "reason": "Missing huggingface_hub export is a dependency/import issue, not the transformers env command regression." + "reason": "Missing `is_offline_mode` import from huggingface_hub is separate from the v5 env command breakage." }, { "left": "issue:43827", "right": "issue:44961", "accept": false, - "reason": "Docs referencing pipeline() after v5 removals is unrelated to the placeholder issue." + "reason": "Docs referencing removed pipeline APIs is unrelated to the racoon issue." }, { "left": "issue:33453", "right": "issue:36296", "accept": false, - "reason": "Tokenizer loading regression and tensor-parallel training bug are distinct." + "reason": "Tokenizer loading regression and tensor-parallel training bug are unrelated." }, { "left": "issue:43576", "right": "issue:45071", "accept": false, - "reason": "env command breakage and PretrainedConfig type checking are separate v5 regressions." + "reason": "Environment command breakage and `PretrainedConfig` type checking are different regressions." }, { "left": "issue:43576", "right": "issue:43824", "accept": false, - "reason": "Different import/export failures for unrelated symbols." + "reason": "v5 env command issue is unrelated to missing Qwen2.5-VL import/export." }, { "left": "issue:43673", "right": "issue:44492", "accept": false, - "reason": "Missing GenerationMixin cache in chunked_prefill is unrelated to a typo in cache strategies." + "reason": "Cache-missing during chunked prefill is not the same as a typo in cache strategies." }, { "left": "issue:38617", "right": "issue:43576", "accept": false, - "reason": "Different issues: missing layer_type_validation export vs broken env command." + "reason": "ImportError for `layer_type_validation` is a specific missing symbol issue, not the env command regression." }, { "left": "issue:33357", "right": "issue:33453", "accept": false, - "reason": "MacOS bus error with CLIP is not the same as tokenizer loading regression." + "reason": "MacOS CLIP bus error and tokenizer loading regression are not the same underlying bug." }, { "left": "issue:44961", "right": "issue:45468", "accept": false, - "reason": "Placeholder issue versus Gemma-4 audio positional encoding bug." + "reason": "Racoon is not meaningfully connected to Gemma-4 audio positional encoding." }, { "left": "issue:31795", "right": "issue:38617", "accept": false, - "reason": "Documentation confusion and ImportError are unrelated." + "reason": "Documentation confusion about forward args is unrelated to the missing import error." }, { "left": "issue:43576", "right": "issue:44351", "accept": false, - "reason": "Broken env command and missing HybridCache import are different v5 issues." + "reason": "v5 env command breakage and missing `HybridCache` import are distinct issues." }, { "left": "issue:44908", "right": "issue:44961", "accept": false, - "reason": "Scheduler kwargs bug is unrelated to the placeholder issue." + "reason": "Scheduler kwargs bug is unrelated to the racoon issue." }, { "left": "issue:41628", "right": "issue:42548", "accept": false, - "reason": "Both are import errors, but for different missing symbols and likely different causes." + "reason": "Both are import errors, but they concern different top-level symbols and no concrete shared code path is evident." }, { "left": "issue:44485", "right": "issue:44961", "accept": false, - "reason": "RoPE implementation discussion is unrelated to the placeholder issue." + "reason": "GLM-5 RoPE implementation discussion is unrelated to the racoon issue." }, { "left": "issue:36296", "right": "issue:43901", "accept": false, - "reason": "Tensor parallel training bug and pipeline docs mismatch are unrelated." + "reason": "Tensor-parallel training bug and TextClassificationPipeline docs mismatch are unrelated." }, { "left": "issue:36296", "right": "issue:43526", "accept": false, - "reason": "Different model/feature bugs: tensor parallelism vs BeitImageProcessorFast label reduction." + "reason": "Tensor-parallel training bug and BeitImageProcessorFast label reduction are unrelated." }, { "left": "issue:41628", "right": "issue:44355", "accept": false, - "reason": "AutoImageProcessor import failure is unrelated to compiled Python file errors." + "reason": "AutoImageProcessor import failure and compiled Python file errors do not share the same bug." }, { "left": "issue:36296", "right": "issue:43012", "accept": false, - "reason": "Tensor parallel training bug is unrelated to the PyTorch warning during bf16 compilation." + "reason": "Tensor parallel training and bf16 compilation warnings are different problems." }, { "left": "issue:44273", "right": "issue:45003", "accept": true, - "reason": "These look like the same lazy-loading failure: a generic lazy loading breakage with a specific root cause in unsafe sys.modules access from modeling_utils." + "reason": "Both point at lazy-loading/import machinery; `sys.modules[]` access in modeling utils is a plausible root cause for lazy loading not working properly." }, { "left": "issue:44261", "right": "issue:44360", "accept": false, - "reason": "MLA rms_norm_eps precision bug is unrelated to the DSA indexer missing ReLU." + "reason": "Missing `rms_norm_eps` precision drift and a DSA indexer ReLU comment are unrelated." }, { "left": "issue:44273", "right": "issue:45542", "accept": false, - "reason": "Lazy loading problems are unrelated to the TensorFlow backend error caused by tensorboard-only installs." + "reason": "Lazy loading problems and the TensorFlow/tensorboard backend error are different failures." }, { "left": "issue:44623", "right": "issue:45230", "accept": false, - "reason": "processor.save_pretrained missing files is unrelated to a generic bug report." + "reason": "Missing processor save files is not the same as a generic bug report." }, { "left": "issue:43576", "right": "issue:44556", "accept": false, - "reason": "Broken env command and checkpoint reload failures after upgrading to v5 are different problems." + "reason": "Broken env command and checkpoint reload incompatibility after upgrade are different migration issues." }, { "left": "issue:41669", "right": "issue:43673", "accept": false, - "reason": "Import-star performance regression is unrelated to chunked_prefill cache handling." + "reason": "Import-star slow import issue and missing cache in chunked prefill are unrelated." }, { "left": "issue:44393", "right": "issue:45071", "accept": false, - "reason": "Qwen3-VL bounding-box output issues are unrelated to PretrainedConfig type checking." + "reason": "Qwen3-VL bbox output error and PretrainedConfig type checking are unrelated." }, { "left": "issue:44038", "right": "issue:45125", "accept": false, - "reason": "General Qwen3-VL-Moe bug and missing _tp_plan for tensor parallelism are distinct." + "reason": "Both involve Qwen3/VL/MoE, but one is broad v5 support and the other is missing `_tp_plan` for tensor parallelism." }, { "left": "issue:44556", "right": "issue:45092", "accept": false, - "reason": "Both concern v5 checkpoint compatibility, but they describe different models and likely different loading paths." + "reason": "Both are upgrade/loading compatibility reports, but they affect different checkpoint/model scenarios and do not look like one bug." }, { "left": "issue:42371", "right": "issue:44492", "accept": false, - "reason": "TF32 API settings are unrelated to a cache-strategies typo." + "reason": "TF32 settings guidance and a cache-strategy typo are unrelated." }, { "left": "issue:42890", "right": "issue:44908", "accept": false, - "reason": "Test flakiness from missing set_seed is unrelated to inverse_sqrt scheduler kwargs." + "reason": "Test flakiness from missing `set_seed()` is unrelated to the inverse_sqrt scheduler kwargs bug." }, { "left": "issue:44704", "right": "issue:44908", "accept": false, - "reason": "AutoProcessor kwargs forwarding and scheduler kwargs handling are different APIs and bugs." + "reason": "AutoProcessor kwargs forwarding and scheduler kwargs handling are different code paths." }, { "left": "issue:44246", "right": "issue:44273", "accept": false, - "reason": "Slow import timing is unrelated to lazy loading failure." + "reason": "Slow import and lazy loading malfunction are related at a high level, but not the same concrete bug." } ] }, "evaluator_result": { "accept": true, - "feedback": "Summary is grounded in the packet and stays conservative: it treats this as a loose cluster with mostly distinct issues, while only tentatively grouping the lazy-loading/sys.modules pair. The best/canonical issue choice is supported by the number of soft links around issue:43576. No soft-edge verdict overstates the evidence." + "feedback": "Grounded overall: the summary stays close to the packet, and the only accepted soft edge (44273 \u2194 45003) is a conservative, plausible lazy-loading/sys.modules match. The other pair judgments are generally cautious and do not overclaim shared bugs." }, "evaluator_used": true, "retried": false, @@ -13319,13 +13276,13 @@ "issue:37428", "issue:41669", "issue:42371", + "issue:42890", "issue:42913", "issue:43066", "issue:43352", "issue:43408", "issue:43502", "issue:43519", - "issue:43525", "issue:43576", "issue:43606", "issue:43618", @@ -13401,326 +13358,319 @@ "issue:44263|issue:45468", "issue:44261|issue:45468", "issue:43606|issue:45468", - "issue:43704|issue:45341", + "issue:42890|issue:43704", "issue:42913|issue:43066", "issue:44908|issue:44998", "issue:44861|issue:45290", - "issue:44998|issue:45341", "issue:44492|issue:44998", + "issue:44998|issue:45341", "issue:43618|issue:44998", - "issue:44556|issue:44998", - "issue:43525|issue:44861" + "issue:44556|issue:44998" ], - "prepared_review_unit_hash": "5841439d1bdbc4e8ded142bf817260cf93e7329451ec3ab216ce518dad72b59f", + "prepared_review_unit_hash": "5f102cb6cb356432e6903e2d9e4fbfdc7872607a41d0045ea4f6baaca48c7ffe", "component_budget": { "node_count": 367, "item_count": 367, "soft_pair_count": 1370, - "serialized_chars": 706987, + "serialized_chars": 706986, "estimated_input_tokens": 176747, "estimated_eval_tokens": 353750 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 47, - "serialized_chars": 41101, - "estimated_input_tokens": 10276, - "estimated_eval_tokens": 20808 + "soft_pair_count": 46, + "serialized_chars": 40859, + "estimated_input_tokens": 10215, + "estimated_eval_tokens": 20686 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly heterogeneous, but there is a recurring tokenizer/v5 regression theme. None of the soft pairs is strong enough to merge as a true duplicate under strict triage, so I would keep the broad tokenizer-v5 issue as the representative node.", - "confidence": 0.71, - "canonical_issue_reason": "Issue 42913 is the broadest representative of the tokenizer-v5 regression family and best matches the recurring theme across the cluster.", + "summary": "This cluster is a noisy mix of mostly unrelated Transformers v5 regressions. The only clearly recurring subtheme is tokenizer loading/behavior, so I\u2019d use that as the representative issue, but I would reject all soft edges as true duplicates.", + "confidence": 0.36, + "canonical_issue_reason": "Issue 43723 is the clearest and most central representative of the tokenizer-loading regression subcluster, which appears to be the strongest coherent theme in an otherwise noisy cluster.", "canonical_pr_reason": null, - "best_issue_reason": "42913 is the most umbrella-like issue here: it describes a general tokenizer behavior change from v4 to v5, which can plausibly subsume several narrower tokenizer-loading/save-path reports.", + "best_issue_reason": "Issue 43723 is the best standalone representative because it states the bug concretely, maps to the main recurring theme, and is more actionable than the more generic adjacent tokenizer reports.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43408", "right": "issue:44315", "accept": false, - "reason": "Different subsystems and failure modes: model-type mismatch warning vs Liger Kernel application issue." + "reason": "Different problems: model-type mismatch warning vs. Liger Kernel not being applied in model_init." }, { "left": "issue:44297", "right": "issue:45230", "accept": false, - "reason": "45230 is too generic to establish identity with a tokenizer save_pretrained mismatch." + "reason": "One is a specific tokenizer.save_pretrained mismatch; the other is a generic bug report with no concrete shared failure mode." }, { "left": "issue:43576", "right": "issue:44908", "accept": false, - "reason": "Unrelated features: env command breakage vs inverse_sqrt scheduler kwargs handling." + "reason": "Unrelated subsystems: CLI env command vs. inverse_sqrt scheduler kwargs handling." }, { "left": "issue:43352", "right": "issue:44393", "accept": false, - "reason": "Different models and symptoms; one is Flash Attention support, the other is a Qwen3-VL bbox output problem." + "reason": "Different model families and different failures: Flash Attention 2 support vs. Qwen3-VL bounding-box output issues." }, { "left": "issue:43976", "right": "issue:44393", "accept": false, - "reason": "Python version compatibility regression is unrelated to the Qwen3-VL output issue." + "reason": "Python version compatibility regression is unrelated to the Qwen3-VL output bug." }, { "left": "issue:44908", "right": "issue:45071", "accept": false, - "reason": "Scheduler kwargs bug is not the same as PretrainedConfig type-checking breakage." + "reason": "Scheduler kwarg propagation and PretrainedConfig type checking are distinct bugs." }, { "left": "issue:44295", "right": "issue:45230", "accept": false, - "reason": "No evidence the generic bug report matches the position_ids buffer error." + "reason": "Specific position_ids buffer access issue does not match a generic bug report." }, { "left": "issue:41669", "right": "issue:44351", "accept": false, - "reason": "Import-star cleanup and HybridCache import failure are not the same concrete bug." + "reason": "Import-star cleanup/performance issue is not the same as the HybridCache import error." }, { "left": "issue:41669", "right": "issue:45341", "accept": false, - "reason": "Broad import-star performance complaint does not match a testing_utils bug." + "reason": "Broad import cleanup is unrelated to a testing_utils bug." }, { "left": "issue:44492", "right": "issue:44945", "accept": false, - "reason": "Cache strategy typo and pipeline-parallelism output correctness are different issues." + "reason": "A cache-strategy typo is unrelated to incorrect pipeline-parallel LLM output." }, { "left": "issue:41669", "right": "issue:44492", "accept": false, - "reason": "These are only loosely related by general code cleanup; not the same bug." + "reason": "These are different code changes and different symptoms: import cleanup vs. cache-strategy typo." }, { "left": "issue:44315", "right": "issue:44829", "accept": false, - "reason": "Model-init applying Liger Kernel and flash_attention_3 training degeneration are different code paths." + "reason": "Model_init/Liger Kernel application is not the same underlying bug as flash_attention_3 training degeneration." }, { "left": "issue:44393", "right": "issue:45478", "accept": false, - "reason": "Both mention Qwen3.5-MoE, but the reported failures are not the same concrete bug." + "reason": "Same broad area only by model family name; the concrete failure modes are different Qwen bugs." }, { "left": "issue:44077", "right": "issue:44861", "accept": false, - "reason": "Optional post_init validation and _get_tied_weight_keys AttributeError are unrelated." + "reason": "Optional post_init validation and _tied_weights_keys list handling are separate model-internals issues." }, { "left": "issue:43827", "right": "issue:44908", "accept": false, - "reason": "Docs referencing pipeline() and scheduler kwargs handling are unrelated." + "reason": "Docs still referencing pipeline() is unrelated to scheduler kwargs being ignored." }, { "left": "issue:31795", "right": "issue:33453", "accept": false, - "reason": "Confusing forward docs and tokenizer loading regression are not the same underlying problem." + "reason": "Documentation confusion around forward args is not the same as tokenizer loading regression." }, { "left": "issue:41669", "right": "issue:44908", "accept": false, - "reason": "Import-star usage and scheduler behavior are unrelated." + "reason": "Import cleanup is unrelated to scheduler behavior." }, { "left": "issue:43723", "right": "issue:44393", "accept": false, - "reason": "Tokenizer loading in v5 is unrelated to the Qwen3-VL hallucination/bbox issue." + "reason": "Tokenizer loading in v5 is unrelated to Qwen3-VL bounding-box output errors." }, { "left": "issue:44655", "right": "issue:44861", "accept": false, - "reason": "Pipeline save_pretrained failure and tied-weight key AttributeError are different bugs." + "reason": "Pipeline save_pretrained problems do not match _get_tied_weight_keys crashing on a list." }, { "left": "issue:43576", "right": "issue:44273", "accept": false, - "reason": "Broken env command and lazy loading malfunction are different features." + "reason": "transformers env command breakage is not the same as lazy loading failures." }, { "left": "issue:44360", "right": "issue:45468", "accept": false, - "reason": "DSA indexer ReLU discussion does not match Gemma-4 audio positional encoding." + "reason": "A DSA indexer ReLU discussion and Gemma4AudioRelPositionalEncoding are unrelated." }, { "left": "issue:45230", "right": "issue:45310", "accept": false, - "reason": "A generic bug report is insufficient to treat as the same as a specific Qwen3.5-MoE from_pretrained error." + "reason": "A generic bug report cannot be treated as the same concrete from_pretrained error." }, { "left": "issue:44273", "right": "issue:44908", "accept": false, - "reason": "Lazy loading and inverse_sqrt scheduler kwargs are unrelated." + "reason": "Lazy loading and scheduler kwargs are different areas with different failure modes." }, { "left": "issue:41669", "right": "issue:43704", "accept": false, - "reason": "Import-star performance and VRAM leak in dataloader threads are different code paths." + "reason": "Import-star removal performance issue is unrelated to VRAM leakage across dataloader threads." }, { "left": "issue:43519", "right": "issue:44485", "accept": false, - "reason": "Qwen3VL timestamp calculation and GLM-5 RoPE implementation are different model-specific bugs." + "reason": "Timestamp calculation in Qwen3VL does not match a GLM-5 RoPE implementation bug." }, { "left": "issue:43519", "right": "issue:44908", "accept": false, - "reason": "Timestamp calculation bug is unrelated to scheduler kwargs handling." + "reason": "Processor timestamp math and scheduler kwargs are unrelated." }, { "left": "issue:31795", "right": "issue:34689", "accept": false, - "reason": "Forward-doc confusion and Llama model-loading regression are not the same bug." + "reason": "Documentation ambiguity around forward args is not the same as model loading regression for Llama 3.2 Vision." }, { "left": "issue:31795", "right": "issue:37428", "accept": false, - "reason": "Documentation confusion is unrelated to a flash-attention import error." + "reason": "Forward-doc confusion is unrelated to the flash_attention_utils import failure." }, { "left": "issue:44315", "right": "issue:45092", "accept": false, - "reason": "Liger Kernel application and InternVL2 meta-init incompatibility are different failure modes." + "reason": "Liger Kernel application in model_init and remote-code meta-init incompatibility are different loading paths." }, { "left": "issue:43704", "right": "issue:44485", "accept": false, - "reason": "VRAM leak in multi-threaded use and GLM-5 RoPE implementation are unrelated." + "reason": "VRAM leakage in threaded inference is unrelated to RoPE implementation concerns." }, { "left": "issue:44485", "right": "issue:45468", "accept": false, - "reason": "Both are model-specific, but the model and bug behavior do not match." + "reason": "Different model families and different bugs: GLM-5 RoPE vs. Gemma4Audio positional encoding." }, { "left": "issue:42371", "right": "issue:43704", "accept": false, - "reason": "TF32 behavior settings and VRAM leak in threads are unrelated." + "reason": "TF32 API settings and VRAM leaks are unrelated." }, { "left": "issue:43502", "right": "issue:43519", "accept": false, - "reason": "local_files_only API requests and timestamp calculation are different bugs." + "reason": "local_files_only network leakage and Qwen3VL timestamp calculation are unrelated." }, { "left": "issue:43519", "right": "issue:44623", "accept": false, - "reason": "Processor timestamp math and missing processor.save_pretrained files are unrelated." + "reason": "Processor timestamp calculation is not the same as processor.save_pretrained missing files." }, { "left": "issue:43519", "right": "issue:43704", "accept": false, - "reason": "Timestamp calculation bug does not match the VRAM leak report." + "reason": "Timestamp calculation bug and VRAM leak are different failure modes." }, { "left": "issue:44263", "right": "issue:45468", "accept": false, - "reason": "torch.split return values in GlmMoeDsaIndexer are unrelated to Gemma-4 audio encoding." + "reason": "torch.split return-value issue is unrelated to Gemma4AudioRelPositionalEncoding." }, { "left": "issue:44261", "right": "issue:45468", "accept": false, - "reason": "MLA q_a_layernorm epsilon issue is not the same as Gemma-4 audio positional encoding." + "reason": "MLA q_a_layernorm precision/config issue is unrelated to Gemma4Audio positional encoding." }, { "left": "issue:43606", "right": "issue:45468", "accept": false, - "reason": "CPU offload device mismatch is unrelated to Gemma-4 audio positional encoding." + "reason": "CPU offload device mismatch is not the same as Gemma4AudioRelPositionalEncoding." }, { - "left": "issue:43704", - "right": "issue:45341", + "left": "issue:42890", + "right": "issue:43704", "accept": false, - "reason": "VRAM leak and a testing_utils bug are different issues." + "reason": "Missing set_seed in tests is unrelated to VRAM leakage in dataloader threads." }, { "left": "issue:42913", "right": "issue:43066", "accept": false, - "reason": "Both concern tokenizer v5, but one is a broad behavior change and the other is a specific decoder-type mismatch." + "reason": "Tokenizer behavior differences and wrong decoder type are related only loosely, not the same underlying bug." }, { "left": "issue:44908", "right": "issue:44998", "accept": false, - "reason": "Scheduler kwargs bug is unrelated to the obviously invalid 'Unemployment' issue." + "reason": "Scheduler kwarg handling and a nonsense issue title are clearly unrelated." }, { "left": "issue:44861", "right": "issue:45290", "accept": false, - "reason": "Tied-weight key handling and chat-template tool-call crash are different code paths." + "reason": "Tied weight key handling is unrelated to apply_chat_template crashing on tool-call assistant messages." }, { - "left": "issue:44998", - "right": "issue:45341", + "left": "issue:44492", + "right": "issue:44998", "accept": false, - "reason": "Invalid/non-technical issue and a testing_utils bug are unrelated." + "reason": "Cache-strategy typo and a nonsense issue title are unrelated." }, { - "left": "issue:44492", - "right": "issue:44998", + "left": "issue:44998", + "right": "issue:45341", "accept": false, - "reason": "A typo in cache strategies is unrelated to the invalid issue entry." + "reason": "Nonsense title and a testing_utils bug are not the same issue." }, { "left": "issue:43618", "right": "issue:44998", "accept": false, - "reason": "CLIPOutput attentions regression and the invalid issue entry are unrelated." + "reason": "CLIPOutput attentions regression is unrelated to the nonsense issue title." }, { "left": "issue:44556", "right": "issue:44998", "accept": false, - "reason": "Checkpoint reload regression and the invalid issue entry are unrelated." - }, - { - "left": "issue:43525", - "right": "issue:44861", - "accept": false, - "reason": "Llama4Config pad_token_id missing and tied-weight key AttributeError are different bugs." + "reason": "Checkpoint reload regression across versions is unrelated to the nonsense issue title." } ] }, @@ -13751,6 +13701,7 @@ "issue:36683", "issue:38617", "issue:43298", + "issue:43525", "issue:43572", "issue:44230", "issue:44623", @@ -13758,6 +13709,7 @@ "issue:45310" ], "soft_pairs": [ + "issue:43525|issue:44861", "issue:43572|issue:44861", "issue:31515|issue:38617", "issue:44230|issue:44623", @@ -13768,88 +13720,94 @@ "issue:31515|issue:36683", "issue:43298|issue:44230" ], - "prepared_review_unit_hash": "581c6ccb77eb9c1044476fffdc6585157930bc4c98caacd860142d3809e81318", + "prepared_review_unit_hash": "eea735345ed49354c57614eab97c57d5a078991791107e80839aae5cccf288ba", "component_budget": { "node_count": 367, "item_count": 367, "soft_pair_count": 1370, - "serialized_chars": 706987, + "serialized_chars": 706986, "estimated_input_tokens": 176747, "estimated_eval_tokens": 353750 }, "budget": { - "node_count": 12, - "item_count": 12, - "soft_pair_count": 9, - "serialized_chars": 9455, - "estimated_input_tokens": 2364, - "estimated_eval_tokens": 4984 + "node_count": 13, + "item_count": 13, + "soft_pair_count": 10, + "serialized_chars": 10321, + "estimated_input_tokens": 2581, + "estimated_eval_tokens": 5418 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is heterogeneous: most items are distinct loading/configuration/inference issues, plus one documentation issue. The soft similarities are mostly superficial (shared terms like from_pretrained, loading, or model family) rather than the same underlying bug.", - "confidence": 0.94, - "canonical_issue_reason": "If a single issue must represent the cluster, #45310 is the closest fit because it is a concrete, current from_pretrained regression with a specific model family; however, the set is not truly duplicate enough for a strong canonical issue.", + "summary": "This set is not a true duplicate cluster: it mixes unrelated loading, config, docs, serialization, and feature-request issues across different code paths. No single issue cleanly represents the whole group.", + "confidence": 0.95, + "canonical_issue_reason": "No canonical issue fits the cluster: the reports describe different underlying bugs and requests rather than one shared regression.", "canonical_pr_reason": null, - "best_issue_reason": "#45310 is the most actionable and representative of the loading-path failures in this set, while the others are either unrelated, documentation-only, or concern different subsystems/models.", + "best_issue_reason": "Issue 45310 is the most concrete recent from_pretrained regression, but it is still only one narrow problem and does not represent the rest of the mixed cluster.", "best_pr_reason": null, "soft_edge_verdicts": [ + { + "left": "issue:43525", + "right": "issue:44861", + "accept": false, + "reason": "Different bugs: Llama4Config missing pad_token_id vs _get_tied_weight_keys crashing on list-valued _tied_weights_keys." + }, { "left": "issue:43572", "right": "issue:44861", "accept": false, - "reason": "Both are runtime errors, but one is a StableLmConfig padding-field regression and the other is an AttributeError in tied-weight-key handling; different code paths and fixes." + "reason": "StableLmConfig pad_token_idx regression is unrelated to tied-weight key handling." }, { "left": "issue:31515", "right": "issue:38617", "accept": false, - "reason": "Checkpoint loading slowness and an ImportError for layer_type_validation are unrelated bugs." + "reason": "One is slow checkpoint loading; the other is an ImportError for layer_type_validation. Different failure modes and code paths." }, { "left": "issue:44230", "right": "issue:44623", "accept": false, - "reason": "FP8 inference support for Qwen3-VL/Qwen3.5 MoE is not the same as processor.save_pretrained() omitting files; different functionality." + "reason": "FP8 Qwen3-VL/MoE inference support is a feature request; processor.save_pretrained missing files is a serialization bug." }, { "left": "issue:31515", "right": "issue:36296", "accept": false, - "reason": "from_pretrained performance and tensor-parallel training behavior are separate problems." + "reason": "from_pretrained checkpoint slowness is unrelated to tensor-parallel training behavior." }, { "left": "issue:31515", "right": "issue:31795", "accept": false, - "reason": "Loading speed and documentation confusion about input_ids/past_key_values are not the same underlying issue." + "reason": "Checkpoint loading performance and forward API documentation confusion are distinct issues." }, { "left": "issue:44230", "right": "issue:45310", "accept": false, - "reason": "Both mention Qwen3.5 MoE, but one is FP8 support/inference and the other is a specific from_pretrained error; not the same concrete bug." + "reason": "FP8 support request for Qwen3-VL/MoE is not the same as a Qwen3.5 MoE from_pretrained error." }, { "left": "issue:31515", "right": "issue:33453", "accept": false, - "reason": "Both concern loading, but one is slow checkpoint loading while the other is a tokenizer-loading regression." + "reason": "Slow from_pretrained checkpoint loading and tokenizer loading regression are different loader problems." }, { "left": "issue:31515", "right": "issue:36683", "accept": false, - "reason": "Checkpoint loading slowness and Gemma3Config missing vocab_size are unrelated." + "reason": "Checkpoint load slowness does not match the Gemma3Config missing vocab_size attribute bug." }, { "left": "issue:43298", "right": "issue:44230", "accept": false, - "reason": "Continuous batching audio-model support and FP8 Qwen3-VL/Qwen3.5 MoE inference support are different feature gaps." + "reason": "Continuous batching audio-model support and FP8 inference support are separate feature gaps, not one defect." } ] }, @@ -13879,12 +13837,12 @@ "soft_pairs": [ "issue:26350|issue:33260" ], - "prepared_review_unit_hash": "baebfafb0752bddacd5d146084f8a4f91935df26877ca18d273940cecc64234e", + "prepared_review_unit_hash": "ff47fbaf803419612d8855ebedada103273e93ac4d94c9442c704ff31af5379c", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1589, + "serialized_chars": 1590, "estimated_input_tokens": 398, "estimated_eval_tokens": 1052 }, @@ -13892,7 +13850,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1589, + "serialized_chars": 1590, "estimated_input_tokens": 398, "estimated_eval_tokens": 1052 }, @@ -13901,18 +13859,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are not duplicates: one requests Flash Attention 2 support across more architectures, the other requests GGUF support across more architectures. They concern different features, code paths, and implementation work.", - "confidence": 0.99, - "canonical_issue_reason": null, + "summary": "These are two separate feature-request issues about adding support for more architectures, but they target different technologies (Flash Attention 2 vs GGUF) and do not describe the same underlying change.", + "confidence": 0.96, + "canonical_issue_reason": "No canonical duplicate issue: the requests are about different capabilities and code paths, not the same bug or feature.", "canonical_pr_reason": null, - "best_issue_reason": "No single issue is a clear canonical representative because the two issues are unrelated feature requests rather than duplicate reports.", + "best_issue_reason": "Issue 26350 is the better representative issue only by discussion activity and age, but it is not a duplicate of 33260.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:26350", "right": "issue:33260", "accept": false, - "reason": "Different requested capabilities (Flash Attention 2 vs GGUF support); they do not describe the same underlying bug or change." + "reason": "Different feature requests: one is for Flash Attention 2 support, the other for GGUF support. They are related only at a broad 'more architectures' theme, not the same underlying change." } ] }, @@ -13949,12 +13907,12 @@ "issue:43834|issue:44365", "issue:44365|issue:44370" ], - "prepared_review_unit_hash": "eac5735c559ad8904cd71fff66e30d66dbf924494c1f2010923c687cdedf47ea", + "prepared_review_unit_hash": "adc762a3c4bfe0ed98700f7c411845279288a0eab9a9651160efbb067a233c1e", "component_budget": { "node_count": 4, "item_count": 4, "soft_pair_count": 6, - "serialized_chars": 4453, + "serialized_chars": 4455, "estimated_input_tokens": 1114, "estimated_eval_tokens": 2484 }, @@ -13962,7 +13920,7 @@ "node_count": 4, "item_count": 4, "soft_pair_count": 6, - "serialized_chars": 4453, + "serialized_chars": 4455, "estimated_input_tokens": 1114, "estimated_eval_tokens": 2484 }, @@ -13971,48 +13929,48 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are translation-request issues with very similar template text, but they are not clearly the same underlying task: one is specifically for Turkish, while the others are generic language placeholders. The generic issues look templated rather than true duplicates of the Turkish request.", - "confidence": 0.78, - "canonical_issue_reason": "issue:27088 is the only concrete, language-specific report (Turkish), so it is the clearest representative of this cluster.", + "summary": "These are templated translation-request issues, not a clear duplicate set. The Turkish-specific issue is concrete, while the others are generic placeholders or separate language targets, so they should not be merged as the same underlying request.", + "confidence": 0.91, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "issue:27088 is the most specific and informative issue; the others are generic template instances with no concrete language filled in.", + "best_issue_reason": "issue:27088 is the most concrete and specific item in the cluster, with an explicit Turkish translation request; the others are generic template variants with no distinct target language shown.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:27088", "right": "issue:44370", "accept": false, - "reason": "Similar translation template, but one is Turkish-specific and the other is a generic placeholder; not clearly the same underlying request." + "reason": "Different or unspecified language targets; similar template wording is not enough to treat them as the same issue." }, { "left": "issue:27088", "right": "issue:44365", "accept": false, - "reason": "High textual similarity, but the language-target details are not the same concrete issue." + "reason": "This is a generic placeholder-style translation request, not the same concrete Turkish request." }, { "left": "issue:27088", "right": "issue:43834", "accept": false, - "reason": "Looks like the same template category, not the same exact translation task." + "reason": "Template similarity only; no evidence these are the same underlying translation task." }, { "left": "issue:43834", "right": "issue:44370", "accept": false, - "reason": "Near-identical template text, but no evidence they refer to the same concrete language or request." + "reason": "Both look like generic template instances, but they do not identify the same concrete language target or work item." }, { "left": "issue:43834", "right": "issue:44365", "accept": false, - "reason": "Template-level similarity only; insufficient evidence of a true duplicate." + "reason": "High textual similarity alone is insufficient; no specific shared translation target is established." }, { "left": "issue:44365", "right": "issue:44370", "accept": false, - "reason": "Same placeholder-style issue template, but not enough evidence that they are the same underlying issue." + "reason": "These appear to be separate template-style requests rather than one identical issue." } ] }, @@ -14077,178 +14035,184 @@ "issue:43837|issue:45027", "issue:43837|issue:45177" ], - "prepared_review_unit_hash": "7249c26c99bd9d35c50d97cf822be25ff964af3ebe920aec3e1f24ab47e1cea3", + "prepared_review_unit_hash": "eb25d1614f619d939c8ee4daa842bc13ddbb473e669bdab817440b9c484fc3bd", "component_budget": { "node_count": 13, "item_count": 13, "soft_pair_count": 25, - "serialized_chars": 16009, - "estimated_input_tokens": 4003, - "estimated_eval_tokens": 8262 + "serialized_chars": 16008, + "estimated_input_tokens": 4002, + "estimated_eval_tokens": 8260 }, "budget": { "node_count": 13, "item_count": 13, "soft_pair_count": 25, - "serialized_chars": 16009, - "estimated_input_tokens": 4003, - "estimated_eval_tokens": 8262 + "serialized_chars": 16008, + "estimated_input_tokens": 4002, + "estimated_eval_tokens": 8260 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are mostly unrelated model-addition requests across different model families and tasks. They share only a broad \"add support\" theme, so there is no true duplicate cluster; 43671 is just the most connected anchor by similarity.", - "confidence": 0.94, - "canonical_issue_reason": "issue:43671 is the most central node in the similarity graph, but it is still a specific Qwen3-TTS request rather than a true umbrella duplicate.", + "summary": "This is a loose cluster of model-addition requests, not a true duplicate set. The items cover distinct models and often different modalities/tasks (BERT variants, ASR/TTS, OCR, sparse transformers, vision-language encoders), so the soft similarities are mostly generic wording rather than the same underlying change.", + "confidence": 0.95, + "canonical_issue_reason": "If forced to pick an anchor, issue:37015 is the broadest and most active generic model-add request, but it is not a true duplicate of the others.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43671 has the strongest connectivity among the issues and is the best available representative anchor for this loose cluster.", + "best_issue_reason": "issue:37015 has the strongest discussion activity and is the most representative generic request in this otherwise non-duplicate set.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:36817", "right": "issue:41211", "accept": false, - "reason": "EuroBert config support and DEIMv2 are different model requests with no shared code-path evidence." + "reason": "Different model targets (EuroBert config vs DEIMv2); shared theme is only generic model addition." }, { "left": "issue:36817", "right": "issue:43671", "accept": false, - "reason": "EuroBert config and Qwen3-TTS are unrelated model additions." + "reason": "EuroBert config request and Qwen3-TTS support are different model families and tasks." }, { "left": "issue:37015", "right": "issue:43837", "accept": false, - "reason": "NeoBERT and Qwen3-ASR are different concrete model support requests." + "reason": "NeoBERT and Qwen3-ASR are unrelated model additions; not the same underlying change." }, { "left": "issue:37015", "right": "issue:44121", "accept": false, - "reason": "NeoBERT and the weight-sparse transformer request target different model/code paths." + "reason": "NeoBERT vs circuit-sparsity model request; different architecture and code path." }, { "left": "issue:37015", "right": "issue:43909", "accept": false, - "reason": "NeoBERT and LFM2.5 Audio are different model families." + "reason": "NeoBERT and LFM2.5 Audio are different model families and modalities." }, { "left": "issue:37015", "right": "issue:42503", "accept": false, - "reason": "NeoBERT and ModernVBERT are separate model requests." + "reason": "NeoBERT and ModernVBERT are distinct model requests, not one duplicate." }, { "left": "issue:37015", "right": "issue:45027", "accept": false, - "reason": "NeoBERT and Voxtral-4B-TTS support are unrelated changes." + "reason": "NeoBERT and Voxtral TTS are separate model support requests with different implementation needs." }, { "left": "issue:41211", "right": "issue:43671", "accept": false, - "reason": "DEIMv2 and Qwen3-TTS are distinct requests for different models." + "reason": "DEIMv2 and Qwen3-TTS are unrelated requests; only broad model-add similarity." }, { "left": "issue:42503", "right": "issue:43837", "accept": false, - "reason": "ModernVBERT and Qwen3-ASR are different concrete additions." + "reason": "ModernVBERT models and Qwen3-ASR are different tasks and code paths." }, { "left": "issue:42503", "right": "issue:43671", "accept": false, - "reason": "ModernVBERT and Qwen3-TTS are not the same underlying change." + "reason": "ModernVBERT and Qwen3-TTS are not the same model or feature." }, { "left": "issue:42503", "right": "issue:42738", "accept": false, - "reason": "ModernVBERT request and the broad RoPE-themed BERT-like models issue are not a single concrete bug/change." + "reason": "ModernVBERT models vs generic BERT-like RoPE support; related theme, but not the same change." }, { "left": "issue:42503", "right": "issue:45027", "accept": false, - "reason": "ModernVBERT and Voxtral-4B-TTS support are unrelated model additions." + "reason": "ModernVBERT and Voxtral-TTS are different model support requests." }, { "left": "issue:42738", "right": "issue:45027", "accept": false, - "reason": "A generic BERT-like RoPE request and Voxtral TTS support do not describe the same change." + "reason": "BERT-like RoPE support and Voxtral-TTS are not duplicate changes." }, { "left": "issue:43671", "right": "issue:43837", "accept": false, - "reason": "Qwen3-TTS and Qwen3-ASR are related by family name only; they are different model capabilities and likely separate work." + "reason": "Qwen3-TTS and Qwen3-ASR are different modalities and implementation paths; same family name is insufficient." }, { "left": "issue:43671", - "right": "issue:43909", + "right": "issue:45027", "accept": false, - "reason": "Qwen3-TTS and LFM2.5 Audio are different models and support paths." + "reason": "Qwen3-TTS vs Voxtral-TTS are separate TTS model support requests." }, { "left": "issue:43671", - "right": "issue:44121", + "right": "issue:45177", "accept": false, - "reason": "Qwen3-TTS support and the weight-sparse transformer request are unrelated." + "reason": "Qwen3-TTS and DeepSeek-OCR2 are unrelated model requests." }, { "left": "issue:43671", - "right": "issue:45027", + "right": "issue:44121", "accept": false, - "reason": "Both mention TTS, but Qwen3-TTS and Voxtral-4B-TTS are different model integrations." + "reason": "TTS support and weight-sparse transformer support are different feature/code changes." }, { "left": "issue:43671", - "right": "issue:45177", - "accept": false, - "reason": "Qwen3-TTS and DeepSeek-OCR2 are unrelated model requests." - }, - { - "left": "issue:43837", "right": "issue:43909", "accept": false, - "reason": "Qwen3-ASR and LFM2.5 Audio are different concrete model requests." + "reason": "Qwen3-TTS and LFM2.5 Audio are both audio-related but clearly different models and integrations." }, { "left": "issue:43837", - "right": "issue:44121", + "right": "issue:43909", "accept": false, - "reason": "Qwen3-ASR and the weight-sparse transformer request are not the same underlying issue." + "reason": "Qwen3-ASR and LFM2.5 Audio are distinct audio model requests, not duplicates." }, { "left": "issue:43837", "right": "issue:45322", "accept": false, - "reason": "Qwen3-ASR and EUPE are different model additions." + "reason": "ASR support and EUPE encoder support are different model families and tasks." }, { "left": "issue:43837", - "right": "issue:45027", + "right": "issue:44121", "accept": false, - "reason": "Qwen3-ASR support and Voxtral TTS support are separate tasks." + "reason": "Qwen3-ASR and circuit-sparsity transformer support are unrelated changes." }, { "left": "issue:43909", "right": "issue:44205", "accept": false, - "reason": "LFM2.5 Audio and SAM3-LiteText target different models and functionality." + "reason": "Audio model request vs SAM3-LiteText; different model types and implementation work." }, { "left": "issue:44121", "right": "issue:45322", "accept": false, - "reason": "Weight-sparse transformer support and EUPE are unrelated requests." + "reason": "Weight-sparse transformer request and EUPE encoder request are different underlying changes." + }, + { + "left": "issue:43837", + "right": "issue:45027", + "accept": false, + "reason": "Qwen3-ASR and Voxtral-TTS are different speech models and code paths." + }, + { + "left": "issue:43837", + "right": "issue:45177", + "accept": false, + "reason": "Qwen3-ASR and DeepSeek-OCR2 are unrelated model additions." } ] }, @@ -14289,7 +14253,7 @@ "issue:43668|issue:44031", "issue:43668|issue:43813" ], - "prepared_review_unit_hash": "5ed7737059fb436b74e1606eed85113db8803d29db774ec691677aaf92ec4465", + "prepared_review_unit_hash": "7cce7d1db6d3ce2f9ee54f6d2ba6dfb053d8aa4790ff7aea72db216b46d22c12", "component_budget": { "node_count": 6, "item_count": 6, @@ -14311,60 +14275,60 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "No duplicate cluster here: the six issues cover unrelated model, config, loss, quantization, and tokenizer problems, so none should be merged as the same bug.", - "confidence": 0.95, - "canonical_issue_reason": "No canonical issue stands out because the items are heterogeneous and do not appear to describe the same underlying defect.", + "summary": "False-positive cluster: the issues span unrelated bugs across different models and subsystems, with only superficial keyword overlap.", + "confidence": 0.98, + "canonical_issue_reason": "No canonical issue stands out; the reports cover distinct problems (normalization behavior, Mask2Former loss handling, missing kwargs propagation, a type-hint typo, a quantization attribute typo, and a tokenizer regex warning).", "canonical_pr_reason": null, - "best_issue_reason": "No single issue is a good representative duplicate target; each item is about a different code path or even a different class of bug (bug vs typo vs type hint).", + "best_issue_reason": "No issue is a strong representative of the cluster. If forced, 44031 is the broadest user-facing regression report, but it still does not align with the others.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:37934", "right": "issue:43668", "accept": false, - "reason": "Different concerns: Llama4TextL2Norm behavior vs ModernBERTConfig type annotation. No shared bug or code path." + "reason": "Different problem domains: Llama4 normalization semantics vs a ModernBERT config type-hint typo." }, { "left": "issue:42936", "right": "issue:43240", "accept": false, - "reason": "Both are loss-related, but one is about an unused ignore_value and the other about kwargs not being forwarded to the loss function. Related area, not the same defect." + "reason": "Mask2Former ignore_value handling is separate from kwargs not being forwarded into the loss function." }, { "left": "issue:42936", "right": "issue:44031", "accept": false, - "reason": "Mask2former ignore_value handling and tokenizer regex warnings are unrelated subsystems with different symptoms." + "reason": "Unrelated subsystems: Mask2Former loss/ignore_value vs tokenizer regex warnings after a version bump." }, { "left": "issue:42936", "right": "issue:43813", "accept": false, - "reason": "Unused ignore_value in Mask2former is unrelated to a typo in a quantization_operation attribute name." + "reason": "No shared code path; one is a segmentation loss bug, the other is a typo in a quantization attribute name." }, { "left": "issue:42936", "right": "issue:43668", "accept": false, - "reason": "Mask2former loss parameter handling and ModernBERTConfig norm_eps typing are unrelated." + "reason": "Different artifact types and bugs: runtime loss behavior vs a config annotation mistake." }, { "left": "issue:43240", "right": "issue:43668", "accept": false, - "reason": "Loss kwargs forwarding bug vs config type-hint mistake; different code paths and fixes." + "reason": "Missing loss kwargs forwarding is not the same as an incorrect type hint in ModernBERTConfig." }, { "left": "issue:43668", "right": "issue:44031", "accept": false, - "reason": "Incorrect type hint in a config class is not the same issue as tokenizer regex warning regressions." + "reason": "A type-hint bug and a tokenizer regex warning regression are unrelated." }, { "left": "issue:43668", "right": "issue:43813", "accept": false, - "reason": "A type-hint correction and a string-attribute typo are distinct, unrelated issues." + "reason": "ModernBERT config typing issue vs a literal typo in a different conversion path." } ] }, @@ -14394,7 +14358,7 @@ "soft_pairs": [ "issue:41084|issue:43037" ], - "prepared_review_unit_hash": "1f465789384842b4448a871e1bb33d219f79a946fd265250b1e2c164baccca87", + "prepared_review_unit_hash": "7f26a9e50cf93f54877d8d9b20dbe86c1c27e39deaa15fed3c9c8d86e8ebc753", "component_budget": { "node_count": 2, "item_count": 2, @@ -14416,18 +14380,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The two issues do not appear to describe the same underlying problem: one is a generic decoding change request, the other is a specific DeepSeek v3.2 support request. I would not cluster them as duplicates.", - "confidence": 0.94, - "canonical_issue_reason": "No canonical issue for duplication purposes: the titles point to different asks and there is no shared target or evidence of the same bug/change.", + "summary": "These are not duplicates: one is a generic request for block decoding, the other is a feature request for DeepSeek v3.2 support. They do not describe the same underlying bug or change.", + "confidence": 0.92, + "canonical_issue_reason": "Issue #41084 is the broader, more general request around block decoding and is the better representative of that topic.", "canonical_pr_reason": null, - "best_issue_reason": "issue #43037 is the better representative only because it is more specific and actionable; #41084 is too terse and ambiguous to serve as a cluster anchor.", + "best_issue_reason": "Issue #41084 is the best issue candidate because it is the more foundational/general problem statement, while #43037 is model-specific.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:41084", "right": "issue:43037", "accept": false, - "reason": "The requests are unrelated: \"Set Block Decoding\" is a generic decoding concern, while \"DeepSeek v3.2 support\" is a model-specific support request. No evidence they are the same issue." + "reason": "The titles and likely scopes differ: block decoding versus DeepSeek v3.2 support. They are not the same bug or change and there is no explicit linkage or shared target." } ] }, @@ -14457,12 +14421,12 @@ "soft_pairs": [ "issue:43598|issue:43600" ], - "prepared_review_unit_hash": "938a106bd0fec204fda2faf9c5f473611537e0d9381060aa996a0826c890a683", + "prepared_review_unit_hash": "c24bd1fb854b312ac387e1c1df5f53f35cd2822c20b28cbdc1dbcd54acd2979e", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1584, + "serialized_chars": 1583, "estimated_input_tokens": 396, "estimated_eval_tokens": 1048 }, @@ -14470,7 +14434,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1584, + "serialized_chars": 1583, "estimated_input_tokens": 396, "estimated_eval_tokens": 1048 }, @@ -14479,18 +14443,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are two separate Trainer customization topics: one about `remove_unused_column` behavior, the other about FLOPS estimation for custom trainers. They are related at a broad subsystem level but not the same underlying issue.", - "confidence": 0.92, - "canonical_issue_reason": "Issue #43598 is the more concrete and broadly representative Trainer customization request, centered on a specific behavior (`remove_unused_column`) and likely to be easier to use as the cluster anchor.", + "summary": "These are two distinct Trainer-related feature requests: one about `remove_unused_column` customizability, the other about FLOPS estimation for custom trainers. They share only a broad subsystem and are not duplicates.", + "confidence": 0.97, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "Issue #43598 is the best standalone issue because it has a clearer, actionable scope and directly targets a specific Trainer customization pain point.", + "best_issue_reason": "No single issue cleanly represents both topics; they are unrelated enough that neither should be treated as the canonical issue for the pair.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43598", "right": "issue:43600", "accept": false, - "reason": "Both concern Trainer customization, but one is about unused-column handling and the other about FLOPS estimation. They affect different code paths and are not the same bug/change." + "reason": "Different underlying requests: one changes input-column pruning behavior, the other adjusts FLOPS estimation for custom trainers. Same area (Trainer) but not the same bug or fix." } ] }, @@ -14520,7 +14484,7 @@ "soft_pairs": [ "issue:43684|issue:44405" ], - "prepared_review_unit_hash": "4c8b175bd2183ed075472800fc75e6e280afd2821d013dd0dfd351439e55a483", + "prepared_review_unit_hash": "1dd9098894c31059fd5cb3a2fddeaef92e33859d1b73baa60e2353c63c5e07ef", "component_budget": { "node_count": 2, "item_count": 2, @@ -14542,18 +14506,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are two separate feature requests for different Qwen model families and different AutoModel mappings, so they should not be deduplicated.", - "confidence": 0.89, - "canonical_issue_reason": "Issue #43684 is the broader, more central request: add Qwen3-Omni registration to AutoModel and AutoModelForConditionalGeneration.", + "summary": "These are related only at a very high level (Qwen auto-class support), but they target different model families and different AutoModel APIs, so they are not duplicates.", + "confidence": 0.94, + "canonical_issue_reason": "Issue #43684 is broader and more general: it requests registration for Qwen3-Omni across core auto classes, making it the better representative if this cluster is grouped at all.", "canonical_pr_reason": null, - "best_issue_reason": "Issue #43684 is the best representative because it targets core model registration support, whereas #44405 is a narrower request for sequence classification on Qwen3.5.", + "best_issue_reason": "#43684 is the better canonical issue because it describes a broader registration gap, while #44405 is a narrower task for a specific sequence-classification head on Qwen3.5.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43684", "right": "issue:44405", "accept": false, - "reason": "Different model families and different tasks: Qwen3-Omni registration for base/conditional generation vs Qwen3.5 sequence classification support. They do not look like the same underlying bug or change." + "reason": "Different underlying changes: one asks for AutoModel/AutoModelForConditionalGeneration registration for Qwen3-Omni, the other asks for AutoModelForSequenceClassification support for Qwen3.5. They involve different model variants and different auto-class paths, so they should not be merged as the same bug/request." } ] }, @@ -14590,70 +14554,70 @@ "issue:43845|issue:43859", "issue:43846|issue:43859" ], - "prepared_review_unit_hash": "9aaaa889ef6b2278228976457b3a5ba8890b0c33277c46b124e5a6a04a01f431", + "prepared_review_unit_hash": "ca92f410b4da238dd16083ef795d5da4b6dcb52957138ef7dc6ad2bbffa8c25b", "component_budget": { "node_count": 4, "item_count": 4, "soft_pair_count": 6, - "serialized_chars": 4266, - "estimated_input_tokens": 1067, - "estimated_eval_tokens": 2390 + "serialized_chars": 4262, + "estimated_input_tokens": 1066, + "estimated_eval_tokens": 2388 }, "budget": { "node_count": 4, "item_count": 4, "soft_pair_count": 6, - "serialized_chars": 4266, - "estimated_input_tokens": 1067, - "estimated_eval_tokens": 2390 + "serialized_chars": 4262, + "estimated_input_tokens": 1066, + "estimated_eval_tokens": 2388 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "All four are sparse, closed issues with no actionable details; three share the same vague title, but there is not enough evidence to prove a single underlying bug beyond superficial similarity.", - "confidence": 0.28, - "canonical_issue_reason": "Issue 43845 is the best canonical placeholder because it matches the repeated title used by the majority of the cluster, but the available content is too thin to establish a real root problem.", + "summary": "Three issues share the same generic title \"huggingface \" and look like near-duplicate placeholders; one issue titled \"Fett\" is unrelated. I would keep the earliest huggingface report as the representative issue, but confidence is limited because the bodies contain no actionable detail.", + "confidence": 0.57, + "canonical_issue_reason": "issue:43845 is the earliest of the near-identical \"huggingface \" reports and is a reasonable representative for the cluster.", "canonical_pr_reason": null, - "best_issue_reason": "None of the issues is strongly distinguishable on substance; 43845 is the least arbitrary choice only because it is one of the repeated 'huggingface' entries.", + "best_issue_reason": "43845 is the best issue to retain because it is the earliest among the repeated huggingface-titled reports; however, the cluster is too underspecified to prove an exact duplicate set.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43835", "right": "issue:43845", "accept": false, - "reason": "Only weak lexical similarity; titles are different and there is no body/context tying them to the same bug." + "reason": "Different titles; \"Fett\" does not look like the same underlying report as the huggingface placeholder issues." }, { "left": "issue:43835", "right": "issue:43846", "accept": false, - "reason": "Only weak lexical similarity; titles are different and there is no body/context tying them to the same bug." + "reason": "Different titles and no shared details beyond generic similarity." }, { "left": "issue:43835", "right": "issue:43859", "accept": false, - "reason": "Only weak lexical similarity; titles are different and there is no body/context tying them to the same bug." + "reason": "Different titles; no evidence these describe the same problem." }, { "left": "issue:43845", "right": "issue:43846", "accept": false, - "reason": "Identical titles are not enough here; no description or issue target confirms the same underlying problem." + "reason": "The titles match, but the bodies are too sparse to confirm the same underlying bug with confidence." }, { "left": "issue:43845", "right": "issue:43859", "accept": false, - "reason": "Identical titles are not enough here; no description or issue target confirms the same underlying problem." + "reason": "Same generic title, but the available text is too minimal to safely treat them as the same issue." }, { "left": "issue:43846", "right": "issue:43859", "accept": false, - "reason": "Identical titles are not enough here; no description or issue target confirms the same underlying problem." + "reason": "Same generic title, but no concrete evidence shows they are the same underlying report." } ] }, @@ -14694,88 +14658,88 @@ "issue:44572|issue:44937", "issue:44573|issue:44937" ], - "prepared_review_unit_hash": "43da57b120ef45442d6f37610d2be07998042d9071100e7e4d26bf76c496f906", + "prepared_review_unit_hash": "19f232c873586bd704c8948bd44c826959351073a84ace2172e5815a0e44ba47", "component_budget": { "node_count": 5, "item_count": 5, "soft_pair_count": 9, - "serialized_chars": 5794, - "estimated_input_tokens": 1449, - "estimated_eval_tokens": 3154 + "serialized_chars": 5791, + "estimated_input_tokens": 1448, + "estimated_eval_tokens": 3152 }, "budget": { "node_count": 5, "item_count": 5, "soft_pair_count": 9, - "serialized_chars": 5794, - "estimated_input_tokens": 1449, - "estimated_eval_tokens": 3154 + "serialized_chars": 5791, + "estimated_input_tokens": 1448, + "estimated_eval_tokens": 3152 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "All items look like unrelated low-information/spam issues; none share a concrete bug or change. The only somewhat substantive item is the documentation suggestion, but it still does not match the others.", - "confidence": 0.86, - "canonical_issue_reason": "Issue 44418 is the most legible and substantive of the set; the others are clearly spam-like or similarly content-free.", + "summary": "These are mostly spammy or placeholder issues with no evidence they describe the same underlying bug or request. None of the soft similarity links are strong enough to treat as duplicates.", + "confidence": 0.93, + "canonical_issue_reason": "issue:44418 is the only item that resembles a real user-facing request (a documentation enhancement) rather than obvious spam, so it is the least-bad issue to anchor the cluster if one must be chosen.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 44418 is the best representative only because it has an actual topic and reads like a real report, unlike the spam-titled issues.", + "best_issue_reason": "issue:44418 is the best standalone representative because it is the only non-spam-looking issue; the others are generic spam or unrelated text and should not be merged with it.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44371", "right": "issue:44572", "accept": false, - "reason": "Both are spam-like placeholders with no evidence of the same underlying issue." + "reason": "Both titles are generic spam placeholders; no concrete shared bug or change is described." }, { "left": "issue:44371", "right": "issue:44573", "accept": false, - "reason": "No shared bug, code path, or concrete topic; both appear to be spam." + "reason": "Generic spam titles with no actionable content; not the same underlying issue." }, { "left": "issue:44371", "right": "issue:44418", "accept": false, - "reason": "One is spam-like, the other is a documentation suggestion; not the same issue." + "reason": "One is spam, the other is a documentation enhancement suggestion; unrelated." }, { "left": "issue:44371", "right": "issue:44937", "accept": false, - "reason": "Spam placeholder vs. unrelated named-topic issue; no duplicate signal." + "reason": "Spam placeholder versus a specific quoted phrase; no duplicate signal." }, { "left": "issue:44418", "right": "issue:44572", "accept": false, - "reason": "Documentation enhancement and spam placeholder are not the same underlying report." + "reason": "Documentation enhancement request and spam placeholder are not the same issue." }, { "left": "issue:44418", "right": "issue:44573", "accept": false, - "reason": "Documentation request does not match a spam-like placeholder issue." + "reason": "Documentation enhancement request versus spam placeholder; no shared bug/change." }, { "left": "issue:44572", "right": "issue:44573", "accept": false, - "reason": "Both are spam-like, but there is no evidence they describe the same concrete problem." + "reason": "Both are spam placeholders, but there is no evidence they refer to the same concrete problem." }, { "left": "issue:44572", "right": "issue:44937", "accept": false, - "reason": "No meaningful overlap beyond low-quality titles; not a duplicate." + "reason": "Spam placeholder versus a different quoted-text title; unrelated." }, { "left": "issue:44573", "right": "issue:44937", "accept": false, - "reason": "These do not describe the same bug or change." + "reason": "Spam placeholder versus a different quoted-text title; not a duplicate." } ] }, @@ -14805,7 +14769,7 @@ "soft_pairs": [ "issue:45145|issue:45151" ], - "prepared_review_unit_hash": "837be90f695d49915c46ee966086daabc409b17548a3cef838de9eb9fdbce0a8", + "prepared_review_unit_hash": "415c4aa3e00e24ce1931372625b1dcfeb68b94dc4d819f3e1ae3de82e9df4ae9", "component_budget": { "node_count": 2, "item_count": 2, @@ -14827,28 +14791,28 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The two closed issues are very similar and likely duplicate reports, but the evidence is still soft: same title, similar body size, and one inbound reference on #45145, with no explicit links or targets. #45145 remains the better canonical issue candidate, but the soft edge should be treated conservatively and not accepted outright.", - "confidence": 0.82, - "canonical_issue_reason": "#45145 is the better canonical issue because it has the same title as #45151 and an inbound reference, making it the slightly more established duplicate target.", + "summary": "The two closed issues have the same title and very similar body size/content, with no distinguishing links or filenames. They look like duplicate reports of the same Energy/N6 Arithmetic idea.", + "confidence": 0.98, + "canonical_issue_reason": "Issue #45145 is the better canonical because it has the same title as #45151 and slightly more inbound reference activity, suggesting it is the more connected duplicate target.", "canonical_pr_reason": null, - "best_issue_reason": "#45145 is the best overall issue candidate because it is the more connected of the two and appears to represent the same underlying report.", + "best_issue_reason": "#45145 is the best representative of this issue cluster since it appears equivalent to #45151 but has a bit more discussion/linkage.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45145", "right": "issue:45151", - "accept": false, - "reason": "Strong similarity is present, but there are no explicit links, targets, or code references, and the soft-similarity signal is not deterministic. That is enough to rank them as likely duplicates, but not enough to confidently accept the edge as the same underlying issue." + "accept": true, + "reason": "Same title, same apparent topic, and no evidence of a different underlying problem; these are duplicates." } ] }, "evaluator_result": { - "accept": false, - "feedback": "The packet supports strong similarity, but the analyst overstates the certainty. We only have same titles, similar body lengths, no explicit links/targets, and a soft-similarity signal with deterministic_accept=false; that is not enough to say the issues are \"near-identical duplicates\" or that they have the \"same title/content.\" The inbound reference does make #45145 a reasonable canonical candidate, but the soft-edge verdict should stay conservative rather than accepted outright." + "accept": true, + "feedback": "The conclusions are grounded in the packet: the issues have identical titles, similar body lengths, and only soft similarity evidence, with no conflicting links or filenames. The canonical choice is also conservative and based on the slightly higher inbound reference count for #45145. The soft-edge accept is reasonable here." }, "evaluator_used": true, - "retried": true, - "accepted_nontrivial_soft_edge": false, + "retried": false, + "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null }, @@ -14917,9 +14881,10 @@ "source": "llm", "cache_hit": false, "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-39895-10", + "cluster_id": "cluster-39895-12", "nodes": [ "pull_request:39895", + "pull_request:41855", "pull_request:42668", "pull_request:43251", "pull_request:43254", @@ -14928,10 +14893,13 @@ "pull_request:43995", "pull_request:44827", "pull_request:45170", - "pull_request:45213" + "pull_request:45213", + "pull_request:45587" ], "soft_pairs": [ "pull_request:39895|pull_request:43424", + "pull_request:41855|pull_request:43823", + "pull_request:41855|pull_request:43424", "pull_request:42668|pull_request:43823", "pull_request:43251|pull_request:43823", "pull_request:43251|pull_request:44827", @@ -14939,134 +14907,153 @@ "pull_request:43424|pull_request:43823", "pull_request:43823|pull_request:43995", "pull_request:43823|pull_request:44827", - "pull_request:43823|pull_request:45213", "pull_request:43995|pull_request:45213", "pull_request:43995|pull_request:44827", + "pull_request:44827|pull_request:45587", "pull_request:44827|pull_request:45170", "pull_request:44827|pull_request:45213", + "pull_request:43823|pull_request:45213", "pull_request:43251|pull_request:43254" ], - "prepared_review_unit_hash": "ceb48e1e4dc2622130fc591837e0399b00709a73f3a106d462142e3d6f177771", + "prepared_review_unit_hash": "9bd24c94d579fabe4f58899fc0b189a5c31340ff58adfb148d48f1825e471544", "component_budget": { - "node_count": 10, - "item_count": 10, - "soft_pair_count": 14, - "serialized_chars": 27589, - "estimated_input_tokens": 6898, - "estimated_eval_tokens": 14052 + "node_count": 12, + "item_count": 12, + "soft_pair_count": 17, + "serialized_chars": 32981, + "estimated_input_tokens": 8246, + "estimated_eval_tokens": 16748 }, "budget": { - "node_count": 10, - "item_count": 10, - "soft_pair_count": 14, - "serialized_chars": 27589, - "estimated_input_tokens": 6898, - "estimated_eval_tokens": 14052 + "node_count": 12, + "item_count": 12, + "soft_pair_count": 17, + "serialized_chars": 32981, + "estimated_input_tokens": 8246, + "estimated_eval_tokens": 16748 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "This cluster is mostly a set of unrelated pull requests. The only true duplicate-looking pair is the two fixed_cross_entropy patches (43251 and 43254), which both change the same function in the same file to forward extra cross-entropy kwargs. All other soft-similarity pairs touch different model additions, refactors, or tests and should stay separate.", + "summary": "Almost all pairs are unrelated across distinct model additions, refactors, tests, docs, or drafts. The only true duplicate is the loss-utils cross-entropy kwargs fix: 43251 and 43254 both target issue 43240 and patch the same function in the same code path.", "confidence": 0.95, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 43251 is the best canonical PR because it is the more complete cross-entropy kwargs fix, explicitly targets issue #43240, and makes the concrete code-path change in loss_utils.py; 43254 is just a smaller variant of the same patch.", + "canonical_pr_reason": "PR 43251 is a good canonical representative for the only true duplicate pair because it and 43254 both target issue 43240 and modify src/transformers/loss/loss_utils.py in the same cross-entropy forwarding path; the choice is representative, not because it is materially more complete.", "best_issue_reason": null, - "best_pr_reason": "PR 43251 is the strongest representative of the only real duplicate cluster here: it has the explicit issue link, the main implementation change, and broadest completeness for the fixed_cross_entropy kwargs support.", + "best_pr_reason": "PR 43251 is the best overall PR representative for deduping because it belongs to the only confirmed duplicate pair and is explicitly linked to issue 43240, making it a clean anchor for the shared loss-utils fix.", "soft_edge_verdicts": [ { "left": "pull_request:39895", "right": "pull_request:43424", "accept": false, - "reason": "Videoprism model addition vs executorch dynamic-shape test; different code paths and unrelated changes." + "reason": "VideoPrism model addition vs executorch dynamic-shape export test; unrelated changes." + }, + { + "left": "pull_request:41855", + "right": "pull_request:43823", + "accept": false, + "reason": "Tokenizer method addition vs new MobileLLM model implementation; different features and code paths." + }, + { + "left": "pull_request:41855", + "right": "pull_request:43424", + "accept": false, + "reason": "Tokenizer API work vs executorch export testing; not the same underlying bug or change." }, { "left": "pull_request:42668", "right": "pull_request:43823", "accept": false, - "reason": "Processor robustness refactor vs adding a new MobileLLM model; not the same underlying change." + "reason": "Processor auto-construction robustness vs new model integration; unrelated." }, { "left": "pull_request:43251", "right": "pull_request:43823", "accept": false, - "reason": "Cross-entropy kwargs fix vs new model addition; no shared bug or implementation target." + "reason": "Cross-entropy kwargs fix vs MobileLLM model addition; no shared code-path problem." }, { "left": "pull_request:43251", "right": "pull_request:44827", "accept": false, - "reason": "Loss utility patch vs Mistral4 test/model cleanup; unrelated functionality." + "reason": "Loss utility change vs Mistral4 test fixes; different concrete issues." }, { "left": "pull_request:43424", "right": "pull_request:44827", "accept": false, - "reason": "Executorch dynamic-shape test vs Mistral4 fix; different subsystems and goals." + "reason": "Executorch export test vs Mistral4 test fixes; both are tests but not the same fix." }, { "left": "pull_request:43424", "right": "pull_request:43823", "accept": false, - "reason": "Executorch export test vs MobileLLM model addition; not the same concrete change." + "reason": "Executorch exportability test vs MobileLLM model implementation; unrelated." }, { "left": "pull_request:43823", "right": "pull_request:43995", "accept": false, - "reason": "New MobileLLM model vs Falcon output-interface refactor; distinct work." + "reason": "MobileLLM model addition vs Falcon output-interface refactor; different model families and change types." }, { "left": "pull_request:43823", "right": "pull_request:44827", "accept": false, - "reason": "MobileLLM addition vs Mistral4 test/model fix; no shared underlying bug." + "reason": "New MobileLLM model vs Mistral4 test fixes; no common bug/change." }, { - "left": "pull_request:43823", + "left": "pull_request:43995", "right": "pull_request:45213", "accept": false, - "reason": "MobileLLM model addition vs model-creation skill/docs meta-change; unrelated." + "reason": "Falcon refactor vs draft model-creation skill content; not duplicates." }, { "left": "pull_request:43995", - "right": "pull_request:45213", + "right": "pull_request:44827", "accept": false, - "reason": "Falcon refactor vs model-creation skill/meta PR; not the same change." + "reason": "Falcon output-collection refactor vs Mistral4 test fixes; unrelated." }, { - "left": "pull_request:43995", - "right": "pull_request:44827", + "left": "pull_request:44827", + "right": "pull_request:45587", "accept": false, - "reason": "Falcon standardized-output refactor vs Mistral4 fix; different code paths." + "reason": "Mistral4 test fixes vs continuous batching docs update; different subsystems and purposes." }, { "left": "pull_request:44827", "right": "pull_request:45170", "accept": false, - "reason": "Mistral4 test/fix PR vs layernorm typo rename across CLIP-like models; separate issues." + "reason": "Mistral4 test fixes vs CLIP-like weight-renaming cleanup; unrelated." }, { "left": "pull_request:44827", "right": "pull_request:45213", "accept": false, - "reason": "Mistral4 fix PR vs model-creation skill; unrelated." + "reason": "Mistral4 test fixes vs draft model-creation skill; shared paths are incidental, not the same change." + }, + { + "left": "pull_request:43823", + "right": "pull_request:45213", + "accept": false, + "reason": "MobileLLM model addition vs model-creation skill draft; no shared underlying bug or feature." }, { "left": "pull_request:43251", "right": "pull_request:43254", "accept": true, - "reason": "Both patch src/transformers/loss/loss_utils.py to add the same fixed_cross_entropy kwargs support for the same underlying issue; they are plausibly mergeable into one PR." + "reason": "Both PRs patch src/transformers/loss/loss_utils.py to forward supported kwargs to nn.functional.cross_entropy for issue 43240; they are near-identical variants of the same fix." } ] }, "evaluator_result": { - "accept": true, - "feedback": "Grounded overall. The only accepted pair, 43251/43254, is a conservative mergeable duplicate: same file, same function, same issue target, and the diffs are near-identical kwargs-forwarding fixes. The other soft pairs are reasonably rejected as unrelated or at least not clearly the same bugfix based on the packet." + "accept": false, + "feedback": "The duplicate-pair acceptance is grounded, but the canonical PR reasoning overstates the evidence: both 43251 and 43254 explicitly target issue 43240 and patch the same function, so the packet does not support claiming 43251 is 'more complete' than 43254. The rest of the soft-edge rejects look conservative." }, "evaluator_used": true, - "retried": false, + "retried": true, "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null @@ -15090,12 +15077,12 @@ "soft_pairs": [ "pull_request:40546|pull_request:42781" ], - "prepared_review_unit_hash": "f19069bb41e72ee23a7689547797e4c59c570fbd67bbdf868544319caea37d9b", + "prepared_review_unit_hash": "5eea437ee5dac868dd94391308c8598c023d2279d6c26568c4557a0784104e08", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 7225, + "serialized_chars": 7226, "estimated_input_tokens": 1807, "estimated_eval_tokens": 3870 }, @@ -15103,7 +15090,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 7225, + "serialized_chars": 7226, "estimated_input_tokens": 1807, "estimated_eval_tokens": 3870 }, @@ -15112,18 +15099,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both PRs target VibeVoice, but they are not duplicates: #40546 is the base VibeVoice integration, while #42781 is a separate VibeVoice Realtime extension with extra tokenizer/model and generation changes. The overlap is mostly in shared model plumbing and docs, not the same concrete fix or one mergeable change set.", + "summary": "These are related VibeVoice PRs, but they are not the same change: #40546 adds the base VibeVoice implementation, while #42781 adds a separate VibeVoice Realtime extension with additional acoustic-tokenizer and generation changes. They should not be treated as duplicates.", "confidence": 0.89, "canonical_issue_reason": null, - "canonical_pr_reason": "#40546 is the more canonical PR for the cluster because it establishes the initial VibeVoice support and broader integration surface; #42781 is a later, narrower realtime follow-on rather than the primary implementation.", + "canonical_pr_reason": "#40546 is the foundational, more complete VibeVoice implementation and has substantial review activity; it is the better representative of the main model addition.", "best_issue_reason": null, - "best_pr_reason": "#40546 is the best representative PR for this cluster because it is the foundational VibeVoice implementation and has the broader, more complete scope. #42781 is better treated as a separate enhancement.", + "best_pr_reason": "#40546 is the best cluster representative because it covers the core VibeVoice codepath and docs, whereas #42781 is a narrower follow-up for realtime support.", "soft_edge_verdicts": [ { "left": "pull_request:40546", "right": "pull_request:42781", "accept": false, - "reason": "Reject: both are about VibeVoice, but #42781 adds a realtime variant with additional component changes; it is not the same concrete code-path problem as the base implementation in #40546 and would not naturally merge into one PR." + "reason": "Same model family, but different concrete scope: base VibeVoice implementation vs. realtime variant with extra tokenizer/generation work. Not the same mergeable change." } ] }, @@ -15153,12 +15140,12 @@ "soft_pairs": [ "pull_request:40755|pull_request:41224" ], - "prepared_review_unit_hash": "4b16aef37077ebfb3a69b5d139dc7c84f7b44c6684b37ece61239dae2f589590", + "prepared_review_unit_hash": "89411a8e362a683daa60f23c7aafa79d12a1e1ef588cc0fcbb617fe6f7cf58df", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4674, + "serialized_chars": 4675, "estimated_input_tokens": 1169, "estimated_eval_tokens": 2594 }, @@ -15166,7 +15153,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4674, + "serialized_chars": 4675, "estimated_input_tokens": 1169, "estimated_eval_tokens": 2594 }, @@ -15175,18 +15162,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are two unrelated pull requests that both add new model functionality, not the same underlying change: one adds TimesFM covariate forecasting, the other adds DINOv3 image classification support.", - "confidence": 0.98, + "summary": "The cluster contains two unrelated pull requests: one adds covariate forecasting for TimesFM, the other adds DINOv3 image classification support. They touch different models, code paths, and documentation, 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": "PR #41224 is the stronger representative because it is a focused, concrete feature addition with aligned code, tests, and docs for a single model path.", + "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "pull_request:40755", "right": "pull_request:41224", "accept": false, - "reason": "Different models and different capabilities: TimesFM covariate forecasting vs. DINOv3 image classification. They do not fix the same code-path problem and would not be merged as one PR." + "reason": "Different model families and different concrete features: TimesFM covariate regression vs DINOv3 image classification support. No shared bug/change target, and they would not plausibly be one merged PR." } ] }, @@ -15205,8 +15192,8 @@ "review_unit_count": 1, "status": "reviewed", "reason": null, - "source": "llm", - "cache_hit": false, + "source": "cache", + "cache_hit": true, "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-41356-2", "nodes": [ @@ -15279,12 +15266,12 @@ "soft_pairs": [ "pull_request:41797|pull_request:45186" ], - "prepared_review_unit_hash": "2f8b076ed799a1ac2907658e6dacd7d505295a8005be415cc38fd780dbd137d5", + "prepared_review_unit_hash": "ade0862fa5b40ee50a77ef3df4cf3e247c883fdfb6bee5f5a9b8bac09c191aae", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 6722, + "serialized_chars": 6723, "estimated_input_tokens": 1681, "estimated_eval_tokens": 3618 }, @@ -15292,7 +15279,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 6722, + "serialized_chars": 6723, "estimated_input_tokens": 1681, "estimated_eval_tokens": 3618 }, @@ -15301,18 +15288,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are two unrelated model-addition PRs that happen to share Hugging Face model-registration scaffolding. They do not fix the same bug or implement the same change, so they should not be clustered as duplicates.", - "confidence": 0.98, + "summary": "Both pull requests add entirely different model integrations (DeepSeek OCR vs. Isaac). They share only the usual Transformers boilerplate and registry updates, so they are not the same underlying change and should not be clustered as duplicates.", + "confidence": 0.94, "canonical_issue_reason": null, - "canonical_pr_reason": "No single PR is a canonical duplicate target here; if one must be chosen as the representative, PR #41797 is slightly more established and reviewed, but it covers DeepSeek OCR, not Isaac.", + "canonical_pr_reason": "No single PR is a true canonical for this pair because they target different models and separate codepaths; if forced, the DeepSeek OCR PR is the earlier, more established integration but it is not a duplicate of Isaac.", "best_issue_reason": null, - "best_pr_reason": "PR #41797 is the better representative only because it has more review activity and a fuller implementation/test set, but it is not a duplicate of PR #45186 and should not be merged with it.", + "best_pr_reason": "PR 41797 is the better representative only by convention: it is the earlier, more complete-looking model addition with broader review activity. However, it is not equivalent to 45186 and should not be treated as a duplicate of it.", "soft_edge_verdicts": [ { "left": "pull_request:41797", "right": "pull_request:45186", "accept": false, - "reason": "Different model additions for different concrete code paths (DeepSeek OCR vs Isaac). Shared auto-registration and docs plumbing is expected for independent model PRs, not evidence of the same underlying change." + "reason": "Different model additions for different architectures (DeepSeek OCR vs. Isaac). Shared files are only autogenerated registries/docs that appear in many model-integration PRs, not evidence of the same bug or change." } ] }, @@ -15342,12 +15329,12 @@ "soft_pairs": [ "pull_request:42166|pull_request:45426" ], - "prepared_review_unit_hash": "5e8eb21b6b05542698babb6054ba5fd64481e08559f788e46eda4ef76c82c709", + "prepared_review_unit_hash": "0e3ce74c15c13905f748f482878fa9c11bf16b4a3107f6b7950a1f80fc152023", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 5407, + "serialized_chars": 5408, "estimated_input_tokens": 1352, "estimated_eval_tokens": 2960 }, @@ -15355,7 +15342,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 5407, + "serialized_chars": 5408, "estimated_input_tokens": 1352, "estimated_eval_tokens": 2960 }, @@ -15364,18 +15351,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These two PRs are both model-addition scaffolds that touch shared registry files, but they implement different models entirely: InternVLFlash vs AXK1. The overlap is boilerplate, not the same underlying change.", - "confidence": 0.98, + "summary": "Both items are feature PRs adding different models (InternVLFlash vs AXK1). They overlap only in shared registration/auto-discovery files, not in the underlying code change, so they are not duplicates.", + "confidence": 0.97, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 42166 is the more established representative of the cluster, with much higher review activity and a fuller docs/test-oriented model addition, but it is not a duplicate of PR 45426.", + "canonical_pr_reason": "No single canonical PR fits this cluster: the PRs introduce different models with different implementations and test/docs surfaces.", "best_issue_reason": null, - "best_pr_reason": "If a representative PR is needed, 42166 is the better anchor because it has more discussion and a broader, more complete implementation surface; however, it still addresses a different model than 45426.", + "best_pr_reason": "No global best PR is appropriate here because the PRs are unrelated feature additions; the overlap is limited to shared model registry plumbing.", "soft_edge_verdicts": [ { "left": "pull_request:42166", "right": "pull_request:45426", "accept": false, - "reason": "Reject: these are separate feature PRs for different model families. The shared files are only auto-registry boilerplate; the concrete additions and code paths (InternVLFlash vs AXK1) are unrelated and not plausibly mergeable into one PR." + "reason": "Reject: these are different model-addition PRs for different code paths and cannot plausibly be merged into one duplicate fix/change." } ] }, @@ -15412,7 +15399,7 @@ "pull_request:43291|pull_request:44815", "pull_request:43665|pull_request:44815" ], - "prepared_review_unit_hash": "952f8d0cba5211a97978d8bfd693fdaa72a18591214204ed04115b2f9dfdef1d", + "prepared_review_unit_hash": "98c092ccb827bdf8fac102a0302171bd8565b7b3c8a930b4b1b511b61e369372", "component_budget": { "node_count": 4, "item_count": 4, @@ -15434,48 +15421,48 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These four PRs are not duplicates of one another. They span unrelated areas: Whisper pipeline language return, Whisper tokenizer/tests, vision model test files, and FP8/dequant model loading changes.", - "confidence": 0.94, + "summary": "These are four unrelated PRs that only share loose topical similarity (mostly Transformers/Whisper/test changes). They do not look like duplicates of the same underlying bug or change.", + "confidence": 0.89, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #42227 is the strongest representative of the cluster because it is a concrete, narrowly scoped bug fix with an explicit linked issue and a clear code-path change in Whisper ASR handling.", + "canonical_pr_reason": "PR 42227 is the cleanest canonical candidate: it has an explicit issue target, a focused code-path fix in the ASR pipeline, and a clear bug description.", "best_issue_reason": null, - "best_pr_reason": "PR #42227 is the best standalone representative: it addresses a specific runtime bug, has an explicit issue target, and its diff is directly about the underlying behavior rather than test-only or broad refactoring changes.", + "best_pr_reason": "PR 42227 is the most representative and mergeable-looking change in the set; the others are either unrelated test edits or a different FP8/dequantization effort.", "soft_edge_verdicts": [ { "left": "pull_request:42227", "right": "pull_request:43291", "accept": false, - "reason": "Both are Whisper-related, but they fix different things: #42227 propagates return_language through the ASR pipeline, while #43291 changes tokenizer decoding behavior and tests. They are not the same underlying bug or mergeable into one fix." + "reason": "Both mention Whisper, but one fixes ASR return_language plumbing while the other changes Whisper tokenizer/test behavior. Different code paths and different bugs." }, { "left": "pull_request:42227", "right": "pull_request:43665", "accept": false, - "reason": "Unrelated domains. #42227 fixes Whisper ASR language return; #43665 only edits CLIP/ViT tests. Same broad repo, but not the same code-path problem." + "reason": "42227 is a Whisper pipeline fix; 43665 is unrelated model test edits for CLIP/ViT. No shared underlying issue." }, { "left": "pull_request:42227", "right": "pull_request:44815", "accept": false, - "reason": "Completely different functionality: Whisper ASR language return versus FP8/dequant loading logic and Mistral-related integration changes. Not a duplicate." + "reason": "Whisper return-language plumbing is unrelated to FP8/dequantization/model-loading work. Different subsystem and bug class." }, { "left": "pull_request:43291", "right": "pull_request:43665", "accept": false, - "reason": "These are test-oriented changes in different model areas (Whisper tokenizer/tests vs CLIP/ViT tests) and do not indicate the same bug or change." + "reason": "Tokenizer/test changes for Whisper do not match CLIP/ViT test edits. Similar in being tests, but not the same change or bug." }, { "left": "pull_request:43291", "right": "pull_request:44815", "accept": false, - "reason": "Whisper tokenizer decoding changes are unrelated to FP8/dequant and model loading fixes. No shared underlying bug." + "reason": "Whisper decode/test cleanup and FP8 dequant fix are unrelated code paths and should not be merged as duplicates." }, { "left": "pull_request:43665", "right": "pull_request:44815", "accept": false, - "reason": "CLIP/ViT test edits are unrelated to the FP8/dequant implementation change; they do not fix the same concrete issue." + "reason": "The shared similarity is only that both touch tests somewhere, but the actual changes target unrelated model families and problems." } ] }, @@ -15505,7 +15492,7 @@ "soft_pairs": [ "pull_request:42975|pull_request:42976" ], - "prepared_review_unit_hash": "8eb4d0d874e9472100c5ddc8c59bdf4b00d18ed50b0dee117fb21648571708aa", + "prepared_review_unit_hash": "ff018e71bbef2f94bf22e7703b445616f369c8290cf00ecac795da660087aeec", "component_budget": { "node_count": 2, "item_count": 2, @@ -15527,18 +15514,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both pull requests are workflow dependency bumps in GitHub Actions, but they target different action families and different upgrade goals: one is specifically for Node 24 compatibility across several JS actions, while the other updates docker/build-push-action and related workflow actions to newer versions. They are related, but not the same concrete change.", - "confidence": 0.91, + "summary": "Two unrelated GitHub Actions maintenance PRs: one upgrades many workflow actions for Node 24 compatibility, the other bumps a smaller set of workflow/action versions, mainly docker/build-push-action. They overlap in files but do not look like the same fix or change.", + "confidence": 0.92, "canonical_issue_reason": null, - "canonical_pr_reason": "No single canonical PR: these are separate workflow maintenance changes with different action sets and different upgrade intents.", + "canonical_pr_reason": "PR #42975 is the better canonical representative because it is the broader, more clearly scoped workflow-action upgrade aimed at Node 24 compatibility, touching many of the same automation files.", "best_issue_reason": null, - "best_pr_reason": "PR 42975 is the broader umbrella upgrade and most representative of the cluster because it touches more workflows and has the explicit Node 24 compatibility goal.", + "best_pr_reason": "PR #42975 is the best overall representative of the cluster: it has the clearest stated purpose, the widest workflow coverage, and summarizes the common theme of action-version maintenance.", "soft_edge_verdicts": [ { "left": "pull_request:42975", "right": "pull_request:42976", "accept": false, - "reason": "Different underlying changes: 42975 upgrades several GitHub actions for Node 24 compatibility, while 42976 mainly bumps docker/build-push-action and a smaller set of workflow actions. Overlapping workflows alone is not enough to treat them as one PR." + "reason": "Both are workflow dependency bumps, but they upgrade different actions for different reasons: #42975 targets Node 24 compatibility across many workflows, while #42976 mainly updates docker/build-push-action to v6. Shared workflow filenames alone are not enough to treat them as the same underlying change." } ] }, @@ -15575,7 +15562,7 @@ "pull_request:43448|pull_request:45490", "pull_request:43451|pull_request:45490" ], - "prepared_review_unit_hash": "910ac29c81cb83f5e53b77100e4e2e96b52edbeb5694430c676b74c9918312c1", + "prepared_review_unit_hash": "6e144bf9570e2f5eb42f778e4a10f9079c2d1dc41c36d893b1121a5bb28f73a6", "component_budget": { "node_count": 4, "item_count": 4, @@ -15597,48 +15584,48 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These PRs are four separate model-addition changes (ViT NEPA, Molmo, Molmo2, CTSM). They share transformer-registration/docs patterns, but they do not fix the same underlying bug or implement the same change.", + "summary": "These four PRs are separate model-integration additions (ViT NEPA, Molmo, Molmo2, and CTSM). They share some boilerplate paths like docs and auto-registration, but they do not describe the same bug or the same code change.", "confidence": 0.98, "canonical_issue_reason": null, - "canonical_pr_reason": "No canonical PR exists here: the items are distinct model integrations rather than duplicates of one another, despite overlapping auto-mapping and docs edits.", + "canonical_pr_reason": "No canonical PR: the four PRs target different model families and different implementation files, so none is a duplicate representative.", "best_issue_reason": null, - "best_pr_reason": "PR 43451 is the most substantial and broadest integration, but it still represents a different model addition and should not be treated as the cluster canonical.", + "best_pr_reason": "PR 43451 is the broadest single integration, with the most files and review activity, but it is still not a duplicate of the others.", "soft_edge_verdicts": [ { "left": "pull_request:42978", "right": "pull_request:43448", "accept": false, - "reason": "Different model additions (ViT NEPA vs Molmo); shared docs/auto-registry files are incidental, not the same change." + "reason": "ViT NEPA vs Molmo: different model integrations and different code paths; overlap is only in shared framework boilerplate." }, { "left": "pull_request:42978", "right": "pull_request:43451", "accept": false, - "reason": "ViT NEPA and Molmo2 are unrelated model integrations with different code paths and model families." + "reason": "ViT NEPA vs Molmo2 are unrelated model additions, not the same fix or feature." }, { "left": "pull_request:42978", "right": "pull_request:45490", "accept": false, - "reason": "ViT NEPA and CTSM are separate model additions; overlap is only in common Transformers plumbing." + "reason": "ViT NEPA and CTSM add different models with different implementations; not mergeable as one duplicate PR." }, { "left": "pull_request:43448", "right": "pull_request:43451", "accept": false, - "reason": "Molmo and Molmo2 are related by naming but implement different model families and files, so they are not the same PR." + "reason": "Molmo vs Molmo2 are distinct model families; shared docs/auto-registration does not make them duplicates." }, { "left": "pull_request:43448", "right": "pull_request:45490", "accept": false, - "reason": "Molmo and CTSM address different model additions; shared registration/docs do not make them duplicates." + "reason": "Molmo and CTSM are separate model integrations touching different model code." }, { "left": "pull_request:43451", "right": "pull_request:45490", "accept": false, - "reason": "Molmo2 and CTSM are distinct model integrations with different concrete implementations and targets." + "reason": "Molmo2 and CTSM are different model additions with no shared underlying bug or concrete code-path change." } ] }, @@ -15671,7 +15658,7 @@ "pull_request:43488|pull_request:45476", "pull_request:43532|pull_request:45476" ], - "prepared_review_unit_hash": "0961dd83ad0107402e162949af3c6994d395a0d26617bc48dcc0028c0585057a", + "prepared_review_unit_hash": "c49b0af034f5601017c196ac612076acbebef736d291e1e6baaa8e38bd72d053", "component_budget": { "node_count": 3, "item_count": 3, @@ -15693,30 +15680,30 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "All three PRs are unrelated \u201c[don\u2019t merge]\u201d test/instrumentation changes, not the same underlying fix. The only pair with automatic soft similarity still does not look like one mergeable change.", - "confidence": 0.92, + "summary": "All three items are unrelated placeholder PRs that only overlap on an incidental test file. They do not appear to fix the same underlying bug or change, so no soft-edge merge is warranted.", + "confidence": 0.95, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #43532 is the most coherent representative: it focuses on CI/comment/diff-display plumbing and related test utilities, whereas #43488 and #45476 are more clearly ad hoc bot-check PRs with unrelated dummy edits.", + "canonical_pr_reason": "PR 43532 is the most central of the three: it touches multiple CI/workflow-related files and has the highest discussion/review activity, while the other two are narrower and look even more synthetic.", "best_issue_reason": null, - "best_pr_reason": "#43532 is the best single representative of this cluster because it has the clearest unified intent and the broadest but still coherent code-path change; the other two are more synthetic and less representative of a real fix/change.", + "best_pr_reason": "PR 43532 is the best representative of the cluster because it has the broadest change surface and connects to both other PRs via weak similarity, but the overlap is still incidental rather than a shared fix.", "soft_edge_verdicts": [ { "left": "pull_request:43488", "right": "pull_request:43532", "accept": false, - "reason": "Both touch tests/CI-adjacent files, but one is a setup/version/dependency bump and the other is workflow/test-diff plumbing; they are not the same concrete change and would not plausibly merge into one PR." + "reason": "These are different placeholder changes: one is about setup/version/dependency metadata, the other is about workflow/comment CI plumbing. Shared vit test presence is incidental and not the same code-path problem." }, { "left": "pull_request:43488", "right": "pull_request:45476", "accept": false, - "reason": "#43488 is a packaging/version check PR, while #45476 is a CI workflow caller plus unrelated model/test edits; the overlap is only superficial." + "reason": "One PR changes packaging/dependency setup, the other adds a CI workflow caller and unrelated tokenizer/test noise. No concrete shared bug or mergeable combined fix is evident." }, { "left": "pull_request:43532", "right": "pull_request:45476", "accept": false, - "reason": "Despite shared test filenames and soft similarity, #43532 is about showing diffs in CI/comments, while #45476 is about calling a CI workflow and contains unrelated dummy edits; not the same underlying bug or change." + "reason": "Both touch CI-adjacent files, but they address different areas: comment/scheduled/slack workflow logic versus a new PR CI caller workflow plus tokenizer/test edits. The overlap is too broad and superficial to treat as the same change." } ] }, @@ -15746,7 +15733,7 @@ "soft_pairs": [ "pull_request:43612|pull_request:43613" ], - "prepared_review_unit_hash": "b50a142e8dde3e88a76a40cf16f5af11b467b9b7b2b69e52f72c16ff8921689f", + "prepared_review_unit_hash": "814e77809d208172e5bc54330b49f4b5c45becffc7de3b8b6ac40df6a6bb140c", "component_budget": { "node_count": 2, "item_count": 2, @@ -15768,18 +15755,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are two separate PRs adding different pipeline features: one for Promptable Concept Segmentation and one for Promptable Visual Segmentation. They share scaffolding and doc/test patterns, but they do not implement the same underlying change or fix the same concrete code path.", - "confidence": 0.94, + "summary": "These are two related but distinct feature PRs: one adds a Promptable Concept Segmentation pipeline, the other adds a Promptable Visual Segmentation pipeline. They share documentation and pipeline infrastructure changes, but target different tasks/models and are not the same underlying change.", + "confidence": 0.93, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 43613 is the better representative because it is the broader, more feature-complete pipeline addition, with more backend/model coverage and a larger integration surface.", + "canonical_pr_reason": "PR #43613 is the better canonical representative for the cluster because it is the broader pipeline addition, with more model integrations and test coverage, while #43612 is a narrower sibling feature.", "best_issue_reason": null, - "best_pr_reason": "PR 43613 is the strongest standalone candidate to represent the cluster since it is more comprehensive and touches more of the related visual-segmentation stack.", + "best_pr_reason": "PR #43613 is the best representative overall for this cluster: it has the wider scope, touches more related code paths, and is the more central of the two additions.", "soft_edge_verdicts": [ { "left": "pull_request:43612", "right": "pull_request:43613", "accept": false, - "reason": "They are different features: concept segmentation vs visual segmentation. Shared docs/tests/registration changes are generic pipeline boilerplate, not evidence of the same bug or change." + "reason": "Do not merge: these add different pipelines for different segmentation tasks. The overlap is mostly shared docs/registry plumbing, not the same concrete bug or feature." } ] }, @@ -15809,12 +15796,12 @@ "soft_pairs": [ "pull_request:43651|pull_request:43663" ], - "prepared_review_unit_hash": "8af79e504fb82a09872d8e296bb58c4ea45f5b069846673dc44bd3f9f7917e29", + "prepared_review_unit_hash": "7525d53a7ac22053874b457dedc8a829a6000c4690c14e75d5d8a9b5e3169d11", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4291, + "serialized_chars": 4292, "estimated_input_tokens": 1073, "estimated_eval_tokens": 2402 }, @@ -15822,7 +15809,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4291, + "serialized_chars": 4292, "estimated_input_tokens": 1073, "estimated_eval_tokens": 2402 }, @@ -15831,18 +15818,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are two distinct Trainer PRs that both modify src/transformers/trainer.py but address different behaviors: one changes gradient-accumulation loss scaling, the other adds an overridable signature-column helper for column filtering. They are related in scope but not duplicates.", + "summary": "Two unrelated open PRs in `trainer.py`: one changes gradient-accumulation loss scaling behavior, the other refactors signature-column filtering for custom trainers/PEFT models. They share a file and subsystem but not the same underlying bug or change.", "confidence": 0.96, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 43651 is the more concrete bugfix-style change: it directly alters training_step loss scaling behavior and adds a helper to let custom trainers control that path.", + "canonical_pr_reason": "No canonical PR: these are distinct feature/refactor PRs with different code paths and goals, not duplicates of one another.", "best_issue_reason": null, - "best_pr_reason": "PR 43651 is the best representative because it targets a specific runtime behavior with clear user impact, whereas 43663 is a separate Trainer extensibility refactor.", + "best_pr_reason": "Neither PR is a duplicate candidate; if a representative is needed, `#43651` is slightly more central to trainer training-step behavior, but it should not be treated as the same change as `#43663`.", "soft_edge_verdicts": [ { "left": "pull_request:43651", "right": "pull_request:43663", "accept": false, - "reason": "Do not merge: they touch the same file but fix different code paths. 43651 changes gradient-accumulation loss scaling in training_step; 43663 adds a signature-column helper for column filtering. Same subsystem, not the same underlying bug/change." + "reason": "Both touch `trainer.py`, but one modifies loss scaling in `training_step` and the other adds an overrideable signature-column helper for data filtering. Different bugs/changes, not mergeable as one PR." } ] }, @@ -15874,56 +15861,56 @@ "pull_request:43656|pull_request:43842", "pull_request:43836|pull_request:43842" ], - "prepared_review_unit_hash": "43159e0cb5863f756b7a620adbbef6fbfdaf6c7bde39f192acd7c62633ec7fb3", + "prepared_review_unit_hash": "aaae9ca124a545dcbb9947a6aed2b8214570914fb751060eaf028752f2cbe194", "component_budget": { "node_count": 3, "item_count": 3, "soft_pair_count": 2, - "serialized_chars": 5991, - "estimated_input_tokens": 1498, - "estimated_eval_tokens": 3252 + "serialized_chars": 5993, + "estimated_input_tokens": 1499, + "estimated_eval_tokens": 3254 }, "budget": { "node_count": 3, "item_count": 3, "soft_pair_count": 2, - "serialized_chars": 5991, - "estimated_input_tokens": 1498, - "estimated_eval_tokens": 3252 + "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": "#43842 remains the best standalone PR: it is the narrowest, most targeted fix for the `TypeAdapter`/pydantic import NameError in `serve.py`. #43836 is related but looks broader in the visible diff, with a serve.py refactor and typing cleanup, so it is not safe to treat as the same concrete fix. #43656 is also related but appears to be a different, wider annotation cleanup rather than the same bugfix.", - "confidence": 0.82, + "summary": "All three PRs are small, overlapping fixes in `src/transformers/cli/serve.py` around `TypeAdapter`/typing evaluation at import time. PR 43842 is the cleanest and most directly scoped fix; 43836 looks like a close variant of the same bug; 43656 is similar but less clearly the same concrete failure mode.", + "confidence": 0.74, "canonical_issue_reason": null, - "canonical_pr_reason": "#43842 is the cleanest canonical PR because it directly targets the `TypeAdapter` NameError with a minimal, focused change and the clearest bug-specific intent.", + "canonical_pr_reason": "PR 43842 is the most focused and explicit fix for the runtime `TypeAdapter` NameError, with minimal scope and a clear explanation tied to pydantic being absent.", "best_issue_reason": null, - "best_pr_reason": "#43842 is the best overall PR because it is the most precise and least disruptive standalone fix for the runtime import/typing problem.", + "best_pr_reason": "PR 43842 best represents the cluster because it directly addresses the concrete import-time `TypeAdapter` NameError with the smallest, most targeted change set.", "soft_edge_verdicts": [ { "left": "pull_request:43656", "right": "pull_request:43842", "accept": false, - "reason": "They overlap in `serve.py` and `TypeAdapter`, but #43656\u2019s visible diff is broader typing cleanup/refactoring, not clearly the same concrete pydantic import bugfix as #43842." + "reason": "Both touch the same file and mention `TypeAdapter`, but 43656 appears to be a broader typing/annotation cleanup and has a different issue target; the evidence is not strong enough to say it is the same exact bug/fix." }, { "left": "pull_request:43836", "right": "pull_request:43842", - "accept": false, - "reason": "The packet only shows soft similarity and a shared issue target; #43836\u2019s visible diff is a broader `serve.py` refactor plus annotations cleanup, so it is not confidently the same fix or cleanly mergeable into one PR with #43842." + "accept": true, + "reason": "These are very likely the same underlying `TypeAdapter` NameError fix in `serve.py`: both target the same issue, both are tiny annotation-only changes, and both avoid runtime evaluation of `TypeAdapter`." } ] }, "evaluator_result": { - "accept": false, - "feedback": "The verdict on 43836 vs 43842 is too strong for the evidence in the packet. 43842 is a narrow `TypeAdapter`/pydantic import fix, but 43836\u2019s visible diff is a broader `serve.py` refactor plus annotations cleanup; the packet only shows soft similarity and a shared issue target, not enough to confidently say the two PRs are the same concrete bugfix or cleanly mergeable into one PR. The rejection of 43656 vs 43842 is conservative and fine, but the overall summary overstates the 43836 relationship." + "accept": true, + "feedback": "The summary stays within the packet evidence and is suitably conservative. PR 43842 is clearly the smallest, most targeted TypeAdapter fix. The 43836/43842 soft-accept is defensible because they share the same issue target and file and appear to be alternative solutions to the same import-time typing problem. The 43656/43842 reject is also conservative given the weaker linkage and different issue target." }, "evaluator_used": true, - "retried": true, - "accepted_nontrivial_soft_edge": false, + "retried": false, + "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null }, @@ -16012,7 +15999,7 @@ "soft_pairs": [ "pull_request:43973|pull_request:45283" ], - "prepared_review_unit_hash": "d6c7bd944f59acfc27935dd7a1b4ea5eb4511346f4a24dd7af3c5b6c181287cd", + "prepared_review_unit_hash": "943436408822dc567b5d61dd36502cbe2f8f715d5c54c74363f63b0d61951814", "component_budget": { "node_count": 2, "item_count": 2, @@ -16034,18 +16021,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These two PRs are unrelated: #43973 adds a new Lfm2Audio model integration with docs/tests, while #45283 adds GGUF loading support for Qwen3.5. They do not share the same underlying bug or code change.", - "confidence": 0.99, + "summary": "These two PRs are unrelated: one adds the new Lfm2Audio model package and docs, while the other adds Qwen3.5 GGUF loading support via integration mappings. They do not target the same bug or code path.", + "confidence": 0.97, "canonical_issue_reason": null, - "canonical_pr_reason": "No canonical PR: the changes target different features and code paths, so they are not duplicates or mergeable as one fix.", + "canonical_pr_reason": "PR 43973 is the better canonical PR candidate for this cluster only because it is a self-contained model addition with an explicit issue target and substantial implementation/test/docs coverage.", "best_issue_reason": null, - "best_pr_reason": "Neither PR is a strong global representative of the other; they address separate subsystems and different user-visible goals.", + "best_pr_reason": "PR 43973 is the strongest representative of the cluster, but the cluster is not a duplicate set; PR 45283 is a separate GGUF-loading change for a different model family.", "soft_edge_verdicts": [ { "left": "pull_request:43973", "right": "pull_request:45283", "accept": false, - "reason": "Different workstreams: Lfm2Audio model addition vs Qwen3.5 GGUF mapping/loading support. No shared concrete bug or fix path." + "reason": "Different features and code paths: Lfm2Audio model/docs/tests versus Qwen3.5 GGUF loading mappings. Shared transformer infrastructure is too broad to indicate the same underlying change." } ] }, @@ -16125,13 +16112,13 @@ "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:44059|pull_request:44722", + "pull_request:44085|pull_request:44086", "pull_request:44098|pull_request:44101", "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:44017|pull_request:44161", "pull_request:44068|pull_request:44071", "pull_request:44071|pull_request:44086", "pull_request:44059|pull_request:44071", @@ -16147,8 +16134,8 @@ "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:44056|pull_request:44161", "pull_request:43997|pull_request:44161", "pull_request:44074|pull_request:44161", "pull_request:43997|pull_request:44071", @@ -16172,364 +16159,298 @@ "pull_request:44044|pull_request:44154", "pull_request:44024|pull_request:44076" ], - "prepared_review_unit_hash": "ade9e98caf599ae9204c27749bc325fafea1163ff67b460a19f8f2f6930a048f", + "prepared_review_unit_hash": "cd98c8fbcb8a4ff605858352e22680c7229a3a18d58c4e2a3526abbd1ced83ea", "component_budget": { "node_count": 28, "item_count": 28, "soft_pair_count": 77, - "serialized_chars": 84500, - "estimated_input_tokens": 21125, - "estimated_eval_tokens": 42506 + "serialized_chars": 84503, + "estimated_input_tokens": 21126, + "estimated_eval_tokens": 42508 }, "budget": { "node_count": 28, "item_count": 28, "soft_pair_count": 72, - "serialized_chars": 76932, - "estimated_input_tokens": 19233, - "estimated_eval_tokens": 38722 + "serialized_chars": 76936, + "estimated_input_tokens": 19234, + "estimated_eval_tokens": 38724 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a topical set of model-specific output-tracing refactors. Most PRs share the same modernization pattern but touch different model code paths, so they are not duplicates. The only clear duplicate-like subclusters are GPT-Neo and GPTJ/CodeGen variants of the same decorator migration.", - "confidence": 0.71, - "canonical_issue_reason": null, - "canonical_pr_reason": "PR 44722 is the best representative of the GPTJ/CodeGen output-tracing refactor subcluster: it is the latest, broadest instance of that exact decorator migration and overlaps directly with 44066 and 44085.", - "best_issue_reason": null, - "best_pr_reason": "PR 44722 is the clearest single representative for the repeated GPTJ/CodeGen standardized output-tracing change, with the most complete-looking implementation among the near-duplicates.", + "summary": "This cluster is mostly a refactor series for standardized output tracing across many model implementations. Most pairs are not true duplicates because they touch different models/code paths. The only clear duplicate-like subclusters are GPT-Neo (#44018/#44068) and the GPT-J family (#44066/#44085/#44722), which overlap on the same model files and the same decorator-based output-tracing change.", + "confidence": 0.79, + "canonical_issue_reason": "Issue #43979 appears to be the umbrella tracking issue for the standardized output-tracing refactor across models, which explains the shared lineage of many PRs in this cluster.", + "canonical_pr_reason": "PR #44722 is the best canonical PR: it is the most complete representative of the GPT-J/CodeGen output-tracing refactor, directly targets the shared issue, and covers the same concrete decorator-based code path as the overlapping GPT-J PRs.", + "best_issue_reason": "#43979 is the best issue anchor because it is the shared tracker for the broader refactor work, even though it is an umbrella rather than a narrow bug report.", + "best_pr_reason": "#44722 is the strongest single PR to keep as the representative example of the shared change set because it is explicit, broader than the earlier GPT-J-only variants, and aligns with the same implementation path.", "soft_edge_verdicts": [ { "left": "pull_request:43996", "right": "pull_request:44085", "accept": false, - "reason": "Different model families and files; only the output-tracing theme is shared." + "reason": "Different models and files (CVT/FNet vs GPT-J); same umbrella refactor theme, but not the same concrete code path." }, { "left": "pull_request:43996", "right": "pull_request:44044", "accept": false, - "reason": "Both are output-tracing refactors, but they target different model code paths." + "reason": "Both are output-tracing refactors, but they target unrelated model implementations (CVT/FNet vs DeBERTa-v2)." }, { "left": "pull_request:44066", "right": "pull_request:44085", "accept": true, - "reason": "Both modify GPTJ output tracing in modeling_gptj.py and look like the same decorator-migration change." + "reason": "Both touch GPT-J output tracing in the same model file and appear to be overlapping/superset variants of the same change." }, { "left": "pull_request:44007", "right": "pull_request:44072", "accept": false, - "reason": "RegNet/ResNet-family refactors with different implementations; same general pattern, not the same change." + "reason": "RegNet vs EfficientNet: same style of refactor, but different model code paths." }, { "left": "pull_request:44072", "right": "pull_request:44722", "accept": false, - "reason": "EfficientNet and GPTJ/CodeGen are unrelated code paths despite similar output-tracing cleanup." + "reason": "EfficientNet and GPT-J/CodeGen are unrelated model implementations; shared issue link is not enough." }, { "left": "pull_request:44066", "right": "pull_request:44072", "accept": false, - "reason": "GPTJ/CodeGen vs EfficientNet: different models and different concrete changes." + "reason": "GPT-J/CodeGen vs EfficientNet are separate model refactors, not the same underlying bug or change." }, { "left": "pull_request:44013", "right": "pull_request:44044", "accept": false, - "reason": "MobileNetV2 and DeBERTa are unrelated implementations." + "reason": "MobileNetV2 and DeBERTa-v2 are different code paths despite similar output-tracing cleanup." }, { "left": "pull_request:44066", "right": "pull_request:44086", "accept": false, - "reason": "GPTJ/CodeGen refactor and MGP-STR refactor are different model-specific changes." + "reason": "GPT-J/CodeGen refactor vs MGP-STR refactor; different model internals." }, { "left": "pull_request:44066", "right": "pull_request:44071", "accept": false, - "reason": "GPTJ/CodeGen and MPT are different code paths." + "reason": "GPT-J/CodeGen and MPT are separate architectures; not the same change." }, { "left": "pull_request:44018", "right": "pull_request:44068", "accept": true, - "reason": "Both are GPT-Neo output-tracing refactors in the same file and appear to be the same decorator migration." + "reason": "Both are GPT-Neo output-tracing refactors over the same model file/test set and look like duplicate/superset variants." }, { "left": "pull_request:44066", "right": "pull_request:44068", "accept": false, - "reason": "GPTJ/CodeGen versus GPT-Neo; same pattern, different model code." + "reason": "GPT-J/CodeGen vs GPT-Neo: similar refactor pattern, but different model code paths." }, { "left": "pull_request:44018", "right": "pull_request:44066", "accept": false, - "reason": "GPT-Neo and GPTJ/CodeGen are distinct model changes." + "reason": "GPT-Neo and GPT-J/CodeGen are distinct model refactors." }, { "left": "pull_request:44019", "right": "pull_request:44722", "accept": false, - "reason": "ResNet output tracing and GPTJ/CodeGen output tracing are unrelated changes." + "reason": "ResNet and GPT-J/CodeGen are unrelated; same issue target does not make them duplicates." }, { "left": "pull_request:44086", "right": "pull_request:44722", "accept": false, - "reason": "MGP-STR and GPTJ/CodeGen are different model code paths." + "reason": "MGP-STR and GPT-J/CodeGen are different model implementations and not mergeable as one change." }, { "left": "pull_request:44019", "right": "pull_request:44085", "accept": false, - "reason": "ResNet and GPTJ are different refactors." + "reason": "ResNet vs GPT-J output tracing; different code paths." }, { "left": "pull_request:44018", "right": "pull_request:44019", "accept": false, - "reason": "GPT-Neo and ResNet are unrelated changes." + "reason": "GPT-Neo and ResNet are unrelated model refactors." }, { "left": "pull_request:44019", "right": "pull_request:44071", "accept": false, - "reason": "ResNet and MPT are unrelated model refactors." + "reason": "ResNet vs MPT: same refactor theme, different concrete implementation." }, { "left": "pull_request:44068", "right": "pull_request:44116", "accept": false, - "reason": "GPT-Neo versus Flaubert; same theme, different model implementation." + "reason": "GPT-Neo and Flaubert/XLM are separate model code paths." }, { "left": "pull_request:44068", "right": "pull_request:44722", "accept": false, - "reason": "GPT-Neo and GPTJ/CodeGen are different code paths." + "reason": "GPT-Neo is a different model family from GPT-J/CodeGen." }, { "left": "pull_request:44071", "right": "pull_request:44722", "accept": false, - "reason": "MPT and GPTJ/CodeGen are unrelated concrete changes." + "reason": "MPT and GPT-J/CodeGen are unrelated implementations." }, { "left": "pull_request:44018", - "right": "pull_request:44116", + "right": "pull_request:44129", "accept": false, - "reason": "GPT-Neo and Flaubert are different model-specific refactors." + "reason": "GPT-Neo vs SpeechT5: different model families, only a shared refactor style." }, { "left": "pull_request:44018", "right": "pull_request:44086", "accept": false, - "reason": "GPT-Neo and MGP-STR are unrelated changes." + "reason": "GPT-Neo and MGP-STR are different code paths." }, { "left": "pull_request:44018", "right": "pull_request:44085", "accept": false, - "reason": "GPT-Neo and GPTJ are different model families." + "reason": "GPT-Neo and GPT-J are related only at a high level; not the same model change." }, { "left": "pull_request:44068", "right": "pull_request:44086", "accept": false, - "reason": "GPT-Neo and MGP-STR are not the same underlying change." + "reason": "GPT-Neo and MGP-STR are unrelated model refactors." }, { "left": "pull_request:44071", "right": "pull_request:44085", "accept": false, - "reason": "MPT and GPTJ are different model refactors." + "reason": "MPT vs GPT-J are distinct model implementations." }, { "left": "pull_request:44059", "right": "pull_request:44086", "accept": false, - "reason": "GPT-2 and MGP-STR touch different code paths." - }, - { - "left": "pull_request:44085", - "right": "pull_request:44086", - "accept": false, - "reason": "GPTJ and MGP-STR are unrelated model changes." + "reason": "GPT-2 and MGP-STR are different models; not a duplicate." }, { "left": "pull_request:44059", "right": "pull_request:44722", "accept": false, - "reason": "GPT-2 output tracing and GPTJ/CodeGen output tracing are different changes." + "reason": "GPT-2 output-tracing refactor is separate from GPT-J/CodeGen." }, { - "left": "pull_request:44098", - "right": "pull_request:44101", + "left": "pull_request:44085", + "right": "pull_request:44086", "accept": false, - "reason": "ViLT and Flaubert/XLM are different models and different refactors." + "reason": "GPT-J and MGP-STR are unrelated code paths." }, { - "left": "pull_request:44073", + "left": "pull_request:44098", "right": "pull_request:44101", "accept": false, - "reason": "VisualBert and Flaubert/XLM are distinct code paths." + "reason": "ViLT and Flaubert/XLM are different model families." }, { "left": "pull_request:44018", "right": "pull_request:44071", "accept": false, - "reason": "GPT-Neo and MPT are unrelated model changes." + "reason": "GPT-Neo and MPT are separate implementations." }, { "left": "pull_request:44017", "right": "pull_request:44722", "accept": false, - "reason": "SegFormer and GPTJ/CodeGen are different model refactors." + "reason": "SegFormer and GPT-J/CodeGen are unrelated model refactors." }, { "left": "pull_request:44017", - "right": "pull_request:44161", + "right": "pull_request:44073", "accept": false, - "reason": "SegFormer and LongT5 are unrelated changes." + "reason": "SegFormer vs VisualBert: same pattern, different code paths." }, { "left": "pull_request:44017", - "right": "pull_request:44073", + "right": "pull_request:44161", "accept": false, - "reason": "SegFormer and VisualBert touch different code paths." + "reason": "SegFormer and LongT5 are different models; not a true duplicate." }, { "left": "pull_request:44068", "right": "pull_request:44071", "accept": false, - "reason": "GPT-Neo and MPT are different refactors." + "reason": "GPT-Neo and MPT are different architectures." }, { "left": "pull_request:44071", - "right": "pull_request:44086", - "accept": false, - "reason": "MPT and MGP-STR are different model changes." - }, - { - "left": "pull_request:44059", - "right": "pull_request:44071", - "accept": false, - "reason": "GPT-2 and MPT are unrelated concrete changes." - }, - { - "left": "pull_request:43997", - "right": "pull_request:44073", - "accept": false, - "reason": "RegNet and VisualBert are different model code paths." - }, - { - "left": "pull_request:43997", - "right": "pull_request:44056", + "right": "pull_request:44076", "accept": false, - "reason": "RegNet and MPNet are unrelated refactors." + "reason": "MPT and ImageGPT are unrelated code paths." }, { "left": "pull_request:44074", "right": "pull_request:44101", "accept": false, - "reason": "TextNet and Flaubert/XLM are different model changes." - }, - { - "left": "pull_request:44056", - "right": "pull_request:44161", - "accept": false, - "reason": "MPNet and LongT5 are unrelated changes." + "reason": "TextNet and Flaubert/XLM are different model implementations." }, { "left": "pull_request:43997", - "right": "pull_request:44018", - "accept": false, - "reason": "RegNet and GPT-Neo are different model implementations." - }, - { - "left": "pull_request:44074", - "right": "pull_request:44161", + "right": "pull_request:44056", "accept": false, - "reason": "TextNet and LongT5 are unrelated refactors." + "reason": "RegNet and MPNet are separate model refactors." }, { "left": "pull_request:43997", - "right": "pull_request:44071", - "accept": false, - "reason": "RegNet and MPT are different code paths." - }, - { - "left": "pull_request:44010", - "right": "pull_request:44071", - "accept": false, - "reason": "SqueezeBert and MPT are unrelated model changes." - }, - { - "left": "pull_request:44010", - "right": "pull_request:44073", - "accept": false, - "reason": "SqueezeBert and VisualBert are different model refactors." - }, - { - "left": "pull_request:44018", - "right": "pull_request:44129", + "right": "pull_request:44018", "accept": false, - "reason": "GPT-Neo and SpeechT5 are unrelated changes." + "reason": "RegNet and GPT-Neo are unrelated." }, { "left": "pull_request:43997", - "right": "pull_request:44074", + "right": "pull_request:44161", "accept": false, - "reason": "RegNet and TextNet are different model code paths." + "reason": "RegNet and LongT5 are different model code paths." }, { - "left": "pull_request:43996", + "left": "pull_request:44056", "right": "pull_request:44161", "accept": false, - "reason": "CVT/FNet and LongT5 are unrelated model refactors." + "reason": "MPNet and LongT5 are unrelated model-specific changes." }, { "left": "pull_request:43996", - "right": "pull_request:44017", + "right": "pull_request:44161", "accept": false, - "reason": "CVT/FNet and SegFormer are different model changes." + "reason": "CVT/FNet and LongT5 are different implementations; shared refactor theme is not enough." }, { "left": "pull_request:44010", - "right": "pull_request:44044", - "accept": false, - "reason": "SqueezeBert and DeBERTa are unrelated code paths." - }, - { - "left": "pull_request:44044", - "right": "pull_request:44074", - "accept": false, - "reason": "DeBERTa and TextNet are different model refactors." - }, - { - "left": "pull_request:44044", - "right": "pull_request:44154", + "right": "pull_request:44071", "accept": false, - "reason": "DeBERTa and VITS are unrelated changes." + "reason": "SqueezeBert and MPT are different model code paths." }, { "left": "pull_request:44024", "right": "pull_request:44076", "accept": false, - "reason": "FocalNet and ImageGPT are different model code paths." + "reason": "FocalNet and ImageGPT are separate model implementations; same output-tracing style only." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is grounded in the packet and stays conservative about non-duplicate model-specific refactors. The accepted soft-edge pairs (GPT-Neo and GPTJ/CodeGen variants) look mergeable into single PRs for the same bugfix, and the rejected pairs are appropriately cautious." + "feedback": "Grounded in the packet, and the soft-edge accepts are conservative: both accepted pairs are same-model, same-file overlaps that look mergeable into one PR for the same refactor. The rejections are appropriately cautious about different model code paths." }, "evaluator_used": true, "retried": false, @@ -16564,20 +16485,20 @@ "pull_request:43997|pull_request:44010", "pull_request:44010|pull_request:44129" ], - "prepared_review_unit_hash": "ab5fd9960a19b8b59ff4e0a037ea9f39eb3a3087d500964c589aa796c4477df9", + "prepared_review_unit_hash": "8f6f4ead4fb877334034eba773bcf856d0142d83356f688fa2bd8971c0ecf2f5", "component_budget": { "node_count": 28, "item_count": 28, "soft_pair_count": 77, - "serialized_chars": 84500, - "estimated_input_tokens": 21125, - "estimated_eval_tokens": 42506 + "serialized_chars": 84503, + "estimated_input_tokens": 21126, + "estimated_eval_tokens": 42508 }, "budget": { "node_count": 6, "item_count": 6, "soft_pair_count": 5, - "serialized_chars": 13068, + "serialized_chars": 13067, "estimated_input_tokens": 3267, "estimated_eval_tokens": 6790 }, @@ -16586,42 +16507,42 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These PRs are all related to the same framework-wide output-tracing refactor pattern, but they touch different model implementations and code paths, so they should not be treated as duplicates or merged together.", + "summary": "All items are model-specific refactors toward standardized output tracing/capture_outputs. They share a theme, but each PR targets a different architecture and code path, so they are not true duplicates.", "confidence": 0.93, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #44129 is the strongest representative because it has an explicit issue target and a concrete standardized output-capture refactor in one model implementation.", + "canonical_pr_reason": "PR 44129 is the best cluster representative: it has an explicit issue target, a clear output-tracing refactor scope, and enough context to anchor the shared theme without conflating distinct model implementations.", "best_issue_reason": null, - "best_pr_reason": "PR #44129 is the best global representative of the cluster: it is clearly titled, tied to a tracking issue, and shows a substantive end-to-end output-tracing migration.", + "best_pr_reason": "PR 44129 is the strongest single representative because it is explicitly linked to an issue and cleanly expresses the standardized output capture refactor pattern seen across the cluster.", "soft_edge_verdicts": [ { "left": "pull_request:44073", "right": "pull_request:44074", "accept": false, - "reason": "Same refactor theme, but VisualBert and TextNet are separate model code paths; this is parallel implementation work, not one shared bug/change." + "reason": "Both are output-tracing refactors, but they touch different model implementations (VisualBert vs TextNet) and different forward paths; they are related patterns, not the same concrete change." }, { "left": "pull_request:44056", "right": "pull_request:44073", "accept": false, - "reason": "Both standardize output tracing, but MPNet and VisualBert have distinct model internals and files, so they are not the same concrete PR." + "reason": "These are separate model-specific refactors in MPNet and VisualBert. Similar decorator/capture_outputs work is not enough to treat them as one duplicate PR." }, { "left": "pull_request:44056", "right": "pull_request:44074", "accept": false, - "reason": "Related output-capture migration only; MPNet and TextNet changes are independent model-specific refactors." + "reason": "MPNet and TextNet are different code paths with independent refactors; they may share a migration pattern, but they are not the same underlying bug or change." }, { "left": "pull_request:43997", "right": "pull_request:44010", "accept": false, - "reason": "RegNet and SqueezeBERT are unrelated implementations; similarity is only at the refactor pattern level, not the underlying change." + "reason": "RegNet and SqueezeBert are unrelated model files and implementations. The diff theme is similar, but this is not one mergeable PR for the same fix." }, { "left": "pull_request:44010", "right": "pull_request:44129", "accept": false, - "reason": "Both update output tracing, but they target different models and different forward paths, so they should not be collapsed into one duplicate set." + "reason": "SqueezeBert and SpeechT5 each implement their own output-tracing refactor in separate model code paths; this is a shared migration style, not a duplicate change." } ] }, @@ -16700,268 +16621,268 @@ "pull_request:44028|pull_request:44030", "pull_request:44029|pull_request:44030" ], - "prepared_review_unit_hash": "63c1b9e1f428e203d9a81d7497ed245031af481a5f61f6d0814cbce058b061a8", + "prepared_review_unit_hash": "e8084fb0789b2c511b36c9fcee64ab494a101e04502b2f87854312a4b2b24e80", "component_budget": { "node_count": 13, "item_count": 13, "soft_pair_count": 39, - "serialized_chars": 39780, - "estimated_input_tokens": 9945, - "estimated_eval_tokens": 20146 + "serialized_chars": 39782, + "estimated_input_tokens": 9946, + "estimated_eval_tokens": 20148 }, "budget": { "node_count": 13, "item_count": 13, "soft_pair_count": 39, - "serialized_chars": 39780, - "estimated_input_tokens": 9945, - "estimated_eval_tokens": 20146 + "serialized_chars": 39782, + "estimated_input_tokens": 9946, + "estimated_eval_tokens": 20148 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "This cluster is a set of similarly themed but model-specific output-tracing refactors. They share implementation patterns and naming, but they do not look like the same underlying bug or change, so they should not be deduplicated together.", - "confidence": 0.97, + "summary": "These PRs are all model-specific implementations of the same broad output-tracing refactor theme, but they target different architectures and code paths. They do not look like duplicates of one another, so no soft edges should be merged.", + "confidence": 0.91, "canonical_issue_reason": null, - "canonical_pr_reason": "No canonical PR: these are separate model-specific refactors in different code paths, not one change that subsumes the others.", + "canonical_pr_reason": "PR 44003 is the strongest representative of the cluster because it makes the most infrastructure-like tracing changes and spans closely related Mamba/Falcon-Mamba paths, rather than only applying the pattern in one model.", "best_issue_reason": null, - "best_pr_reason": "No best PR for deduping; at most these are related by a common refactor pattern, but each PR targets a different model implementation.", + "best_pr_reason": "44003 is the clearest anchor PR for this cluster: it has the broadest tracing machinery changes and best captures the underlying refactor theme.", "soft_edge_verdicts": [ { "left": "pull_request:43998", "right": "pull_request:44002", "accept": false, - "reason": "timm_backbone and upernet are different model code paths; same refactor theme, not the same change." + "reason": "Same refactor pattern, but different model implementations (`timm_backbone` vs `upernet`) and no shared concrete code path." }, { "left": "pull_request:43998", "right": "pull_request:44000", "accept": false, - "reason": "Different models and files; only the output-tracing pattern matches." + "reason": "Both are output-tracing refactors, but they affect different models and different forward paths, so they are not the same change." }, { "left": "pull_request:43998", "right": "pull_request:44001", "accept": false, - "reason": "Separate model implementations; no shared concrete bug or patch." + "reason": "Different model families (`timm_backbone` vs `univnet`) with separate forward logic; similar cleanup, not the same bug or patch." }, { "left": "pull_request:43998", "right": "pull_request:44003", "accept": false, - "reason": "timm_backbone vs mamba/falcon_mamba are unrelated model internals." + "reason": "These touch unrelated model code paths (`timm_backbone` vs `mamba/falcon_mamba`), so they are not mergeable as one PR." }, { "left": "pull_request:43998", "right": "pull_request:44029", "accept": false, - "reason": "Different architectures and code paths; similarity is only at the refactor level." + "reason": "Shared theme only; `timm_backbone` and `rwkv` are separate architectures with different tracing behavior." }, { "left": "pull_request:43999", "right": "pull_request:44001", "accept": false, - "reason": "MobileNetV1 and UnivNet are unrelated models; same helper migration pattern only." + "reason": "Both refactor output handling, but `mobilenet_v1` and `univnet` are distinct implementations with no shared underlying bug." }, { "left": "pull_request:43999", "right": "pull_request:44002", "accept": false, - "reason": "Different model families and files; not the same underlying fix." + "reason": "`mobilenet_v1` and `upernet` are separate model paths; this is the same pattern applied independently." }, { "left": "pull_request:43999", "right": "pull_request:44025", "accept": false, - "reason": "MobileNetV1 vs Depth Anything are separate changes in separate models." + "reason": "Different model stacks (`mobilenet_v1` vs `depth_anything/prompt_depth_anything`); no common code path." }, { "left": "pull_request:43999", "right": "pull_request:44026", "accept": false, - "reason": "Different model code paths; shared tracing refactor theme is too broad." + "reason": "These are unrelated encoder/decoder and classification-style model paths, despite both adding tracing helpers." }, { "left": "pull_request:43999", "right": "pull_request:44027", "accept": false, - "reason": "MobileNetV1 and Speech Encoder-Decoder are unrelated implementations." + "reason": "`mobilenet_v1` and `speech_encoder_decoder` are different subsystems; not the same underlying fix." }, { "left": "pull_request:44000", "right": "pull_request:44001", "accept": false, - "reason": "Vision-Text Dual Encoder and UnivNet do not fix the same concrete issue." + "reason": "`vision_text_dual_encoder` and `univnet` have different forward paths and output structures, so this is not one duplicate change." }, { "left": "pull_request:44000", "right": "pull_request:44002", "accept": false, - "reason": "Different model-specific API refactors; not mergeable as one PR." + "reason": "Same refactor motif, but different models and APIs; no shared concrete code path." }, { "left": "pull_request:44000", "right": "pull_request:44003", "accept": false, - "reason": "Vision-Text Dual Encoder vs Mamba/Falcon Mamba are distinct code paths." + "reason": "`vision_text_dual_encoder` and `mamba/falcon_mamba` are unrelated code paths, so they should not be merged together." }, { "left": "pull_request:44000", "right": "pull_request:44025", "accept": false, - "reason": "Separate models and separate output contracts; not a duplicate patch." + "reason": "These modify different model families and different output contracts; only the general theme matches." }, { "left": "pull_request:44001", "right": "pull_request:44002", "accept": false, - "reason": "UnivNet and UperNet are unrelated models; only the refactor pattern is shared." + "reason": "`univnet` and `upernet` are separate model implementations; no shared bug or single PR target." }, { "left": "pull_request:44001", "right": "pull_request:44003", "accept": false, - "reason": "Different model families and different implementation targets." + "reason": "Different architectures and different forward behavior; not the same underlying change." }, { "left": "pull_request:44001", "right": "pull_request:44025", "accept": false, - "reason": "UnivNet vs Depth Anything is not the same underlying change." + "reason": "`univnet` and `depth_anything` are unrelated code paths, so this is not a duplicate." }, { "left": "pull_request:44002", "right": "pull_request:44029", "accept": false, - "reason": "UperNet and RWKV are separate model implementations with different patches." + "reason": "`upernet` and `rwkv` are separate model families; similar output-tracing work is not enough to merge them." }, { "left": "pull_request:44002", "right": "pull_request:44003", "accept": false, - "reason": "UperNet and Mamba/Falcon Mamba are unrelated code paths." + "reason": "These touch unrelated model internals and do not fix the same concrete issue." }, { "left": "pull_request:44003", "right": "pull_request:44004", "accept": false, - "reason": "Both are output refactors, but CodeGen and Mamba/Falcon Mamba are different concrete model changes." + "reason": "Both are refactors, but `mamba/falcon_mamba` and `codegen` are distinct code paths with different output semantics." }, { "left": "pull_request:43998", "right": "pull_request:44004", "accept": false, - "reason": "timm_backbone and CodeGen are unrelated model implementations." + "reason": "`timm_backbone` and `codegen` are unrelated model implementations; same theme, different changes." }, { "left": "pull_request:44001", "right": "pull_request:44004", "accept": false, - "reason": "UnivNet and CodeGen do not share a single code path or bug." + "reason": "`univnet` and `codegen` are separate model families, so they are not the same underlying fix." }, { "left": "pull_request:44002", "right": "pull_request:44004", "accept": false, - "reason": "UperNet and CodeGen are separate changes in different models." + "reason": "`upernet` and `codegen` are unrelated forward paths and cannot plausibly be one merged PR." }, { "left": "pull_request:44000", "right": "pull_request:44004", "accept": false, - "reason": "Vision-Text Dual Encoder and CodeGen are unrelated beyond the generic refactor theme." + "reason": "`vision_text_dual_encoder` and `codegen` differ in both architecture and output handling; not a duplicate." }, { "left": "pull_request:44025", "right": "pull_request:44026", "accept": false, - "reason": "Depth Anything and Vision-Encoder-Decoder are distinct model-specific PRs." + "reason": "Different model families (`depth_anything` vs `vision_encoder_decoder`) and separate output-tracing changes." }, { "left": "pull_request:44025", "right": "pull_request:44027", "accept": false, - "reason": "Depth Anything and Speech Encoder-Decoder are separate implementations." + "reason": "`depth_anything` and `speech_encoder_decoder` are distinct code paths; these are parallel refactors, not one change." }, { "left": "pull_request:44025", "right": "pull_request:44028", "accept": false, - "reason": "Depth Anything and SuperPoint are unrelated models and patches." + "reason": "`depth_anything` and `superpoint` are unrelated model implementations with different output structures." }, { "left": "pull_request:44025", "right": "pull_request:44029", "accept": false, - "reason": "Depth Anything and RWKV are different code paths with no shared concrete bug." + "reason": "These target different architectures (`depth_anything` vs `rwkv`), so they are not the same underlying bug or patch." }, { "left": "pull_request:44025", "right": "pull_request:44030", "accept": false, - "reason": "Depth Anything and DPR are separate model refactors, not one mergeable change." + "reason": "`depth_anything` and `dpr` are separate models; similar refactoring does not make them duplicates." }, { "left": "pull_request:44026", "right": "pull_request:44029", "accept": false, - "reason": "Vision-Encoder-Decoder and RWKV are unrelated model families." + "reason": "`vision_encoder_decoder` and `rwkv` are different code paths with no shared fix." }, { "left": "pull_request:44026", "right": "pull_request:44027", "accept": false, - "reason": "Vision-Encoder-Decoder and Speech Encoder-Decoder are different model-specific changes." + "reason": "These update different encoder-decoder stacks and should be triaged separately." }, { "left": "pull_request:44026", "right": "pull_request:44028", "accept": false, - "reason": "Vision-Encoder-Decoder and SuperPoint do not share the same concrete fix." + "reason": "`vision_encoder_decoder` and `superpoint` are unrelated model implementations." }, { "left": "pull_request:44026", "right": "pull_request:44030", "accept": false, - "reason": "Different models and different output-handling code." + "reason": "Different subsystems (`vision_encoder_decoder` vs `dpr`); same pattern, not the same change." }, { "left": "pull_request:44027", "right": "pull_request:44028", "accept": false, - "reason": "Speech Encoder-Decoder and SuperPoint are unrelated model implementations." + "reason": "`speech_encoder_decoder` and `superpoint` do not share a concrete code path or bug." }, { "left": "pull_request:44027", "right": "pull_request:44029", "accept": false, - "reason": "Speech Encoder-Decoder and RWKV are distinct patches for different code paths." + "reason": "These are separate model families (`speech_encoder_decoder` vs `rwkv`) with independent tracing adjustments." }, { "left": "pull_request:44027", "right": "pull_request:44030", "accept": false, - "reason": "Speech Encoder-Decoder and DPR are not the same underlying change." + "reason": "`speech_encoder_decoder` and `dpr` are unrelated implementations, so not a duplicate pair." }, { "left": "pull_request:44028", "right": "pull_request:44029", "accept": false, - "reason": "SuperPoint and RWKV are different models; similarity is only thematic." + "reason": "`superpoint` and `rwkv` are different model paths; no single concrete fix connects them." }, { "left": "pull_request:44028", "right": "pull_request:44030", "accept": false, - "reason": "SuperPoint and DPR are separate refactors in separate models." + "reason": "These are distinct model implementations with different output contracts, so they should not be merged." }, { "left": "pull_request:44029", "right": "pull_request:44030", "accept": false, - "reason": "RWKV and DPR are unrelated concrete changes." + "reason": "`rwkv` and `dpr` are separate architectures and do not appear to fix the same underlying issue." } ] }, @@ -16994,7 +16915,7 @@ "pull_request:44662|pull_request:45075", "pull_request:45075|pull_request:45401" ], - "prepared_review_unit_hash": "a6804fbf456883f0f64206da648370eaf7865071a474d758f725c0c78aa0f1a8", + "prepared_review_unit_hash": "788c14159f465e1d89c72ab5c5ff680b9654cbe795fb988d0e53c93e3bb67f0b", "component_budget": { "node_count": 3, "item_count": 3, @@ -17016,30 +16937,30 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "All three PRs are separate model-implementation additions that happen to share the same Transformers boilerplate and auto-registration files; they are not duplicates of one another.", + "summary": "All three PRs are separate model-integration changes that only overlap in generic registry/docs wiring; they do not appear to be duplicates or the same underlying code-path fix.", "confidence": 0.97, "canonical_issue_reason": null, - "canonical_pr_reason": "No single canonical PR: PenguinVL, DeepSeek-OCR-2, and Voxtral TTS are different models with distinct codepaths and test suites.", + "canonical_pr_reason": "PR 45075 is the strongest standalone candidate: it is a complete, non-draft model addition with converter, model, image-processing, processing, and test coverage, whereas the others are distinct integrations.", "best_issue_reason": null, - "best_pr_reason": "None of these is a better global representative than the others; each is an independent model addition rather than the same fix/change.", + "best_pr_reason": "PR 45075 is the best overall representative because it is the most production-ready and self-contained of the three model additions.", "soft_edge_verdicts": [ { "left": "pull_request:44662", "right": "pull_request:45401", "accept": false, - "reason": "Different model integrations (PenguinVL vs Voxtral TTS). Shared auto/docs scaffolding is expected boilerplate, not the same underlying change." + "reason": "Different models and use cases (PenguinVL vision-language vs Voxtral TTS). Shared auto/TOC files are generic integration boilerplate, not evidence of the same change." }, { "left": "pull_request:44662", "right": "pull_request:45075", "accept": false, - "reason": "Different models and implementations. Similar registration/docs files do not indicate the same bug or mergeable combined PR." + "reason": "Different model integrations (PenguinVL vs DeepSeek-OCR-2). Overlap is limited to standard Transformers registry/docs files, so they are not the same underlying bug/change." }, { "left": "pull_request:45075", "right": "pull_request:45401", "accept": false, - "reason": "Distinct model support additions (DeepSeek-OCR-2 vs Voxtral TTS) with different codepaths; overlap is only template-level Transformers integration." + "reason": "DeepSeek-OCR-2 OCR support and Voxtral TTS support are unrelated modalities and code paths; the overlap is only in common model-registration files." } ] }, @@ -17069,7 +16990,7 @@ "soft_pairs": [ "pull_request:44979|pull_request:45363" ], - "prepared_review_unit_hash": "59c9ada948fc2080d23c66590cbcd737af0887a930b61588dd75e2eb19c4b85b", + "prepared_review_unit_hash": "dea6054d98e767949526b3bc34d9ea3286ebd7ae67c804cc3b2d92bd1ef09084", "component_budget": { "node_count": 2, "item_count": 2, @@ -17091,18 +17012,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both PRs are about fusion mechanisms in Transformers, but they address different layers: one introduces a generic module-fusion API, the other adds `KernelConfig`-driven n-to-1 kernel fusion and integration plumbing. They are related in theme, not the same concrete change.", - "confidence": 0.93, + "summary": "Both PRs target kernel/module fusion, but they are not the same change: #44979 introduces a low-level Module Fusion API, while #45363 implements n-to-1 kernel fusion through KernelConfig and integrates it into model loading/runtime. They look related and potentially dependent, not duplicate.", + "confidence": 0.91, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #44979 is the more foundational, self-contained fusion API change: it adds a new `module_fusion.py` abstraction and tests, making it the clearest representative of the cluster theme.", + "canonical_pr_reason": "#45363 is the better cluster representative because it is the more user-facing, end-to-end feature implementation and touches the main integration path for KernelConfig-based fusion.", "best_issue_reason": null, - "best_pr_reason": "PR #44979 is the best single representative because it defines the underlying fusion API directly, while #45363 is a separate, more implementation-specific kernel-fusion integration.", + "best_pr_reason": "#45363 has the clearest concrete scope and broader impact; #44979 is a foundational API piece rather than the full feature implementation.", "soft_edge_verdicts": [ { "left": "pull_request:44979", "right": "pull_request:45363", "accept": false, - "reason": "Related fusion topic, but not the same change: #44979 adds a generic module-fusion API, while #45363 wires `KernelConfig` into model loading for n-to-1 kernel fusion. They could not realistically be merged as one PR without conflating distinct features." + "reason": "Related fusion work, but different layers of the stack: one adds the Module Fusion API, the other wires n-to-1 kernel fusion through KernelConfig and integrations. Not the same underlying change and not something that would naturally merge into one PR." } ] }, @@ -17132,12 +17053,12 @@ "soft_pairs": [ "pull_request:45086|pull_request:45317" ], - "prepared_review_unit_hash": "915a547f788c2f1239d646fbde2d4334a86a30f1eb578bca81dadc748813a826", + "prepared_review_unit_hash": "13b67ffbb1eec714308edaae58d66fd0bfdf86eb975552fcd9cdbf66b846017b", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4413, + "serialized_chars": 4414, "estimated_input_tokens": 1104, "estimated_eval_tokens": 2464 }, @@ -17145,7 +17066,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4413, + "serialized_chars": 4414, "estimated_input_tokens": 1104, "estimated_eval_tokens": 2464 }, @@ -17154,18 +17075,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both PRs implement the same fix for the Mistral regex patch AttributeError in the same file; 45317 is a more complete version with tests, so these look like duplicate PRs for the same underlying bug fix.", - "confidence": 0.95, + "summary": "Both pull requests address the same Mistral regex AttributeError in `_patch_mistral_regex` by changing the same pre-tokenizer access path. PR 45317 is the more complete version because it includes tests in addition to the code fix, so it is the best canonical PR.", + "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 test coverage, making it the more complete representative of the change.", + "canonical_pr_reason": "PR 45317 is the better canonical PR because it contains the same core fix as 45086 plus test coverage, making it the more complete representation of the change.", "best_issue_reason": null, - "best_pr_reason": "PR 45317 is the strongest representative overall: same code-path fix, same target issue, and added tests.", + "best_pr_reason": "PR 45317 is the best overall PR candidate: same concrete bug fix as 45086, but with added tests and slightly broader implementation coverage.", "soft_edge_verdicts": [ { "left": "pull_request:45086", "right": "pull_request:45317", "accept": true, - "reason": "Same bug fix in the same function and file, with identical core code change; the second PR just adds test coverage, so they are the same underlying change." + "reason": "Same underlying bug and same code path: both replace `backend_tokenizer.pre_tokenizer` with `tokenizer.pre_tokenizer` in `_patch_mistral_regex`. The second PR appears to be a superset with tests, so these look mergeable into one change." } ] }, @@ -17195,7 +17116,7 @@ "soft_pairs": [ "pull_request:45115|pull_request:45569" ], - "prepared_review_unit_hash": "e248c62f136eeddad1d0d7c4e4d95143d3ceded5a43b78479c127910e652e9a5", + "prepared_review_unit_hash": "44422564ac9e4ff89f16e3bb3734c0b3ca8b2263e88db1dc8e6f7f6236f2895c", "component_budget": { "node_count": 2, "item_count": 2, @@ -17217,18 +17138,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These PRs are related to NemotronH, but they are not the same change: #45115 is a focused inheritance/refactor plus conversion-mapping work, while #45569 is a broader implementation split that introduces distinct Dense/Sparse model packages, docs, auto-mapping updates, and tests.", - "confidence": 0.88, + "summary": "These are related NemotronH-family refactors, but they target different concrete changes: one rewires NemotronH to inherit GraniteMoeHybrid and adds conversion mapping, while the other splits the family into NemotronH Dense/Sparse variants with docs, auto-mapping, and tests. They are not the same bug or change set.", + "confidence": 0.82, "canonical_issue_reason": null, - "canonical_pr_reason": "#45569 is the better canonical PR for this cluster because it is the broader, more complete NemotronH architecture change and includes the concrete model/package split, docs, and tests.", + "canonical_pr_reason": null, "best_issue_reason": null, - "best_pr_reason": "#45569 best represents the cluster because it covers the larger end-state change; #45115 is a narrower precursor/refactor centered on inheritance and conversion mappings.", + "best_pr_reason": "PR #45569 is the better cluster representative because it is broader and more self-contained around the NemotronH family split, covering model registration, docs, and tests; #45115 is a narrower internal refactor/conversion-mapping change.", "soft_edge_verdicts": [ { "left": "pull_request:45115", "right": "pull_request:45569", "accept": false, - "reason": "Reject: they share a subsystem and some files, but the concrete change is different. #45115 focuses on refactoring NemotronH to inherit GraniteMoeHybrid-compatible structure and mapping keys, while #45569 introduces separate NemotronHDense/NemotronHSparse models, docs, auto-mapping, and tests. They would not plausibly merge as one single PR without collapsing distinct workstreams." + "reason": "Overlap is only at the shared NemotronH base model files. #45115 focuses on GraniteMoeHybrid inheritance and checkpoint conversion renames, while #45569 introduces separate Dense/Sparse model variants and documentation changes. Different concrete changes, so not a duplicate pair." } ] }, @@ -17258,7 +17179,7 @@ "soft_pairs": [ "pull_request:45332|pull_request:45333" ], - "prepared_review_unit_hash": "59449caedd769c3a356725ea8f1d27c3894895ae06e3ea3dc58835882c44a8d5", + "prepared_review_unit_hash": "82d7423e3ef463a26b389d7162745ae0f1b4e4b8fcf6a2efcdba87d1a4003768", "component_budget": { "node_count": 2, "item_count": 2, @@ -17280,18 +17201,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Two related feature PRs in the heterogeneous-config/modeling area, but they are not the same concrete change: one adds per-layer config plumbing, the other adds broader heterogeneous runtime/model support. They overlap in shared helper/config files but target different layers of the feature stack.", - "confidence": 0.86, + "summary": "The two PRs are related but not duplicates: #45333 adds heterogeneous per-layer configuration support, while #45332 builds on that with broader heterogeneous modeling/cache integration. They share utility files, but they target different layers of the feature stack.", + "confidence": 0.92, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #45332 is the broader representative: it covers heterogeneous modeling/runtime support in addition to the config plumbing, so it best captures the combined feature direction of the cluster.", + "canonical_pr_reason": "PR #45332 is the better representative because it covers the larger end-to-end heterogeneous model support path, including cache/modeling changes, while #45333 is mainly the configuration foundation.", "best_issue_reason": null, - "best_pr_reason": "PR #45332 is the most complete and central artifact for the heterogeneous feature work; PR #45333 is narrower and mainly foundational config support.", + "best_pr_reason": "#45332 is the strongest overall PR in the cluster since it subsumes config support and extends into concrete model/runtime behavior, making it the most complete implementation candidate.", "soft_edge_verdicts": [ { "left": "pull_request:45332", "right": "pull_request:45333", "accept": false, - "reason": "Related, but not duplicate: #45333 adds per-layer configuration support, while #45332 adds heterogeneous model/runtime behavior that consumes such config. They touch overlapping helpers but do not fix the same concrete code-path problem and would not be merged as one PR without losing separation of concerns." + "reason": "Related but not the same change: #45333 is config plumbing for per-layer config, whereas #45332 adds heterogeneous model/runtime support and cache handling on top. They share subsystem files but would not plausibly be merged as one duplicate PR." } ] }, @@ -17321,7 +17242,7 @@ "soft_pairs": [ "pull_request:45415|pull_request:45425" ], - "prepared_review_unit_hash": "f3ca094bb5f7a2dbe8b71b86b15f4868b20cad39a174e6c2dacdc9621cc8e94d", + "prepared_review_unit_hash": "5a60dc715725ed1187461e6d2d4b7d532f26cc41f9d956e0e924e2e7301d71c4", "component_budget": { "node_count": 2, "item_count": 2, @@ -17343,18 +17264,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both PRs are typing-related, but they are not the same change: #45415 is a broad repo-wide type-checking sweep, while #45425 adds specific typing protocols/aliases and updates a few targeted files. I would not merge them as duplicates.", - "confidence": 0.91, + "summary": "Two unrelated typing-focused PRs: one is a broad repo-wide type-checking sweep, the other adds specific typing helpers and annotations for modeling_utils/PEFT integration. They overlap on a couple of shared files but do not fix the same concrete bug/change.", + "confidence": 0.94, "canonical_issue_reason": null, - "canonical_pr_reason": "#45425 is the better canonical PR for this cluster because it is a focused typing-infrastructure change centered on _typing.py and model-related annotations, whereas #45415 is a much broader mechanical type-checking pass across many files.", + "canonical_pr_reason": "45425 is the more focused, concrete typing change centered on modeling_utils and PEFT-related protocols; 45415 is a broad type-checking cleanup across many unrelated files and should not be treated as the same change.", "best_issue_reason": null, - "best_pr_reason": "#45425 is the best representative PR here: it is more self-contained and easier to identify as a concrete typing fix/change. #45415 is too broad and looks like a general sweep rather than the same underlying patch.", + "best_pr_reason": "45425 is the best representative of the cluster because it has a narrowly scoped, coherent purpose and a single code-path theme, unlike the large cross-file typing sweep in 45415.", "soft_edge_verdicts": [ { "left": "pull_request:45415", "right": "pull_request:45425", "accept": false, - "reason": "Reject: both are about typing, but they address different scopes and concrete changes. #45415 is a wide package-level type-checking pass, while #45425 adds specific typing aliases/protocols for modeling_utils/peft-related code. They do not look like the same underlying fix and would not plausibly be merged as one PR." + "reason": "Both are typing-related, but 45415 is a broad static-type cleanup across many modules while 45425 adds specific protocols and annotations for modeling_utils/PEFT. Shared filenames alone do not indicate the same underlying fix, so they should not be merged as duplicates." } ] }, diff --git a/analysis/current/manifest.json b/analysis/current/manifest.json index 70436907c0130d870fa77d11f3b76d31dd73b395..4f8cc3675c71114c39c5c639ffd5b62b0028e807 100644 --- a/analysis/current/manifest.json +++ b/analysis/current/manifest.json @@ -1,8 +1,8 @@ { - "analysis_id": "hybrid-model-20260422t180025z", + "analysis_id": "hybrid-model-20260423t000027z", "archived_artifacts": { - "hybrid": "snapshots/20260422T180025Z/analysis-runs/hybrid-model-20260422t180025z/analysis-report-hybrid.json", - "hybrid_reviews": "snapshots/20260422T180025Z/analysis-runs/hybrid-model-20260422t180025z/analysis-report-hybrid.llm-reviews.json" + "hybrid": "snapshots/20260423T000027Z/analysis-runs/hybrid-model-20260423t000027z/analysis-report-hybrid.json", + "hybrid_reviews": "snapshots/20260423T000027Z/analysis-runs/hybrid-model-20260423t000027z/analysis-report-hybrid.llm-reviews.json" }, "artifacts": { "hybrid": "analysis/current/analysis-report-hybrid.json", @@ -10,9 +10,9 @@ }, "channel": "canonical", "model": null, - "published_at": "2026-04-22T18:09:33Z", + "published_at": "2026-04-23T00:09:12Z", "repo": "huggingface/transformers", "schema_version": 1, - "snapshot_id": "20260422T180025Z", + "snapshot_id": "20260423T000027Z", "variant": "hybrid" } diff --git a/snapshots/20260423T000027Z/analysis-runs/hybrid-model-20260423t000027z/analysis-report-hybrid.json b/snapshots/20260423T000027Z/analysis-runs/hybrid-model-20260423t000027z/analysis-report-hybrid.json new file mode 100644 index 0000000000000000000000000000000000000000..b4c1e45a67baf68f41b072bb914b0157fa308e4e --- /dev/null +++ b/snapshots/20260423T000027Z/analysis-runs/hybrid-model-20260423t000027z/analysis-report-hybrid.json @@ -0,0 +1,1750 @@ +{ + "schema_version": "1.0", + "repo": "huggingface/transformers", + "snapshot_id": "20260423T000027Z", + "generated_at": "2026-04-23T00:09:10Z", + "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": 0.156, + "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-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-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-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." + } + ], + "best_issue": { + "cluster_id": "cluster-41211-3", + "issue_number": 41211, + "reason": "Issue #41211 is the strongest global issue candidate because it is open, belongs to a cluster with 3 artifacts, and carries 6 discussion comments plus 6 inbound references.", + "score": 148.0 + }, + "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/20260423T000027Z/analysis-runs/hybrid-model-20260423t000027z/analysis-report-hybrid.llm-reviews.json b/snapshots/20260423T000027Z/analysis-runs/hybrid-model-20260423t000027z/analysis-report-hybrid.llm-reviews.json new file mode 100644 index 0000000000000000000000000000000000000000..6166aae5918ee6694fe9ba0059022384e01d40b6 --- /dev/null +++ b/snapshots/20260423T000027Z/analysis-runs/hybrid-model-20260423t000027z/analysis-report-hybrid.llm-reviews.json @@ -0,0 +1,17290 @@ +{ + "schema_version": "1.0", + "repo": "huggingface/transformers", + "snapshot_id": "20260423T000027Z", + "generated_at": "2026-04-23T00:09:10Z", + "model": "gpt-5.4-mini?service_tier=flex", + "reviews": [ + { + "label": "issue", + "component_index": 1, + "component_count": 13, + "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": "66bfc44967010832d2937d318ab9877350859211b88141bb644aeaced5a638a0", + "component_budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 2476, + "estimated_input_tokens": 619, + "estimated_eval_tokens": 1494 + }, + "budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 2476, + "estimated_input_tokens": 619, + "estimated_eval_tokens": 1494 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These three issues are only loosely related by TensorFlow/Transformers usage, but they describe different bugs: tokenization mismatch, an authentication/error-message problem, and a model-loading failure. I would not merge them as duplicates.", + "confidence": 0.97, + "canonical_issue_reason": "Issue #16618 is the closest cluster anchor because it is the only one with soft similarity links to both other issues and it concerns a concrete model-loading failure, but it is still not a true duplicate of the others.", + "canonical_pr_reason": null, + "best_issue_reason": "#16618 is the best representative of this cluster\u2019s narrow overlap because it is the most central node in the similarity graph and has the clearest, self-contained bug report.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:13244", + "right": "issue:16618", + "accept": false, + "reason": "Different underlying problems: Tapas tokenization parity vs TensorFlow code is unrelated to a BERT model-loading failure." + }, + { + "left": "issue:15980", + "right": "issue:16618", + "accept": false, + "reason": "One is about a bad login/error message for private model downloads; the other is about failing to load bert-base-uncased. Same subsystem, but 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": 2, + "component_count": 13, + "review_unit_index": 1, + "review_unit_count": 42, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-367", + "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:43296", + "issue:43366", + "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" + ], + "soft_pairs": [ + "issue:44322|issue:45464", + "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:43975|issue:45237", + "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:43541|issue:43716", + "issue:43828|issue:44292", + "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:43716|issue:44918", + "issue:43039|issue:43828", + "issue:30277|issue:36331", + "issue:43039|issue:44805", + "issue:34928|issue:36010", + "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:43425|issue:45070", + "issue:39290|issue:43531", + "issue:41093|issue:43638", + "issue:43404|issue:43866", + "issue:45362|issue:45464", + "issue:43792|issue:45070", + "issue:43541|issue:43828", + "issue:36010|issue:39290", + "issue:44841|issue:45084", + "issue:41093|issue:41762", + "issue:43296|issue:43366" + ], + "prepared_review_unit_hash": "15fedc60ae7a6503aa0f525b091f17a2e02913e7b665b8cd4e7fbf2325a10fa2", + "component_budget": { + "node_count": 367, + "item_count": 367, + "soft_pair_count": 1370, + "serialized_chars": 706986, + "estimated_input_tokens": 176747, + "estimated_eval_tokens": 353750 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 63, + "serialized_chars": 46483, + "estimated_input_tokens": 11621, + "estimated_eval_tokens": 23498 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Mostly a mixed cluster of unrelated model-support, dtype/config, docs, and DeepSpeed/FSDP issues. Only one soft duplicate pair looks high-confidence: the two DeepSpeed ZeRO-3 BERT out-of-bounds reports.", + "confidence": 0.74, + "canonical_issue_reason": "issue:43596 is the cleanest anchor for the accepted duplicate pair: it states the DeepSpeed ZeRO-3 + BERT index-0 out-of-bounds failure clearly and concisely.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43596 is the best global issue candidate because it most clearly represents the only strong duplicate pair in this set; the other items are mostly separate bugs or docs problems.", + "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/completions error; not the same code path." + }, + { + "left": "issue:43716", + "right": "issue:45237", + "accept": false, + "reason": "Mistral-3 image-preprocessor dtype mismatch is unrelated to GPT-OSS AMD GPU failure." + }, + { + "left": "issue:43828", + "right": "issue:45237", + "accept": false, + "reason": "Different models and different failure modes; both are dtype/runtime issues only at a very broad level." + }, + { + "left": "issue:45237", + "right": "issue:45507", + "accept": false, + "reason": "GPT-OSS on AMD GPUs vs GraniteMoEHybrid invalid-method error are unrelated." + }, + { + "left": "issue:44292", + "right": "issue:45237", + "accept": false, + "reason": "Qwen-3-8B-NVFP4 runtime error is a different model/path from GPT-OSS AMD GPU support." + }, + { + "left": "issue:30277", + "right": "issue:35545", + "accept": false, + "reason": "DeepSpeed collective mismatch and ModernBERT ONNX export are unrelated bugs." + }, + { + "left": "issue:35545", + "right": "issue:42915", + "accept": false, + "reason": "ModernBERT ONNX export vs Qwen3Moe FP8 failure are different models and subsystems." + }, + { + "left": "issue:35545", + "right": "issue:41720", + "accept": false, + "reason": "ONNX export issue is unrelated to Qwen3 auto device-mapping cuda assert." + }, + { + "left": "issue:44509", + "right": "issue:44512", + "accept": false, + "reason": "Both are docs cleanup, but they reference different removed features and different pages; too broad to treat as the same bug." + }, + { + "left": "issue:35545", + "right": "issue:36010", + "accept": false, + "reason": "ONNX export bug and GenerationMixin import regression are unrelated." + }, + { + "left": "issue:43638", + "right": "issue:44805", + "accept": false, + "reason": "Both mention indexing errors, but the model/setup and error shape differ enough that the underlying bug is not clearly the same." + }, + { + "left": "issue:43596", + "right": "issue:44805", + "accept": false, + "reason": "Zero-3+BERT index-out-of-bounds vs mask/tensor shape mismatch are different concrete failures." + }, + { + "left": "issue:44805", + "right": "issue:45161", + "accept": false, + "reason": "Shape-mismatch indexing error and GPT-OSS tensor-parallel issue are unrelated." + }, + { + "left": "issue:44805", + "right": "issue:44918", + "accept": false, + "reason": "Different failure signatures and code paths; not the same bug." + }, + { + "left": "issue:44661", + "right": "issue:44805", + "accept": false, + "reason": "Tokenizer-mapping add-new-model-like failure is unrelated to the mask/tensor shape mismatch." + }, + { + "left": "issue:34634", + "right": "issue:35545", + "accept": false, + "reason": "BarkProcessor voice_preset issue is unrelated to ModernBERT ONNX export." + }, + { + "left": "issue:30277", + "right": "issue:43638", + "accept": false, + "reason": "Both involve DeepSpeed ZeRO-3, but the collective mismatch and BERT index error are different bugs." + }, + { + "left": "issue:43596", + "right": "issue:43638", + "accept": true, + "reason": "Near-identical DeepSpeed ZeRO-3 BERT out-of-bounds reports with the same error pattern and same underlying code path." + }, + { + "left": "issue:43975", + "right": "issue:44457", + "accept": false, + "reason": "Detokenization regression and LoRA merge/reload inconsistency are separate problems." + }, + { + "left": "issue:43039", + "right": "issue:45161", + "accept": false, + "reason": "Liger Kernel cross-entropy behavior is unrelated to GPT-OSS tensor-parallel failure." + }, + { + "left": "issue:44292", + "right": "issue:44457", + "accept": false, + "reason": "Qwen NVFP4 runtime error and LoRA merge/save mismatch are unrelated." + }, + { + "left": "issue:34928", + "right": "issue:44805", + "accept": false, + "reason": "FSDP checkpointing recompute mismatch is not the same as the mask/tensor indexing error." + }, + { + "left": "issue:24643", + "right": "issue:30277", + "accept": false, + "reason": "Both mention DeepSpeed, but the weight-dimension error and collective mismatch are distinct failures." + }, + { + "left": "issue:43975", + "right": "issue:45237", + "accept": false, + "reason": "Detokenization bug vs GPU backend support issue; unrelated." + }, + { + "left": "issue:43638", + "right": "issue:45161", + "accept": false, + "reason": "BERT ZeRO-3 indexing bug and GPT-OSS tensor-parallel bug are different issues." + }, + { + "left": "issue:34928", + "right": "issue:36331", + "accept": false, + "reason": "Activation-checkpointing/FSDP issue is unrelated to CustomTrainer compute_loss signature regression." + }, + { + "left": "issue:43596", + "right": "issue:45161", + "accept": false, + "reason": "Different models and different failure classes." + }, + { + "left": "issue:43716", + "right": "issue:44292", + "accept": false, + "reason": "Mistral-3 dtype mismatch and Qwen NVFP4 runtime error do not share the same bug." + }, + { + "left": "issue:39290", + "right": "issue:45081", + "accept": false, + "reason": "Gemma3 vLLM config attribute error and Mistral tokenizer regex patch crash are unrelated." + }, + { + "left": "issue:43828", + "right": "issue:45161", + "accept": false, + "reason": "Autocast dtype mismatch and GPT-OSS TP failure are not the same underlying problem." + }, + { + "left": "issue:30277", + "right": "issue:36331", + "accept": false, + "reason": "DeepSpeed collective mismatch and trainer method signature change are unrelated." + }, + { + "left": "issue:43039", + "right": "issue:44805", + "accept": false, + "reason": "Liger Kernel cross-entropy path is unrelated to the indexing shape mismatch." + }, + { + "left": "issue:34928", + "right": "issue:36010", + "accept": false, + "reason": "FSDP activation checkpointing bug and GenerationMixin import error are unrelated." + }, + { + "left": "issue:41762", + "right": "issue:43596", + "accept": false, + "reason": "Both are ZeRO-3 index errors, but they involve different model families and enough context difference to avoid merging." + }, + { + "left": "issue:43901", + "right": "issue:44512", + "accept": false, + "reason": "Both are docs regressions, but they concern different removed APIs and different documentation targets." + }, + { + "left": "issue:39290", + "right": "issue:41720", + "accept": false, + "reason": "Gemma3 vLLM config regression and Qwen3 auto-device CUDA assert are unrelated." + }, + { + "left": "issue:43039", + "right": "issue:44560", + "accept": false, + "reason": "Liger Kernel cross-entropy issue and Qwen3-vl video StopIteration are unrelated." + }, + { + "left": "issue:43866", + "right": "issue:44863", + "accept": false, + "reason": "Corrupted checkpoint report and NemotronH loading failure are not the same bug." + }, + { + "left": "issue:43854", + "right": "issue:43866", + "accept": false, + "reason": "Unit-test loading failure vs corrupted checkpoint; too different to duplicate." + }, + { + "left": "issue:43425", + "right": "issue:45070", + "accept": false, + "reason": "Torch version incompatibility and pydantic PretrainedConfig field breakage are unrelated." + }, + { + "left": "issue:39290", + "right": "issue:43531", + "accept": false, + "reason": "Different Qwen/Gemma sliding-window issues and different configs; not clearly the same bug." + }, + { + "left": "issue:41093", + "right": "issue:43638", + "accept": false, + "reason": "Both are indexing/shape errors, but they occur in different model/setup contexts and are not clearly duplicates." + }, + { + "left": "issue:43404", + "right": "issue:43866", + "accept": false, + "reason": "Mistral3 weight-tie bug and Ovis2 corrupted checkpoint are unrelated." + }, + { + "left": "issue:45362", + "right": "issue:45464", + "accept": false, + "reason": "Qwen3.5 chat crash and streaming API failure are different symptoms with different likely causes." + }, + { + "left": "issue:43792", + "right": "issue:45070", + "accept": false, + "reason": "Whisper load failure and pydantic model regression are unrelated." + }, + { + "left": "issue:43541", + "right": "issue:43828", + "accept": false, + "reason": "Grouped_mm tracing error on Mixtral and autocast dtype mismatch on Phi-tiny-MoE are different code paths." + }, + { + "left": "issue:36010", + "right": "issue:39290", + "accept": false, + "reason": "GenerationMixin import regression and Gemma3 sliding-window config error are unrelated." + }, + { + "left": "issue:44841", + "right": "issue:45084", + "accept": false, + "reason": "Voxtral processor failure and template-node compile error are unrelated." + }, + { + "left": "issue:41093", + "right": "issue:41762", + "accept": false, + "reason": "Both are mask/index errors, but they are on different models and not clearly the same root cause." + }, + { + "left": "issue:43296", + "right": "issue:43366", + "accept": false, + "reason": "PaddleOCR-VL loading failure and GGUF gpt-oss support request are unrelated." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet and stays conservative overall. The only accepted soft pair, issue:43596 vs issue:43638, is supported by very similar DeepSpeed ZeRO-3 + index-out-of-bounds reports. The remaining rejections are generally cautious and do not overclaim duplication from broad similarity alone." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 13, + "review_unit_index": 2, + "review_unit_count": 42, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-367", + "nodes": [ + "issue:36331", + "issue:39186", + "issue:39290", + "issue:39692", + "issue:40990", + "issue:41093", + "issue:41553", + "issue:41762", + "issue:42915", + "issue:43054", + "issue:43366", + "issue:43404", + "issue:43531", + "issue:43541", + "issue:43572", + "issue:43596", + "issue:43644", + "issue:43645", + "issue:43742", + "issue:43828", + "issue:43866", + "issue:43873", + "issue:43883", + "issue:43950", + "issue:43975", + "issue:43994", + "issue:44291", + "issue:44292", + "issue:44360", + "issue:44387", + "issue:44403", + "issue:44492", + "issue:44512", + "issue:44530", + "issue:44534", + "issue:44560", + "issue:44589", + "issue:44661", + "issue:44863", + "issue:44918", + "issue:44928", + "issue:44960", + "issue:45005", + "issue:45070", + "issue:45161", + "issue:45399", + "issue:45464", + "issue:45507" + ], + "soft_pairs": [ + "issue:43572|issue:45070", + "issue:44387|issue:45005", + "issue:43541|issue:45161", + "issue:43828|issue:45464", + "issue:44291|issue:45070", + "issue:43366|issue:45464", + "issue:43531|issue:45070", + "issue:44661|issue:45464", + "issue:39290|issue:43742", + "issue:43054|issue:43994", + "issue:43541|issue:44589", + "issue:41093|issue:43596", + "issue:42915|issue:43541", + "issue:40990|issue:44960", + "issue:39186|issue:44918", + "issue:43404|issue:45005", + "issue:43645|issue:45399", + "issue:44530|issue:45005", + "issue:44863|issue:45005", + "issue:43873|issue:45005", + "issue:44360|issue:44512", + "issue:42915|issue:45399", + "issue:44918|issue:45507", + "issue:44492|issue:44512", + "issue:43975|issue:44292", + "issue:43950|issue:44534", + "issue:43644|issue:43950", + "issue:44292|issue:45507", + "issue:43828|issue:44928", + "issue:39692|issue:43054", + "issue:41553|issue:43883", + "issue:43866|issue:44403", + "issue:36331|issue:41762", + "issue:44560|issue:45507" + ], + "prepared_review_unit_hash": "8ef8a5bf83a138b19ad11e4fbf89972cf8ba6404e8221ff3813f09ee968e3c06", + "component_budget": { + "node_count": 367, + "item_count": 367, + "soft_pair_count": 1370, + "serialized_chars": 706986, + "estimated_input_tokens": 176747, + "estimated_eval_tokens": 353750 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 34, + "serialized_chars": 37565, + "estimated_input_tokens": 9392, + "estimated_eval_tokens": 19040 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is mostly heterogeneous, but it contains two clear duplicate subclusters: non-persistent buffer corruption in Transformers v5, and a SigLIP2 output-quality regression. Most other soft edges are only broad topical similarity and should be rejected.", + "confidence": 0.82, + "canonical_issue_reason": "issue:43950 is the clearest representative of the non-persistent-buffer regression and has direct duplicates in the set (issue:43644, issue:44534).", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43950 is the most actionable and well-scoped issue here: concrete regression, clear symptom, and it anchors the strongest duplicate cluster.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43572", + "right": "issue:45070", + "accept": false, + "reason": "Different bugs: one is a missing config field, the other is a pydantic/PretrainedConfig regression." + }, + { + "left": "issue:44387", + "right": "issue:45005", + "accept": false, + "reason": "Both mention v5 regressions, but the symptoms and code paths are unrelated (int4 OOM vs tied weights)." + }, + { + "left": "issue:43541", + "right": "issue:45161", + "accept": false, + "reason": "Different MoE failures: torch dynamo grouped_mm tracing error vs tensor-parallel support issue." + }, + { + "left": "issue:43828", + "right": "issue:45464", + "accept": false, + "reason": "Unrelated runtime failures: autocast dtype mismatch vs streaming inference API failure." + }, + { + "left": "issue:44291", + "right": "issue:45070", + "accept": false, + "reason": "One is init_empty_weights/_is_hf_initialized, the other is a pydantic model field breakage." + }, + { + "left": "issue:43366", + "right": "issue:45464", + "accept": false, + "reason": "Different domains: GGUF architecture support vs Qwen3.5 streaming inference." + }, + { + "left": "issue:43531", + "right": "issue:45070", + "accept": false, + "reason": "Qwen3-MoE sliding_window behavior is unrelated to PretrainedConfig/pydantic issues." + }, + { + "left": "issue:44661", + "right": "issue:45464", + "accept": false, + "reason": "Tokenizer-mapping model addition bug is unrelated to streaming inference." + }, + { + "left": "issue:39290", + "right": "issue:43742", + "accept": false, + "reason": "Different model-specific loading errors with no clear shared code path." + }, + { + "left": "issue:43054", + "right": "issue:43994", + "accept": true, + "reason": "Both describe SigLIP2 producing bad/nonsensical outputs, likely the same model regression." + }, + { + "left": "issue:43541", + "right": "issue:44589", + "accept": false, + "reason": "Different failures: grouped_mm tracing vs missing Float8 storage type." + }, + { + "left": "issue:41093", + "right": "issue:43596", + "accept": false, + "reason": "Both are shape/index errors, but the model/setup and likely root cause differ." + }, + { + "left": "issue:42915", + "right": "issue:43541", + "accept": false, + "reason": "Different MoE-specific problems and different failure modes." + }, + { + "left": "issue:40990", + "right": "issue:44960", + "accept": false, + "reason": "No meaningful overlap beyond being model-quality/perplexity complaints." + }, + { + "left": "issue:39186", + "right": "issue:44918", + "accept": false, + "reason": "Different training/runtime failures: FSDP weight-shape error vs TRL unpacking embeddings." + }, + { + "left": "issue:43404", + "right": "issue:45005", + "accept": false, + "reason": "Related theme of tied weights, but the model scope and failure mode are not clearly the same bug." + }, + { + "left": "issue:43645", + "right": "issue:45399", + "accept": false, + "reason": "Notebook/custom model init issue is unrelated to flash-attn fallback checks." + }, + { + "left": "issue:44530", + "right": "issue:45005", + "accept": false, + "reason": "PagedAttentionCache/group type crash is unrelated to tied-weight regressions." + }, + { + "left": "issue:44863", + "right": "issue:45005", + "accept": false, + "reason": "NemotronH checkpoint loading is a different model-loading problem." + }, + { + "left": "issue:43873", + "right": "issue:45005", + "accept": false, + "reason": "Quantization/offloading behavior is unrelated to tied weights." + }, + { + "left": "issue:44360", + "right": "issue:44512", + "accept": false, + "reason": "Completely different issues: DSA indexer internals vs outdated docs." + }, + { + "left": "issue:42915", + "right": "issue:45399", + "accept": false, + "reason": "No shared concrete bug; MoE FP8 failure vs flash-attn fallback gating." + }, + { + "left": "issue:44918", + "right": "issue:45507", + "accept": false, + "reason": "Different model families and unrelated runtime errors." + }, + { + "left": "issue:44492", + "right": "issue:44512", + "accept": false, + "reason": "A typo in cache strategies is not the same as stale CLI docs." + }, + { + "left": "issue:43975", + "right": "issue:44292", + "accept": false, + "reason": "Different inference bugs with unrelated symptoms and models." + }, + { + "left": "issue:43950", + "right": "issue:44534", + "accept": true, + "reason": "Both report the same non-persistent-buffer corruption regression in Transformers v5." + }, + { + "left": "issue:43644", + "right": "issue:43950", + "accept": true, + "reason": "Same bug: v5 corrupts non-persistent buffers during load/init." + }, + { + "left": "issue:44292", + "right": "issue:45507", + "accept": false, + "reason": "Qwen NVFP4 runtime error and GraniteMoE invalid method are unrelated." + }, + { + "left": "issue:43828", + "right": "issue:44928", + "accept": false, + "reason": "Both involve dtype/numerics, but the actual failures and code paths differ." + }, + { + "left": "issue:39692", + "right": "issue:43054", + "accept": false, + "reason": "Both mention SigLIP2, but docs example errors and embedding-quality regression are different problems." + }, + { + "left": "issue:41553", + "right": "issue:43883", + "accept": false, + "reason": "Bad tokenizer error message vs missing all_tied_weights_keys are unrelated." + }, + { + "left": "issue:43866", + "right": "issue:44403", + "accept": false, + "reason": "Checkpoint corruption and loading noise are not the same bug." + }, + { + "left": "issue:36331", + "right": "issue:41762", + "accept": false, + "reason": "Different training/loading failures with no shared root cause." + }, + { + "left": "issue:44560", + "right": "issue:45507", + "accept": false, + "reason": "Qwen3-vl video StopIteration and GraniteMoE invalid method are unrelated." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "Grounded in the packet and appropriately conservative. The two supported duplicate clusters are the non-persistent-buffer regression (43950/43644/44534) and the SigLIP2 bad-output pair (43054/43994). The other soft-edge pairs are rejected with cautious, issue-title-based reasoning rather than overclaiming evidence." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 13, + "review_unit_index": 3, + "review_unit_count": 42, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-367", + "nodes": [ + "issue:36331", + "issue:38175", + "issue:41093", + "issue:41553", + "issue:41762", + "issue:42491", + "issue:42915", + "issue:42947", + "issue:43054", + "issue:43257", + "issue:43278", + "issue:43404", + "issue:43493", + "issue:43643", + "issue:43742", + "issue:43782", + "issue:43792", + "issue:43824", + "issue:43827", + "issue:43854", + "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:44863", + "issue:44960", + "issue:44991", + "issue:45020", + "issue:45081", + "issue:45305", + "issue:45313", + "issue:45356", + "issue:45399", + "issue:45520" + ], + "soft_pairs": [ + "issue:43872|issue:44589", + "issue:41553|issue:43881", + "issue:38175|issue:43054", + "issue:42947|issue:44387", + "issue:43054|issue:43493", + "issue:43881|issue:43883", + "issue:44403|issue:44451", + "issue:42947|issue:43856", + "issue:43975|issue:45520", + "issue:43643|issue:45020", + "issue:44403|issue:44488", + "issue:43782|issue:44960", + "issue:42491|issue:44368", + "issue:41762|issue:42915", + "issue:43975|issue:44451", + "issue:45020|issue:45356", + "issue:44661|issue:45020", + "issue:43054|issue:43278", + "issue:43824|issue:43872", + "issue:44991|issue:45399", + "issue:43827|issue:44509", + "issue:43792|issue:44220", + "issue:45081|issue:45399", + "issue:43404|issue:43950", + "issue:43257|issue:43866", + "issue:42915|issue:45020", + "issue:43742|issue:45399", + "issue:43742|issue:45020", + "issue:41553|issue:44991", + "issue:43940|issue:45313", + "issue:43872|issue:44291", + "issue:43854|issue:44863", + "issue:43257|issue:45305", + "issue:41553|issue:44488", + "issue:42915|issue:43872", + "issue:36331|issue:41093" + ], + "prepared_review_unit_hash": "7204b8a2d4879c591ed731bf4a7cc7098cdfe1756b9cc484c1c7a3834e3f68bd", + "component_budget": { + "node_count": 367, + "item_count": 367, + "soft_pair_count": 1370, + "serialized_chars": 706986, + "estimated_input_tokens": 176747, + "estimated_eval_tokens": 353750 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 36, + "serialized_chars": 38577, + "estimated_input_tokens": 9645, + "estimated_eval_tokens": 19546 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The set is mostly unrelated issues, but there are two clear duplicate pairs: the transformers 5.x `_is_hf_initialized` loading TypeError, and the docs regression about removed pipeline tasks still being referenced. Everything else looks too broad or symptom-level to merge as duplicates.", + "confidence": 0.78, + "canonical_issue_reason": "issue:44291 is the clearest root-cause report for the `_is_hf_initialized` TypeError during `init_empty_weights` loading, and it covers the broader failure mode better than the narrower bitsandbytes-specific report in issue:43872.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:44291 is the strongest representative issue overall: it has the most explicit technical cause, a clear reproduction path, and spans the general loading regression rather than one library-specific manifestation.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43872", + "right": "issue:44589", + "accept": false, + "reason": "Different failures: bitsandbytes `_is_hf_initialized` construction error vs a float8 storage lookup error." + }, + { + "left": "issue:41553", + "right": "issue:43881", + "accept": false, + "reason": "Different models and different failure modes; only loosely related around model loading." + }, + { + "left": "issue:38175", + "right": "issue:43054", + "accept": false, + "reason": "Both mention SigLIP2, but the reported symptoms are different enough that they do not look like the same bug." + }, + { + "left": "issue:42947", + "right": "issue:44387", + "accept": false, + "reason": "Unrelated regressions: LoRA gradient checkpointing vs int4 quantization memory growth." + }, + { + "left": "issue:43054", + "right": "issue:43493", + "accept": false, + "reason": "Related to SigLIP2 quality/discrepancy, but not enough evidence they are the same concrete bug." + }, + { + "left": "issue:43881", + "right": "issue:43883", + "accept": false, + "reason": "Different architectures and different errors; no shared code-path is evident." + }, + { + "left": "issue:44403", + "right": "issue:44451", + "accept": false, + "reason": "Generic loading noise vs a specific inability to load a particular model." + }, + { + "left": "issue:42947", + "right": "issue:43856", + "accept": false, + "reason": "Different training concerns: ineffective checkpointing vs MoE memory usage." + }, + { + "left": "issue:43975", + "right": "issue:45520", + "accept": false, + "reason": "Tokenizer detokenization bug and a Python 3.13 flash-attn import key error are unrelated." + }, + { + "left": "issue:43643", + "right": "issue:45020", + "accept": false, + "reason": "Both involve remote-code loading, but one is a specific AutoConfig field-loss bug and the other is a broader regression claim." + }, + { + "left": "issue:44403", + "right": "issue:44488", + "accept": false, + "reason": "Same broad area of model loading, but different models and no concrete shared bug." + }, + { + "left": "issue:43782", + "right": "issue:44960", + "accept": false, + "reason": "Different models and different load-time failures; no direct duplicate evidence." + }, + { + "left": "issue:42491", + "right": "issue:44368", + "accept": false, + "reason": "LoRA compatibility across versions vs a tie_word_embeddings warning; not the same issue." + }, + { + "left": "issue:41762", + "right": "issue:42915", + "accept": false, + "reason": "Different model families and different DeepSpeed/FP8 failure modes." + }, + { + "left": "issue:43975", + "right": "issue:44451", + "accept": false, + "reason": "Tokenizer detokenization regression vs model loading failure; not the same bug." + }, + { + "left": "issue:45020", + "right": "issue:45356", + "accept": false, + "reason": "Both involve recent regressions, but one is general remote-code breakage and the other is a tokenizer codec/fix_mistral_regex issue." + }, + { + "left": "issue:44661", + "right": "issue:45020", + "accept": false, + "reason": "Different problems: tokenizer mapping guardrail vs remote-code loading regressions." + }, + { + "left": "issue:43054", + "right": "issue:43278", + "accept": false, + "reason": "Both mention embeddings, but one is model quality and the other is dtype drift between train/eval." + }, + { + "left": "issue:43824", + "right": "issue:43872", + "accept": false, + "reason": "ImportError for a missing class vs the `_is_hf_initialized` construction TypeError." + }, + { + "left": "issue:44991", + "right": "issue:45399", + "accept": false, + "reason": "Tokenizer loading failure vs flash-attn fallback gating; unrelated code paths." + }, + { + "left": "issue:43827", + "right": "issue:44509", + "accept": true, + "reason": "Same docs regression: both say v5 docs still reference removed pipeline tasks like summarization/translation/text2text-generation." + }, + { + "left": "issue:43792", + "right": "issue:44220", + "accept": false, + "reason": "Possible Whisper-related overlap, but the titles are too vague to conclude the same underlying bug." + }, + { + "left": "issue:45081", + "right": "issue:45399", + "accept": false, + "reason": "Tokenizer regex crash vs flash-attn fallback blocking are different failure paths." + }, + { + "left": "issue:43404", + "right": "issue:43950", + "accept": false, + "reason": "Tied lm_head weights and non-persistent buffer corruption are different regressions." + }, + { + "left": "issue:43257", + "right": "issue:43866", + "accept": false, + "reason": "Different Qwen3 MoE issues: weight conversion/loading vs a corrupted checkpoint." + }, + { + "left": "issue:42915", + "right": "issue:45020", + "accept": false, + "reason": "Different symptoms and different model families; only broad version/regression context overlaps." + }, + { + "left": "issue:43742", + "right": "issue:45399", + "accept": false, + "reason": "Model-specific key error vs flash-attn fallback gating; not the same bug." + }, + { + "left": "issue:41553", + "right": "issue:44488", + "accept": false, + "reason": "Different models and different loading outcomes." + }, + { + "left": "issue:43940", + "right": "issue:45313", + "accept": false, + "reason": "Both are ZeRO-3 load failures, but for different Qwen model variants and different affected tensors." + }, + { + "left": "issue:43872", + "right": "issue:44291", + "accept": true, + "reason": "Same `_is_hf_initialized` TypeError during `init_empty_weights`/model loading; 43872 is a bitsandbytes-specific manifestation of the same incompatibility." + }, + { + "left": "issue:43854", + "right": "issue:44863", + "accept": false, + "reason": "Both are model loading failures, but they concern different models with no shared code path evident." + }, + { + "left": "issue:43257", + "right": "issue:45305", + "accept": false, + "reason": "Different DeepSpeed/Qwen training regressions with different symptoms and causes." + }, + { + "left": "issue:41553", + "right": "issue:44488", + "accept": false, + "reason": "Different model-loading reports without a concrete shared bug." + }, + { + "left": "issue:36331", + "right": "issue:41093", + "accept": false, + "reason": "Unrelated TypeErrors/IndexErrors; the shared surface of an exception is too generic to merge." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet, and the two accepted duplicate pairs are well supported by the issue titles. The soft-edge verdicts are generally conservative and avoid over-merging symptom-level similarities." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 13, + "review_unit_index": 4, + "review_unit_count": 42, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-367", + "nodes": [ + "issue:24643", + "issue:30064", + "issue:36010", + "issue:38175", + "issue:39692", + "issue:41093", + "issue:42491", + "issue:42617", + "issue:42915", + "issue:43317", + "issue:43366", + "issue:43531", + "issue:43550", + "issue:43596", + "issue:43638", + "issue:43644", + "issue:43645", + "issue:43646", + "issue:43720", + "issue:43824", + "issue:43828", + "issue:43872", + "issue:43950", + "issue:43957", + "issue:43994", + "issue:44155", + "issue:44451", + "issue:44479", + "issue:44488", + "issue:44530", + "issue:44534", + "issue:44560", + "issue:44589", + "issue:44661", + "issue:44756", + "issue:44805", + "issue:44863", + "issue:44877", + "issue:44910", + "issue:44918", + "issue:44960", + "issue:44977", + "issue:45030", + "issue:45313", + "issue:45325", + "issue:45357", + "issue:45538", + "issue:45584" + ], + "soft_pairs": [ + "issue:41093|issue:44805", + "issue:43872|issue:45538", + "issue:44479|issue:44560", + "issue:44530|issue:44863", + "issue:43366|issue:43828", + "issue:44155|issue:45584", + "issue:24643|issue:30064", + "issue:43366|issue:43531", + "issue:43550|issue:44910", + "issue:43645|issue:43646", + "issue:43950|issue:45325", + "issue:43366|issue:44589", + "issue:44560|issue:44918", + "issue:43824|issue:43957", + "issue:42915|issue:43957", + "issue:42617|issue:43366", + "issue:38175|issue:43994", + "issue:43646|issue:43950", + "issue:44560|issue:44805", + "issue:36010|issue:42915", + "issue:43638|issue:44661", + "issue:44877|issue:45030", + "issue:43596|issue:44661", + "issue:43644|issue:44534", + "issue:43317|issue:44756", + "issue:45313|issue:45357", + "issue:38175|issue:39692", + "issue:42491|issue:43720", + "issue:44960|issue:44977", + "issue:44451|issue:44488" + ], + "prepared_review_unit_hash": "3bca589c0d67b30538434f8d1195859e9608541fc6ea0c3d1a847f7188cbcd34", + "component_budget": { + "node_count": 367, + "item_count": 367, + "soft_pair_count": 1370, + "serialized_chars": 706986, + "estimated_input_tokens": 176747, + "estimated_eval_tokens": 353750 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 36654, + "estimated_input_tokens": 9164, + "estimated_eval_tokens": 18584 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "A mostly loose cluster with only two convincing duplicate pairs: the Transformers 5.x non-persistent-buffer regression (43644\u219444534) and the custom-model initialization regression (43645\u219443646). The other soft links are mostly title/model-family similarity without enough shared body evidence, so they should stay separate.", + "confidence": 0.82, + "canonical_issue_reason": "issue:43644 is the cleanest canonical anchor: it describes the Transformers 5.0.0 non-persistent-buffer corruption crisply and has a near-exact duplicate in 44534.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43644 is the strongest overall issue candidate because it is narrowly scoped, clearly reproducible, and backed by the closest duplicate match in the packet.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:41093", + "right": "issue:44805", + "accept": false, + "reason": "Same generic IndexError wording, but the packet does not provide enough shared body evidence to prove the same underlying bug." + }, + { + "left": "issue:43872", + "right": "issue:45538", + "accept": false, + "reason": "Unrelated failures: bitsandbytes initialization incompatibility vs tokenizer max-length behavior." + }, + { + "left": "issue:44479", + "right": "issue:44560", + "accept": false, + "reason": "Both are Qwen video-related, but the reported failures differ and the packet lacks evidence they are the same code-path bug." + }, + { + "left": "issue:44530", + "right": "issue:44863", + "accept": false, + "reason": "Different problems: PagedAttentionCache linear_attention crash vs NemotronH checkpoint loading." + }, + { + "left": "issue:43366", + "right": "issue:43828", + "accept": false, + "reason": "Broad model-family overlap only; gpt-oss GGUF support and Phi-tiny-MoE dtype mismatch are different issues." + }, + { + "left": "issue:44155", + "right": "issue:45584", + "accept": false, + "reason": "Different models and failure modes; no evidence they share the same underlying bug." + }, + { + "left": "issue:24643", + "right": "issue:30064", + "accept": false, + "reason": "Completely separate domains: DeepSpeed training weight-shape error vs image processor void segmentation map failure." + }, + { + "left": "issue:43366", + "right": "issue:43531", + "accept": false, + "reason": "Same ecosystem, but support for GGUF gpt-oss and a Qwen3-MoE sliding_window bug are not the same change." + }, + { + "left": "issue:43550", + "right": "issue:44910", + "accept": false, + "reason": "Different concrete failures: Bamba torch.compile/SDPA vs Qwen3.5 flash-attention position_ids handling." + }, + { + "left": "issue:43645", + "right": "issue:43646", + "accept": true, + "reason": "These describe the same custom-model initialization regression; one is a notebook-specific manifestation of the broader failure." + }, + { + "left": "issue:43950", + "right": "issue:45325", + "accept": false, + "reason": "Both are Transformers regressions, but one is non-persistent-buffer corruption and the other is Qwen2.5-VL temporal position_id scaling." + }, + { + "left": "issue:43366", + "right": "issue:44589", + "accept": false, + "reason": "No shared code-path evidence; GGUF architecture support and Float8 storage lookup are distinct bugs." + }, + { + "left": "issue:44560", + "right": "issue:44918", + "accept": false, + "reason": "Both mention Qwen video/embedding behavior, but the reported failures are different enough that they should not be merged on title similarity alone." + }, + { + "left": "issue:43824", + "right": "issue:43957", + "accept": false, + "reason": "Separate bugs: import/export naming for Qwen2.5-VL vs meta-device loading regression." + }, + { + "left": "issue:42915", + "right": "issue:43957", + "accept": false, + "reason": "Both involve model loading, but the concrete failures are different and not the same bug." + }, + { + "left": "issue:42617", + "right": "issue:43366", + "accept": false, + "reason": "A script execution failure and GGUF model support are unrelated." + }, + { + "left": "issue:38175", + "right": "issue:43994", + "accept": false, + "reason": "Shared SigLIP2 model family only; the packet does not show enough shared bug-body evidence to call them duplicates." + }, + { + "left": "issue:43646", + "right": "issue:43950", + "accept": false, + "reason": "Custom-model initialization and non-persistent-buffer corruption are different regressions." + }, + { + "left": "issue:44560", + "right": "issue:44805", + "accept": false, + "reason": "Different failure classes: video StopIteration vs mask/tensor shape mismatch." + }, + { + "left": "issue:36010", + "right": "issue:42915", + "accept": false, + "reason": "ImportError for GenerationMixin is unrelated to Qwen3Moe FP8 loading." + }, + { + "left": "issue:43638", + "right": "issue:44661", + "accept": false, + "reason": "Both are loading/registration issues, but they affect different code paths and are not the same underlying bug." + }, + { + "left": "issue:44877", + "right": "issue:45030", + "accept": false, + "reason": "Similar validation theme, but different models and different concrete config failures." + }, + { + "left": "issue:43596", + "right": "issue:44661", + "accept": false, + "reason": "DeepSpeed ZeRO-3 Bert init failure and TOKENIZER_MAPPING_NAMES restriction are unrelated." + }, + { + "left": "issue:43644", + "right": "issue:44534", + "accept": true, + "reason": "Near-verbatim duplicate: both report Transformers v5 filling non-persistent buffers with junk." + }, + { + "left": "issue:43317", + "right": "issue:44756", + "accept": false, + "reason": "Different issues: dequantized-model offload failure vs mmap/OOM behavior on Strix Halo." + }, + { + "left": "issue:45313", + "right": "issue:45357", + "accept": false, + "reason": "Same model family, but ZeRO-3 weight loading and save_pretrained visual-key serialization are different bugs." + }, + { + "left": "issue:38175", + "right": "issue:39692", + "accept": false, + "reason": "SigLIP2 overlap is too broad; the issues describe different symptoms and one is an example/documentation problem." + }, + { + "left": "issue:42491", + "right": "issue:43720", + "accept": false, + "reason": "Different code paths: Qwen3-MoE LoRA compatibility vs BitNet packed-weight loading." + }, + { + "left": "issue:44960", + "right": "issue:44977", + "accept": false, + "reason": "Both are Qwen/GLM-family generation reports, but they do not appear to be the same concrete bug." + }, + { + "left": "issue:44451", + "right": "issue:44488", + "accept": false, + "reason": "Similar BERT-loading symptom, but the packet does not provide enough evidence that they are the same underlying checkpoint issue." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "Several soft-edge accepts are not conservative enough given the packet. In particular, 41093\u219444805, 44479\u219444560, 38175\u219443994, and 44451\u219444488 are all supported only by broad title similarity/model family overlap; the packet provides no shared issue-body evidence, so treating them as duplicates is too strong. The summary also overstates the number of \u2018clear duplicates\u2019 and is a bit inconsistent with the verdict list. The strong near-exact duplicate 43644\u219444534 looks grounded, but the overall clustering should be more cautious." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 13, + "review_unit_index": 5, + "review_unit_count": 42, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-367", + "nodes": [ + "issue:36010", + "issue:39404", + "issue:39692", + "issue:41720", + "issue:42491", + "issue:42915", + "issue:42947", + "issue:43257", + "issue:43381", + "issue:43404", + "issue:43425", + "issue:43454", + "issue:43644", + "issue:43645", + "issue:43646", + "issue:43708", + "issue:43854", + "issue:43856", + "issue:43873", + "issue:43931", + "issue:43957", + "issue:43994", + "issue:44075", + "issue:44155", + "issue:44164", + "issue:44291", + "issue:44292", + "issue:44387", + "issue:44402", + "issue:44451", + "issue:44479", + "issue:44488", + "issue:44661", + "issue:44683", + "issue:44743", + "issue:44756", + "issue:44863", + "issue:44877", + "issue:44912", + "issue:44928", + "issue:44960", + "issue:44991", + "issue:45216", + "issue:45276", + "issue:45292", + "issue:45335", + "issue:45362", + "issue:45406" + ], + "soft_pairs": [ + "issue:43708|issue:44075", + "issue:43854|issue:43931", + "issue:43856|issue:43873", + "issue:42491|issue:45276", + "issue:41720|issue:42915", + "issue:43257|issue:43854", + "issue:39692|issue:43994", + "issue:43856|issue:44387", + "issue:44488|issue:44991", + "issue:45362|issue:45406", + "issue:43644|issue:43645", + "issue:42915|issue:44756", + "issue:42491|issue:43931", + "issue:44661|issue:44683", + "issue:42491|issue:43856", + "issue:43404|issue:43454", + "issue:45276|issue:45335", + "issue:36010|issue:39404", + "issue:44164|issue:44291", + "issue:41720|issue:44155", + "issue:44402|issue:44488", + "issue:43873|issue:44387", + "issue:39692|issue:44960", + "issue:44387|issue:44756", + "issue:43425|issue:43957", + "issue:43854|issue:45216", + "issue:44292|issue:44912", + "issue:42947|issue:43381", + "issue:44291|issue:44756", + "issue:45276|issue:45292", + "issue:44743|issue:44960", + "issue:43381|issue:44928", + "issue:43957|issue:44756", + "issue:44164|issue:44756", + "issue:43646|issue:43708", + "issue:45292|issue:45335", + "issue:44479|issue:44877", + "issue:44402|issue:44451", + "issue:43645|issue:44164", + "issue:43257|issue:44863" + ], + "prepared_review_unit_hash": "e07bfa254ad8ee572a4915fbba0937250bc39573183d250a41d3950f3325f6e0", + "component_budget": { + "node_count": 367, + "item_count": 367, + "soft_pair_count": 1370, + "serialized_chars": 706986, + "estimated_input_tokens": 176747, + "estimated_eval_tokens": 353750 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 40, + "serialized_chars": 39471, + "estimated_input_tokens": 9868, + "estimated_eval_tokens": 19992 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Mostly a heterogeneous set of unrelated model/load/regression reports. The only clear duplicate family is the `resize_token_embeddings` regression affecting model-specific embedding/output layers across Gemma4, T5Gemma, and the generic output_embeddings report.", + "confidence": 0.86, + "canonical_issue_reason": "Issue 45292 is the best representative of the duplicate family: it states the core, model-agnostic symptom (`resize_token_embeddings` not affecting `output_embeddings`) that the Gemma4 and T5Gemma reports appear to be instances of.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 45292 is the strongest global issue candidate because it is the most generic and reusable description of the shared `resize_token_embeddings` bug, and it is open while the near-duplicates are more model-specific.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43708", + "right": "issue:44075", + "accept": false, + "reason": "Different bugs: checkpoint resumption step calculation vs optimizer arguments being ignored." + }, + { + "left": "issue:43854", + "right": "issue:43931", + "accept": false, + "reason": "Both are loading failures, but for different models and different failure modes." + }, + { + "left": "issue:43856", + "right": "issue:43873", + "accept": false, + "reason": "One is a Qwen3 MoE memory issue; the other is offloading with quantization. Not the same underlying bug." + }, + { + "left": "issue:42491", + "right": "issue:45276", + "accept": false, + "reason": "LoRA compatibility on hf4.x/main branch is unrelated to Gemma4 token-embedding resizing." + }, + { + "left": "issue:41720", + "right": "issue:42915", + "accept": false, + "reason": "Both involve Qwen3-family models, but one is auto device mapping/cuda assert and the other is FineGrainedFP8 config failure." + }, + { + "left": "issue:43257", + "right": "issue:43854", + "accept": false, + "reason": "Accelerate+deepspeed weight conversion is unrelated to GLM-4.7-Flash loading in unit tests." + }, + { + "left": "issue:39692", + "right": "issue:43994", + "accept": false, + "reason": "Both mention SigLIP2, but one is a broken docs example and the other is bad inference outputs for a specific checkpoint." + }, + { + "left": "issue:43856", + "right": "issue:44387", + "accept": false, + "reason": "Memory inefficiency in Qwen3 MoE training is not the same as increased reserved CUDA memory under int4 quantization." + }, + { + "left": "issue:44488", + "right": "issue:44991", + "accept": false, + "reason": "Tokenizer loading failures for different models; same broad area, different underlying breakages." + }, + { + "left": "issue:45362", + "right": "issue:45406", + "accept": false, + "reason": "Qwen3.5 chat crash and Gemma4Processor missing `_tokenizer` are unrelated code paths." + }, + { + "left": "issue:43644", + "right": "issue:43645", + "accept": false, + "reason": "Non-persistent buffer junk and Jupyter custom-model initialization are different regressions." + }, + { + "left": "issue:42915", + "right": "issue:44756", + "accept": false, + "reason": "Qwen3Moe FP8 failure is unrelated to the Strix Halo mmap/OOM report." + }, + { + "left": "issue:42491", + "right": "issue:43931", + "accept": false, + "reason": "LoRA reuse on main branch is not the same as Qwen3-VL weight-shape mismatch." + }, + { + "left": "issue:44661", + "right": "issue:44683", + "accept": false, + "reason": "Tokenizer mapping lookup failure and compiled flex_attention on newer torch are unrelated." + }, + { + "left": "issue:42491", + "right": "issue:43856", + "accept": false, + "reason": "LoRA compatibility and Qwen3 MoE memory usage are separate issues." + }, + { + "left": "issue:43404", + "right": "issue:43454", + "accept": false, + "reason": "Both mention untied lm_head weights, but they are different model implementations and symptom scopes." + }, + { + "left": "issue:45276", + "right": "issue:45335", + "accept": true, + "reason": "Same regression family: `resize_token_embeddings` fails to propagate to model-specific embedding/output layers." + }, + { + "left": "issue:36010", + "right": "issue:39404", + "accept": false, + "reason": "ImportError for GenerationMixin and Whisper return_language pipeline regression are unrelated." + }, + { + "left": "issue:44164", + "right": "issue:44291", + "accept": false, + "reason": "Save/from_pretrained extra_state handling is different from init_empty_weights/_is_hf_initialized argument handling." + }, + { + "left": "issue:41720", + "right": "issue:44155", + "accept": false, + "reason": "Qwen3 device-mapping CUDA assert and AudioFlamingo3 embedding/token leak are unrelated." + }, + { + "left": "issue:44402", + "right": "issue:44488", + "accept": false, + "reason": "Tokenizer vocab mismatch for one model is not the same as loading a different model failing entirely." + }, + { + "left": "issue:43873", + "right": "issue:44387", + "accept": false, + "reason": "Quantization offloading behavior and int4 reserved-memory OOM are different failure mechanisms." + }, + { + "left": "issue:39692", + "right": "issue:44960", + "accept": false, + "reason": "SigLIP2 docs/load issues are unrelated to GLM5." + }, + { + "left": "issue:44387", + "right": "issue:44756", + "accept": false, + "reason": "Int4 memory OOM and disabling mmap on Strix Halo are different problems." + }, + { + "left": "issue:43425", + "right": "issue:43957", + "accept": false, + "reason": "Torch version incompatibility and meta-device model loading failures are not the same bug." + }, + { + "left": "issue:43854", + "right": "issue:45216", + "accept": false, + "reason": "Unit-test loading issues for GLM-4.7-Flash and Qwen3.5 save_pretrained regression are unrelated." + }, + { + "left": "issue:44292", + "right": "issue:44912", + "accept": false, + "reason": "Different model/version loading problems with no shared concrete code path." + }, + { + "left": "issue:42947", + "right": "issue:43381", + "accept": false, + "reason": "Gradient checkpointing inefficacy in LoRA and inability to use checkpointing in eval mode are distinct behaviors." + }, + { + "left": "issue:44291", + "right": "issue:44756", + "accept": false, + "reason": "init_empty_weights constructor mismatch is unrelated to Strix Halo mmap/OOM." + }, + { + "left": "issue:45276", + "right": "issue:45292", + "accept": true, + "reason": "Both report the same `resize_token_embeddings` family of failures affecting output/embedding layers." + }, + { + "left": "issue:44743", + "right": "issue:44960", + "accept": false, + "reason": "Qwen3.5 recurrent-state reset and GLM5 are unrelated." + }, + { + "left": "issue:43381", + "right": "issue:44928", + "accept": false, + "reason": "Eval-mode checkpointing limitation is unrelated to the Qwen3.5 gradient explosion from 3D position_ids." + }, + { + "left": "issue:43957", + "right": "issue:44756", + "accept": false, + "reason": "Meta-device loading regressions and Strix Halo mmap OOM are not the same underlying issue." + }, + { + "left": "issue:44164", + "right": "issue:44756", + "accept": false, + "reason": "extra_state save/load bug is unrelated to mmap/OOM behavior." + }, + { + "left": "issue:43646", + "right": "issue:43708", + "accept": false, + "reason": "Custom model initialization regression and trainer checkpoint step calculation are unrelated." + }, + { + "left": "issue:45292", + "right": "issue:45335", + "accept": true, + "reason": "Same underlying bug: `resize_token_embeddings` does not update model-specific embedding/output modules." + }, + { + "left": "issue:44479", + "right": "issue:44877", + "accept": false, + "reason": "Video-input regression across Qwen variants and strict config preventing granite_speech load are different issues." + }, + { + "left": "issue:44402", + "right": "issue:44451", + "accept": false, + "reason": "Different model/tokenizer load failures, not the same defect." + }, + { + "left": "issue:43645", + "right": "issue:44164", + "accept": false, + "reason": "Jupyter custom-model initialization and extra_state serialization are unrelated." + }, + { + "left": "issue:43257", + "right": "issue:44863", + "accept": false, + "reason": "Qwen3 MOE weight conversion with accelerate/deepspeed is unrelated to NemotronH checkpoint loading." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet and stays conservative about the soft-edge pairs. The only clearly supported duplicate family is the `resize_token_embeddings` regression spanning 45276/45292/45335, and the chosen canonical issue is a reasonable generic representative." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 13, + "review_unit_index": 6, + "review_unit_count": 42, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-367", + "nodes": [ + "issue:32090", + "issue:36032", + "issue:39404", + "issue:40990", + "issue:42947", + "issue:43299", + "issue:43421", + "issue:43425", + "issue:43452", + "issue:43582", + "issue:43596", + "issue:43611", + "issue:43632", + "issue:43644", + "issue:43645", + "issue:43716", + "issue:43828", + "issue:43856", + "issue:43883", + "issue:43901", + "issue:43906", + "issue:43931", + "issue:43957", + "issue:43994", + "issue:44079", + "issue:44164", + "issue:44206", + "issue:44220", + "issue:44265", + "issue:44479", + "issue:44488", + "issue:44509", + "issue:44530", + "issue:44589", + "issue:44671", + "issue:44683", + "issue:44756", + "issue:44805", + "issue:44811", + "issue:44863", + "issue:44898", + "issue:44912", + "issue:44918", + "issue:44991", + "issue:45081", + "issue:45084", + "issue:45161", + "issue:45584" + ], + "soft_pairs": [ + "issue:43425|issue:43828", + "issue:43828|issue:43957", + "issue:44589|issue:44756", + "issue:43906|issue:44530", + "issue:43931|issue:44863", + "issue:43611|issue:43645", + "issue:39404|issue:44912", + "issue:43425|issue:43716", + "issue:44530|issue:45084", + "issue:43901|issue:44509", + "issue:43299|issue:43931", + "issue:44265|issue:44898", + "issue:44479|issue:44805", + "issue:43994|issue:44079", + "issue:44671|issue:44991", + "issue:43856|issue:44756", + "issue:39404|issue:44671", + "issue:40990|issue:42947", + "issue:44589|issue:45084", + "issue:43596|issue:44683", + "issue:43425|issue:44918", + "issue:36032|issue:43452", + "issue:32090|issue:39404", + "issue:43582|issue:43632", + "issue:44991|issue:45081", + "issue:43957|issue:44164", + "issue:40990|issue:43421", + "issue:44164|issue:44479", + "issue:44488|issue:45081", + "issue:43883|issue:44488", + "issue:43425|issue:45161", + "issue:43611|issue:43644", + "issue:44811|issue:45584", + "issue:42947|issue:44756", + "issue:44206|issue:44220" + ], + "prepared_review_unit_hash": "a5db50251de14db2f31c420599a517837bd7aceafaa26103fd8f65fb29ba80fe", + "component_budget": { + "node_count": 367, + "item_count": 367, + "soft_pair_count": 1370, + "serialized_chars": 706986, + "estimated_input_tokens": 176747, + "estimated_eval_tokens": 353750 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 35, + "serialized_chars": 38179, + "estimated_input_tokens": 9545, + "estimated_eval_tokens": 19346 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a loose collection of unrelated Transformers v5 regressions, tokenizer/loading bugs, Whisper/audio issues, and docs complaints rather than a true duplicate set. The soft links are largely broad-similarity false positives.", + "confidence": 0.77, + "canonical_issue_reason": "43611 is the broadest representative of the v5 loading/init regression family and the closest umbrella issue among these mostly separate reports.", + "canonical_pr_reason": null, + "best_issue_reason": "43611 is the most general issue here and best represents the cluster\u2019s dominant theme of Transformers v5 model-loading regressions.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43425", + "right": "issue:43828", + "accept": false, + "reason": "Different bugs: torch version incompatibility vs an autocast dtype-mismatch in Phi-tiny-MoE." + }, + { + "left": "issue:43828", + "right": "issue:43957", + "accept": false, + "reason": "Both are model/runtime failures, but one is autocast dtype mismatch and the other is meta-device loading; not the same code path." + }, + { + "left": "issue:44589", + "right": "issue:44756", + "accept": false, + "reason": "Unrelated failures: Float8 storage lookup vs mmap/OOM behavior on Strix Halo." + }, + { + "left": "issue:43906", + "right": "issue:44530", + "accept": false, + "reason": "The first is a reproduction of an earlier issue; the second is a Qwen3.5 cache/group-type crash. Different root causes." + }, + { + "left": "issue:43931", + "right": "issue:44863", + "accept": false, + "reason": "Both involve loading models, but they are different model families with different checkpoint-shape/implementation failures." + }, + { + "left": "issue:43611", + "right": "issue:43645", + "accept": false, + "reason": "Both are v5 regressions, but one is base_model_prefix loading and the other is custom model init in notebooks; not the same bug." + }, + { + "left": "issue:39404", + "right": "issue:44912", + "accept": false, + "reason": "Whisper return_language pipeline behavior and GPT-OSS MXFP4 quantization are unrelated." + }, + { + "left": "issue:43425", + "right": "issue:43716", + "accept": false, + "reason": "Torch compatibility is not the same as a Mistral image-preprocessor dtype mismatch." + }, + { + "left": "issue:44530", + "right": "issue:45084", + "accept": false, + "reason": "Different failures: PagedAttention cache group-type crash vs template-node compilation error." + }, + { + "left": "issue:43901", + "right": "issue:44509", + "accept": false, + "reason": "Both are docs complaints, but they refer to different removed/changed API surfaces and not one concrete issue." + }, + { + "left": "issue:43299", + "right": "issue:43931", + "accept": false, + "reason": "Same general model family, but one is a Qwen3VL MoE loading regression and the other is a checkpoint shape mismatch; not clearly the same bug." + }, + { + "left": "issue:44265", + "right": "issue:44898", + "accept": false, + "reason": "torch.export with torch_compilable_check and Perceiver interpolation are separate model/export issues." + }, + { + "left": "issue:44479", + "right": "issue:44805", + "accept": false, + "reason": "Both are runtime errors, but one is a video-input regression and the other is a mask/tensor shape mismatch." + }, + { + "left": "issue:43994", + "right": "issue:44079", + "accept": false, + "reason": "Model output quality on SigLIP2 is not the same as incorrect ModelOutput key assignment." + }, + { + "left": "issue:44671", + "right": "issue:44991", + "accept": false, + "reason": "CamemBERT prediction errors and a tokenizer load failure for est-roberta are different regressions." + }, + { + "left": "issue:43856", + "right": "issue:44756", + "accept": false, + "reason": "One is Qwen3 MoE training memory usage; the other is an mmap/OOM workaround on Strix Halo." + }, + { + "left": "issue:39404", + "right": "issue:44671", + "accept": false, + "reason": "Whisper return_language and CamemBERT masked-LM predictions are unrelated bugs." + }, + { + "left": "issue:40990", + "right": "issue:42947", + "accept": false, + "reason": "Perplexity on gpt-oss and gradient checkpointing with LoRA are different problems." + }, + { + "left": "issue:44589", + "right": "issue:45084", + "accept": false, + "reason": "Float8 storage deserialization and template compilation failures are unrelated." + }, + { + "left": "issue:43596", + "right": "issue:44683", + "accept": false, + "reason": "DeepSpeed zero3/BertModel initialization and compiled flex_attention on newer torch versions are separate issues." + }, + { + "left": "issue:43425", + "right": "issue:44918", + "accept": false, + "reason": "Torch compatibility is unrelated to Qwen3.5 input-embedding unpacking in TRL SFT trainer." + }, + { + "left": "issue:36032", + "right": "issue:43452", + "accept": false, + "reason": "T5 tokenizer method-name conflict and gguf_file loading breakage are different tokenizer/model-loading bugs." + }, + { + "left": "issue:32090", + "right": "issue:39404", + "accept": false, + "reason": "Trainer GPU broadcast NoneType error is unrelated to Whisper pipeline return_language behavior." + }, + { + "left": "issue:43582", + "right": "issue:43632", + "accept": false, + "reason": "Apple Silicon allocator warmup TypeError and _is_hf_initialized flag breakage do not share a concrete code path." + }, + { + "left": "issue:44991", + "right": "issue:45081", + "accept": false, + "reason": "Both are tokenizer-loading problems, but they affect different tokenizers and different failure mechanisms." + }, + { + "left": "issue:43957", + "right": "issue:44164", + "accept": false, + "reason": "Meta-device loading regressions and save/from_pretrained extra_state handling are distinct issues." + }, + { + "left": "issue:40990", + "right": "issue:43421", + "accept": false, + "reason": "A perplexity regression is not the same as special-token post-processor updates at runtime." + }, + { + "left": "issue:44164", + "right": "issue:44479", + "accept": false, + "reason": "save/from_pretrained extra_state handling and video-input regressions are unrelated." + }, + { + "left": "issue:44488", + "right": "issue:45081", + "accept": false, + "reason": "Both are tokenizer load failures, but they target different models and different tokenizer code paths." + }, + { + "left": "issue:43883", + "right": "issue:44488", + "accept": false, + "reason": "Missing tied-weights metadata in MolmoForCausalLM is unrelated to loading sleng-bert." + }, + { + "left": "issue:43425", + "right": "issue:45161", + "accept": false, + "reason": "Torch 2.10 compatibility is a dependency issue; GPT-OSS tensor-parallel failure is a model-parallelism bug." + }, + { + "left": "issue:43611", + "right": "issue:43644", + "accept": false, + "reason": "Both are v5 regressions, but one is model-loading with base_model_prefix and the other is non-persistent buffer initialization." + }, + { + "left": "issue:44811", + "right": "issue:45584", + "accept": false, + "reason": "Both involve Whisper token handling, but batch_decode skip_special_tokens and empty-transcription generation after align_special_tokens are different bugs." + }, + { + "left": "issue:42947", + "right": "issue:44756", + "accept": false, + "reason": "LoRA gradient-checkpointing effectiveness and Strix Halo mmap OOM are unrelated." + }, + { + "left": "issue:44206", + "right": "issue:44220", + "accept": false, + "reason": "The titles suggest related audio-feature code, but they describe different failures without enough evidence of one underlying 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": 13, + "review_unit_index": 7, + "review_unit_count": 42, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-367", + "nodes": [ + "issue:34567", + "issue:36032", + "issue:39404", + "issue:40444", + "issue:41553", + "issue:42915", + "issue:43262", + "issue:43317", + "issue:43322", + "issue:43377", + "issue:43404", + "issue:43452", + "issue:43482", + "issue:43540", + "issue:43541", + "issue:43582", + "issue:43632", + "issue:43643", + "issue:43644", + "issue:43645", + "issue:43646", + "issue:43716", + "issue:43819", + "issue:43825", + "issue:43883", + "issue:43994", + "issue:44220", + "issue:44265", + "issue:44403", + "issue:44451", + "issue:44466", + "issue:44479", + "issue:44488", + "issue:44530", + "issue:44534", + "issue:44610", + "issue:44683", + "issue:44756", + "issue:44849", + "issue:44987", + "issue:44991", + "issue:45081", + "issue:45084", + "issue:45127", + "issue:45216", + "issue:45325", + "issue:45357", + "issue:45507" + ], + "soft_pairs": [ + "issue:44488|issue:44987", + "issue:34567|issue:40444", + "issue:43716|issue:45084", + "issue:43883|issue:44991", + "issue:43644|issue:43646", + "issue:43643|issue:44530", + "issue:43452|issue:43482", + "issue:43317|issue:43482", + "issue:44451|issue:44987", + "issue:36032|issue:42915", + "issue:43541|issue:45084", + "issue:39404|issue:43825", + "issue:42915|issue:44530", + "issue:44849|issue:45325", + "issue:43262|issue:43377", + "issue:43404|issue:44466", + "issue:36032|issue:43645", + "issue:43646|issue:44534", + "issue:44987|issue:44991", + "issue:45084|issue:45507", + "issue:43994|issue:44610", + "issue:43645|issue:44534", + "issue:45216|issue:45357", + "issue:43322|issue:43540", + "issue:43377|issue:43819", + "issue:44403|issue:45081", + "issue:44451|issue:45081", + "issue:44265|issue:44610", + "issue:41553|issue:43582", + "issue:43632|issue:44756", + "issue:44220|issue:44683", + "issue:43404|issue:45127", + "issue:42915|issue:45084", + "issue:44987|issue:45081", + "issue:44479|issue:45325" + ], + "prepared_review_unit_hash": "04da4f47d7694292770f2b927958323dd20fff88492dbb9be6b7a593ea0bb7ee", + "component_budget": { + "node_count": 367, + "item_count": 367, + "soft_pair_count": 1370, + "serialized_chars": 706986, + "estimated_input_tokens": 176747, + "estimated_eval_tokens": 353750 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 35, + "serialized_chars": 38188, + "estimated_input_tokens": 9547, + "estimated_eval_tokens": 19350 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Mostly a heterogeneous cluster of unrelated Transformers regressions; only a couple of pairs look like true duplicates, centered on Qwen3.5 save_pretrained serialization and GGUF loading.", + "confidence": 0.37, + "canonical_issue_reason": "The cluster is broadly mixed, but the clearest duplicate family is the Qwen3.5 `save_pretrained` regression; issue 45216 is the broader earlier report, with 45357 as a later, more specific follow-up.", + "canonical_pr_reason": null, + "best_issue_reason": "45216 is the best representative issue because it states the broader checkpoint-serialization failure and predates the narrower variant.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44488", + "right": "issue:44987", + "accept": false, + "reason": "Both are model-loading failures, but they affect different models and appear to be different breakages rather than one shared bug." + }, + { + "left": "issue:34567", + "right": "issue:40444", + "accept": false, + "reason": "Trainer token-count tracking and Qwen2.5-VL iterable-dataset image handling are unrelated code paths." + }, + { + "left": "issue:43716", + "right": "issue:45084", + "accept": false, + "reason": "Mistral-3 dtype mismatch and a template-node compilation TypeError are different failures with no clear shared root cause." + }, + { + "left": "issue:43883", + "right": "issue:44991", + "accept": false, + "reason": "Molmo tied-weights attribute error and tokenizer loading for EMBEDDIA/est-roberta are unrelated." + }, + { + "left": "issue:43644", + "right": "issue:43646", + "accept": false, + "reason": "Both mention v5 regressions, but one is about non-persistent buffers and the other about custom model initialization; not the same concrete bug." + }, + { + "left": "issue:43643", + "right": "issue:44530", + "accept": false, + "reason": "Missing fields from `trust_remote_code=True` config loading is unrelated to the Qwen3.5 linear-attention cache crash." + }, + { + "left": "issue:43452", + "right": "issue:43482", + "accept": true, + "reason": "Both report the same GGUF-loading regression in v5; 43482 looks like the Qwen2.5-specific manifestation of the generic `gguf_file` breakage in 43452." + }, + { + "left": "issue:43317", + "right": "issue:43482", + "accept": false, + "reason": "Device-map/offload loading failure is a different problem from GGUF model loading." + }, + { + "left": "issue:44451", + "right": "issue:44987", + "accept": false, + "reason": "Two separate model-loading failures with different repos and likely different causes." + }, + { + "left": "issue:36032", + "right": "issue:42915", + "accept": false, + "reason": "Tokenizer method-name conflict in T5 and Qwen3MoE FP8 loading are unrelated." + }, + { + "left": "issue:43541", + "right": "issue:45084", + "accept": false, + "reason": "Different compile/tracing errors in different models; not the same code-path issue." + }, + { + "left": "issue:39404", + "right": "issue:43825", + "accept": false, + "reason": "Whisper `return_language` behavior and a pipeline error-message regression are not the same bug." + }, + { + "left": "issue:42915", + "right": "issue:44530", + "accept": false, + "reason": "Qwen3MoE FP8 loading and Qwen3.5 PagedAttentionCache linear-attention crash are unrelated." + }, + { + "left": "issue:44849", + "right": "issue:45325", + "accept": false, + "reason": "Both are Qwen3.5/Qwen2.5-VL regressions, but one is about hidden states and the other about rope/position-id scaling; too different to merge." + }, + { + "left": "issue:43262", + "right": "issue:43377", + "accept": false, + "reason": "Audio chat-template sampling-rate defaulting and MIMI batch-vs-single padding-mask mismatch are separate issues." + }, + { + "left": "issue:43404", + "right": "issue:44466", + "accept": false, + "reason": "Both involve tied weights, but one is a missing tie in model init and the other is device-dependent serialization; different symptoms and likely different fixes." + }, + { + "left": "issue:36032", + "right": "issue:43645", + "accept": false, + "reason": "T5 tokenizer loading conflict and custom-model initialization in notebooks are unrelated." + }, + { + "left": "issue:43646", + "right": "issue:44534", + "accept": false, + "reason": "Both are v5 initialization regressions, but the titles point to different concrete failures; not enough to treat as the same bug." + }, + { + "left": "issue:44987", + "right": "issue:44991", + "accept": false, + "reason": "Two distinct tokenizer/model-loading failures for different model families." + }, + { + "left": "issue:45084", + "right": "issue:45507", + "accept": false, + "reason": "Template compilation error and GraniteMoEHybrid invalid-method call are different issues." + }, + { + "left": "issue:43994", + "right": "issue:44610", + "accept": false, + "reason": "Vision encoder nonsense outputs and OmDet-Turbo processor size mismatch are unrelated." + }, + { + "left": "issue:43645", + "right": "issue:44534", + "accept": false, + "reason": "Notebook custom-model initialization and non-persistent buffer junk are related in theme but not clearly the same underlying bug." + }, + { + "left": "issue:45216", + "right": "issue:45357", + "accept": true, + "reason": "Both describe the same Qwen3.5 `save_pretrained` serialization regression; 45357 is a later, narrower report about the incorrect visual-encoder keys." + }, + { + "left": "issue:43322", + "right": "issue:43540", + "accept": false, + "reason": "Llava Next segfault and Qwen3OmniMoe video-processing ValueError are different multimodal loading bugs." + }, + { + "left": "issue:43377", + "right": "issue:43819", + "accept": false, + "reason": "MIMI padding-mask batching bug and DAC latent reconstruction/STE bug are unrelated." + }, + { + "left": "issue:44403", + "right": "issue:45081", + "accept": false, + "reason": "Generic loading noise and a Mistral regex patch crash are different concerns." + }, + { + "left": "issue:44451", + "right": "issue:45081", + "accept": false, + "reason": "Different model-loading failures with no shared concrete code path." + }, + { + "left": "issue:44265", + "right": "issue:44610", + "accept": false, + "reason": "torch.export failure with `torch_compilable_check` and an OmDet-Turbo processor size bug are unrelated." + }, + { + "left": "issue:41553", + "right": "issue:43582", + "accept": false, + "reason": "Tokenizer error message quality and Apple Silicon allocator warmup TypeError are unrelated." + }, + { + "left": "issue:43632", + "right": "issue:44756", + "accept": false, + "reason": "HF initialization flag regression and Strix Halo mmap OOM mitigation are different issues." + }, + { + "left": "issue:44220", + "right": "issue:44683", + "accept": false, + "reason": "Fbank feature extraction and compiled flex_attention failing on torch>=2.9 are unrelated." + }, + { + "left": "issue:43404", + "right": "issue:45127", + "accept": false, + "reason": "Both mention tied weights, but one is missing weight tying in Mistral3 and the other is LoRA merge collapse with extended vocab; not the same bug." + }, + { + "left": "issue:42915", + "right": "issue:45084", + "accept": false, + "reason": "Qwen3MoE FP8 loading and a template compilation error are unrelated." + }, + { + "left": "issue:44987", + "right": "issue:45081", + "accept": false, + "reason": "Model-loading failure for physical-intelligence/fast and Mistral regex crash are different problems." + }, + { + "left": "issue:44479", + "right": "issue:45325", + "accept": false, + "reason": "Both are Qwen VL regressions, but one is a broad video-input regression and the other is a specific still-image position-id scaling bug; not clearly the same fix." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "Grounded overall: the cluster summary stays conservative, and the two accepted soft pairs are the only ones that are clearly supported by the titles. The representative/canonical issue choice for 45216 vs 45357 is reasonable, and the rest of the false verdicts do not overclaim shared root cause." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 13, + "review_unit_index": 8, + "review_unit_count": 42, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-367", + "nodes": [ + "issue:33666", + "issue:35707", + "issue:36032", + "issue:40444", + "issue:40990", + "issue:41720", + "issue:42175", + "issue:42915", + "issue:43296", + "issue:43317", + "issue:43377", + "issue:43421", + "issue:43425", + "issue:43482", + "issue:43525", + "issue:43531", + "issue:43611", + "issue:43632", + "issue:43644", + "issue:43646", + "issue:43742", + "issue:43749", + "issue:43792", + "issue:43818", + "issue:43856", + "issue:44062", + "issue:44079", + "issue:44265", + "issue:44291", + "issue:44403", + "issue:44534", + "issue:44568", + "issue:44589", + "issue:44683", + "issue:44811", + "issue:44829", + "issue:44877", + "issue:44987", + "issue:44991", + "issue:45020", + "issue:45072", + "issue:45325", + "issue:45357", + "issue:45381", + "issue:45479", + "issue:45491", + "issue:45538", + "issue:45542" + ], + "soft_pairs": [ + "issue:43296|issue:43531", + "issue:42175|issue:45542", + "issue:44062|issue:44987", + "issue:42915|issue:44291", + "issue:40990|issue:44811", + "issue:43611|issue:44534", + "issue:43317|issue:43856", + "issue:36032|issue:43742", + "issue:40444|issue:45325", + "issue:43632|issue:44534", + "issue:43531|issue:44877", + "issue:42915|issue:44589", + "issue:43818|issue:45357", + "issue:43421|issue:44568", + "issue:36032|issue:41720", + "issue:43482|issue:45538", + "issue:44683|issue:45020", + "issue:44829|issue:45479", + "issue:33666|issue:40444", + "issue:43425|issue:44403", + "issue:44403|issue:44991", + "issue:44079|issue:44534", + "issue:35707|issue:40444", + "issue:43377|issue:45491", + "issue:43792|issue:44589", + "issue:43482|issue:43525", + "issue:44291|issue:44589", + "issue:43749|issue:44079", + "issue:44079|issue:45072", + "issue:43644|issue:44079", + "issue:43632|issue:45072", + "issue:44265|issue:45072", + "issue:40444|issue:45381", + "issue:45325|issue:45381", + "issue:43531|issue:44291", + "issue:43611|issue:43646" + ], + "prepared_review_unit_hash": "e88ff0dea395ae7723cb1e1d31972aaef7dbc0cfb68788970213e168b38d9062", + "component_budget": { + "node_count": 367, + "item_count": 367, + "soft_pair_count": 1370, + "serialized_chars": 706986, + "estimated_input_tokens": 176747, + "estimated_eval_tokens": 353750 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 36, + "serialized_chars": 38398, + "estimated_input_tokens": 9600, + "estimated_eval_tokens": 19456 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This is a noisy mixed cluster of mostly unrelated Transformers regression issues. I only see a few genuine duplicate families: the `_is_hf_initialized`/`init_empty_weights` v5 init regression, the non-persistent buffer junk regression, and likely the Qwen2.5-VL rope/position-id bug.", + "confidence": 0.79, + "canonical_issue_reason": "issue:43632 is the clearest representative of a concrete v5 loading/init regression and anchors one of the strongest duplicate pairs in the set.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43632 is the most specific, actionable issue among the duplicate candidates and best represents the shared `_is_hf_initialized` initialization bug.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43296", + "right": "issue:43531", + "accept": false, + "reason": "Different models and failures: PaddleOCR-VL load failure vs Qwen3-MoE sliding-window behavior." + }, + { + "left": "issue:42175", + "right": "issue:45542", + "accept": false, + "reason": "Both mention TF installation/backend, but one is missing TensorFlow in extras while the other is a runtime backend selection error." + }, + { + "left": "issue:44062", + "right": "issue:44987", + "accept": false, + "reason": "Tokenizer AddedToken special-arg conflict and a model load failure for a different repo are not the same bug." + }, + { + "left": "issue:42915", + "right": "issue:44291", + "accept": false, + "reason": "Qwen3Moe FP8 config failure and `_is_hf_initialized` loading TypeError are unrelated regressions." + }, + { + "left": "issue:40990", + "right": "issue:44811", + "accept": false, + "reason": "Perplexity regression on GPT-OSS and Whisper skip_special_tokens decoding are different code paths." + }, + { + "left": "issue:43611", + "right": "issue:44534", + "accept": false, + "reason": "Both are Transformers v5 regressions, but one is model loading with base_model_prefix and the other is buffer initialization junk; no clear same bug." + }, + { + "left": "issue:43317", + "right": "issue:43856", + "accept": false, + "reason": "Device-map/offload loading failure and Qwen3 MoE training memory usage are different issues." + }, + { + "left": "issue:36032", + "right": "issue:43742", + "accept": false, + "reason": "Tokenizer add_special_tokens conflict vs MobileLLM key error are distinct loading failures." + }, + { + "left": "issue:40444", + "right": "issue:45325", + "accept": false, + "reason": "Both involve Qwen2.5-VL, but one is iterable-dataset multi-image finetuning and the other is rope/temporal position scaling." + }, + { + "left": "issue:43632", + "right": "issue:44534", + "accept": false, + "reason": "These are different v5 regressions: `_is_hf_initialized` loading behavior vs non-persistent buffer contents." + }, + { + "left": "issue:43531", + "right": "issue:44877", + "accept": false, + "reason": "Qwen3-MoE sliding-window behavior and granite_speech strict config loading are unrelated." + }, + { + "left": "issue:42915", + "right": "issue:44589", + "accept": false, + "reason": "FP8 config failure and missing Float8 storage are different failure modes." + }, + { + "left": "issue:43818", + "right": "issue:45357", + "accept": false, + "reason": "Video-LLaVA missing temporal attention/shared weights is unrelated to Qwen3.5 save_pretrained key regression." + }, + { + "left": "issue:43421", + "right": "issue:44568", + "accept": false, + "reason": "Runtime post-processor update for special tokens and BOS/EOS omission on add_special_tokens are related in topic but not the same concrete bug." + }, + { + "left": "issue:36032", + "right": "issue:41720", + "accept": false, + "reason": "Tokenizer loading error and Qwen3 auto device-map cuda assert are unrelated." + }, + { + "left": "issue:43482", + "right": "issue:45538", + "accept": false, + "reason": "GGUF loading with transformers v5 and CLIPTokenizer max_length are different issues." + }, + { + "left": "issue:44683", + "right": "issue:45020", + "accept": false, + "reason": "Torch compiled flex_attention failure and remote_code model loading regressions do not share a code path." + }, + { + "left": "issue:44829", + "right": "issue:45479", + "accept": false, + "reason": "Flash-Attention-3 training degeneration and single-label-zero-loss are different sequence-classification problems." + }, + { + "left": "issue:33666", + "right": "issue:40444", + "accept": false, + "reason": "Both are Qwen VL training issues, but one is multi-GPU training and the other is multi-image iterable dataset handling." + }, + { + "left": "issue:43425", + "right": "issue:44403", + "accept": false, + "reason": "Torch 2.10 incompatibility and noisy transformer loading are unrelated." + }, + { + "left": "issue:44403", + "right": "issue:44991", + "accept": false, + "reason": "Noisy loading vs tokenizer load failure for a specific model are different symptoms and code paths." + }, + { + "left": "issue:44079", + "right": "issue:44534", + "accept": false, + "reason": "ModelOutput key assignment and non-persistent buffer junk are unrelated." + }, + { + "left": "issue:35707", + "right": "issue:40444", + "accept": false, + "reason": "Progressive generation with inputs_embeds/past_key_values is unrelated to Qwen2.5-VL multi-image finetuning." + }, + { + "left": "issue:43377", + "right": "issue:45491", + "accept": false, + "reason": "MIMI encoder padding-mask mismatch and Gemma3 sliding-window NaNs are different bugs." + }, + { + "left": "issue:43792", + "right": "issue:44589", + "accept": false, + "reason": "Whisper runtime failure and Float8 storage lookup error are not the same problem." + }, + { + "left": "issue:43482", + "right": "issue:43525", + "accept": false, + "reason": "GGUF loading failure and missing pad_token_id on Llama4Config are unrelated." + }, + { + "left": "issue:44291", + "right": "issue:44589", + "accept": false, + "reason": "`init_empty_weights`/_is_hf_initialized and Float8 storage errors are different regressions." + }, + { + "left": "issue:43749", + "right": "issue:44079", + "accept": false, + "reason": "FSDP CPU RAM efficient loading and ModelOutput key assignment are unrelated." + }, + { + "left": "issue:44079", + "right": "issue:45072", + "accept": false, + "reason": "ModelOutput key handling and bfloat16 dtype mismatch are different issues." + }, + { + "left": "issue:43644", + "right": "issue:44079", + "accept": false, + "reason": "Non-persistent buffer junk and ModelOutput key assignment are unrelated." + }, + { + "left": "issue:43632", + "right": "issue:45072", + "accept": false, + "reason": "Different Transformers v5 regressions: initialization flag handling vs dtype mismatch in inference." + }, + { + "left": "issue:44265", + "right": "issue:45072", + "accept": false, + "reason": "torch.export/torch_compilable_check failure and bfloat16 dtype mismatch are unrelated." + }, + { + "left": "issue:40444", + "right": "issue:45381", + "accept": false, + "reason": "Both are Qwen2.5-VL, but one is multi-image finetuning failure and the other is video position-id corruption." + }, + { + "left": "issue:45325", + "right": "issue:45381", + "accept": true, + "reason": "Both describe the same Qwen2.5-VL rope/position-id computation bug in transformers 5.3.0, just exposed by still-image vs video inputs." + }, + { + "left": "issue:43531", + "right": "issue:44291", + "accept": false, + "reason": "Qwen3-MoE sliding_window behavior and `_is_hf_initialized` loading TypeError are unrelated." + }, + { + "left": "issue:43611", + "right": "issue:43646", + "accept": false, + "reason": "Both are v5 loading/init regressions, but the titles point to different concrete failures and the same code path is not clear enough to merge." + }, + { + "left": "issue:44291", + "right": "issue:43632", + "accept": true, + "reason": "These are the same `_is_hf_initialized` / init-empty-weights Transformers v5 loading regression expressed from two angles." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is conservative and grounded in the packet. The rejected pairs mostly align with distinct titles/failure modes, and the accepted pairs are plausibly the same bug family and mergeable into one fix. No clear overstatement beyond the evidence shown." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 13, + "review_unit_index": 9, + "review_unit_count": 42, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-367", + "nodes": [ + "issue:34634", + "issue:35707", + "issue:39692", + "issue:40444", + "issue:40990", + "issue:41720", + "issue:42175", + "issue:42915", + "issue:43262", + "issue:43296", + "issue:43317", + "issue:43377", + "issue:43454", + "issue:43482", + "issue:43526", + "issue:43550", + "issue:43606", + "issue:43632", + "issue:43701", + "issue:43746", + "issue:43749", + "issue:43792", + "issue:43819", + "issue:43828", + "issue:43873", + "issue:43881", + "issue:43883", + "issue:43957", + "issue:43994", + "issue:44186", + "issue:44265", + "issue:44291", + "issue:44387", + "issue:44466", + "issue:44488", + "issue:44534", + "issue:44610", + "issue:44617", + "issue:44683", + "issue:44877", + "issue:44898", + "issue:44991", + "issue:45072", + "issue:45081", + "issue:45325", + "issue:45357", + "issue:45563", + "issue:45584" + ], + "soft_pairs": [ + "issue:41720|issue:45081", + "issue:43881|issue:44488", + "issue:42175|issue:43296", + "issue:43454|issue:44466", + "issue:43994|issue:44898", + "issue:44387|issue:44617", + "issue:43828|issue:44683", + "issue:42915|issue:43482", + "issue:43526|issue:43883", + "issue:43701|issue:45325", + "issue:43792|issue:43957", + "issue:43262|issue:43746", + "issue:43632|issue:45563", + "issue:40990|issue:45584", + "issue:43317|issue:43873", + "issue:43881|issue:44991", + "issue:43606|issue:44265", + "issue:34634|issue:35707", + "issue:44291|issue:44877", + "issue:44186|issue:44265", + "issue:43606|issue:43819", + "issue:43377|issue:44898", + "issue:35707|issue:41720", + "issue:43454|issue:45072", + "issue:43454|issue:43550", + "issue:44610|issue:45325", + "issue:43377|issue:44610", + "issue:43749|issue:44534", + "issue:39692|issue:40444", + "issue:44466|issue:45357", + "issue:43296|issue:44991", + "issue:43296|issue:44877" + ], + "prepared_review_unit_hash": "90356b718f479d8f0ccfa85bebc20aec1d0225a4f4c99732d3939d6bc9c2a438", + "component_budget": { + "node_count": 367, + "item_count": 367, + "soft_pair_count": 1370, + "serialized_chars": 706986, + "estimated_input_tokens": 176747, + "estimated_eval_tokens": 353750 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 32, + "serialized_chars": 37330, + "estimated_input_tokens": 9333, + "estimated_eval_tokens": 18922 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a set of false-positive similarities around model loading/offloading, quantization, and a few unrelated vision/audio/tokenizer regressions. Only one pair looks like a plausible duplicate-level match: offloading with quantization/device placement.", + "confidence": 0.46, + "canonical_issue_reason": "issue:43873 is the broadest and most central issue in the only real overlap area (offloading + quantization/device placement), and it can naturally absorb the narrower repro in issue:43317.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43873 is the best representative issue overall: it is open, broadly scoped, and already has inbound references, while the other items are mostly narrower one-off bugs or unrelated regressions.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:41720", + "right": "issue:45081", + "accept": false, + "reason": "Different bugs: device-map CUDA assert vs tokenizer regex patch crash." + }, + { + "left": "issue:43881", + "right": "issue:44488", + "accept": false, + "reason": "Both are load failures, but for different models and likely different root causes." + }, + { + "left": "issue:42175", + "right": "issue:43296", + "accept": false, + "reason": "One is backend/extras packaging, the other is a model loading failure in vLLM." + }, + { + "left": "issue:43454", + "right": "issue:44466", + "accept": false, + "reason": "Both mention lm_head/tied weights, but one is a missing tie causing bad generation and the other is device-dependent serialization; not the same concrete bug." + }, + { + "left": "issue:43994", + "right": "issue:44898", + "accept": false, + "reason": "Different vision-model regressions with different symptoms and code paths." + }, + { + "left": "issue:44387", + "right": "issue:44617", + "accept": false, + "reason": "Both are OOMs, but one is an int4 quantization memory regression and the other is a model-specific Sam3Video OOM." + }, + { + "left": "issue:43828", + "right": "issue:44683", + "accept": false, + "reason": "Autocast dtype mismatch and compiled flex_attention failure are unrelated." + }, + { + "left": "issue:42915", + "right": "issue:43482", + "accept": false, + "reason": "Different model families and different failure modes (FP8 config vs GGUF/v5 loading)." + }, + { + "left": "issue:43526", + "right": "issue:43883", + "accept": false, + "reason": "A label-reduction bug is not the same as a missing attribute during model loading." + }, + { + "left": "issue:43701", + "right": "issue:45325", + "accept": false, + "reason": "Checkpoint resume key mismatch vs Qwen2.5-VL rope-index scaling are unrelated." + }, + { + "left": "issue:43792", + "right": "issue:43957", + "accept": false, + "reason": "Whisper runtime failure and meta-device loading regression are different issues." + }, + { + "left": "issue:43262", + "right": "issue:43746", + "accept": false, + "reason": "Audio chat-template sample-rate default bug vs PEFT checkpoint loading bug." + }, + { + "left": "issue:43632", + "right": "issue:45563", + "accept": false, + "reason": "_is_hf_initialized regression and stale paged-generate warning are unrelated." + }, + { + "left": "issue:40990", + "right": "issue:45584", + "accept": false, + "reason": "Perplexity regression and empty-transcription generation failure are different." + }, + { + "left": "issue:43317", + "right": "issue:43873", + "accept": true, + "reason": "Both describe offloading/device_map loading failures in the quantization/dequantization path and could plausibly be fixed together." + }, + { + "left": "issue:43881", + "right": "issue:44991", + "accept": false, + "reason": "Different model/tokenizer loading failures with no clear shared root cause." + }, + { + "left": "issue:43606", + "right": "issue:44265", + "accept": false, + "reason": "CPU offload device mismatch is unrelated to torch.export failures around torch_compilable_check." + }, + { + "left": "issue:34634", + "right": "issue:35707", + "accept": false, + "reason": "Bark voice preset handling and progressive generation with inputs_embeds/past_key_values are unrelated." + }, + { + "left": "issue:44291", + "right": "issue:44877", + "accept": false, + "reason": "Unexpected _is_hf_initialized argument and strict granite_speech config loading are different bugs." + }, + { + "left": "issue:44186", + "right": "issue:44265", + "accept": false, + "reason": "Tokenizer padding/NER crash is unrelated to torch.export compatibility." + }, + { + "left": "issue:43606", + "right": "issue:43819", + "accept": false, + "reason": "CPU offload device mismatch and DAC.from_latents forward mismatch are different code paths." + }, + { + "left": "issue:43377", + "right": "issue:44898", + "accept": false, + "reason": "Padding-mask batching bug and Perceiver image-size issue are unrelated." + }, + { + "left": "issue:35707", + "right": "issue:41720", + "accept": false, + "reason": "Progressive generation bug vs device-mapping CUDA assert are unrelated." + }, + { + "left": "issue:43454", + "right": "issue:45072", + "accept": false, + "reason": "Tied-weights serialization/generation bug vs bf16 dtype mismatch are different." + }, + { + "left": "issue:43454", + "right": "issue:43550", + "accept": false, + "reason": "Tied-weights bug and torch.compile SDPA failure do not share a concrete root cause." + }, + { + "left": "issue:44610", + "right": "issue:45325", + "accept": false, + "reason": "Processor resolution mismatch and temporal rope-index scaling are different bugs." + }, + { + "left": "issue:43377", + "right": "issue:44610", + "accept": false, + "reason": "Batched padding-mask behavior and processor/model input-size mismatch are unrelated." + }, + { + "left": "issue:43749", + "right": "issue:44534", + "accept": false, + "reason": "FSDP RAM-efficient loading and non-persistent buffer initialization junk are different regressions." + }, + { + "left": "issue:39692", + "right": "issue:40444", + "accept": false, + "reason": "Doc-example model/processor mismatch and multi-image IterableDataset fine-tuning failure are not the same bug." + }, + { + "left": "issue:44466", + "right": "issue:45357", + "accept": false, + "reason": "lm_head serialization issue and incorrect visual-encoder key saving are different save_pretrained regressions." + }, + { + "left": "issue:43296", + "right": "issue:44991", + "accept": false, + "reason": "PaddleOCR-VL/vLLM loading failure and tokenizer loading for est-roberta are unrelated." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "Grounded overall and conservatively framed. The representative issue choice is supported by the packet (open, broader scope, inbound refs), and the one accepted soft-edge pair is plausibly the same offloading/quantization bug. The remaining pair rejections are cautious and do not overclaim shared root cause." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 13, + "review_unit_index": 10, + "review_unit_count": 42, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-367", + "nodes": [ + "issue:36010", + "issue:36331", + "issue:41720", + "issue:42617", + "issue:42915", + "issue:43316", + "issue:43317", + "issue:43322", + "issue:43388", + "issue:43404", + "issue:43454", + "issue:43526", + "issue:43531", + "issue:43540", + "issue:43550", + "issue:43572", + "issue:43577", + "issue:43606", + "issue:43708", + "issue:43720", + "issue:43782", + "issue:43792", + "issue:43873", + "issue:43881", + "issue:44186", + "issue:44292", + "issue:44361", + "issue:44423", + "issue:44479", + "issue:44514", + "issue:44545", + "issue:44610", + "issue:44671", + "issue:44683", + "issue:44743", + "issue:44756", + "issue:44877", + "issue:44898", + "issue:44913", + "issue:44918", + "issue:44977", + "issue:44991", + "issue:45059", + "issue:45072", + "issue:45161", + "issue:45357", + "issue:45406", + "issue:45464" + ], + "soft_pairs": [ + "issue:43454|issue:43720", + "issue:41720|issue:43322", + "issue:43792|issue:44683", + "issue:41720|issue:44977", + "issue:43404|issue:45357", + "issue:44423|issue:45406", + "issue:44186|issue:45072", + "issue:43526|issue:43881", + "issue:43873|issue:44756", + "issue:43577|issue:43708", + "issue:42617|issue:42915", + "issue:43550|issue:43720", + "issue:44610|issue:44898", + "issue:44877|issue:44991", + "issue:44479|issue:44918", + "issue:44186|issue:44610", + "issue:36331|issue:42915", + "issue:36331|issue:42617", + "issue:44671|issue:44743", + "issue:43316|issue:44913", + "issue:43550|issue:43606", + "issue:42915|issue:43317", + "issue:41720|issue:42617", + "issue:44292|issue:44479", + "issue:41720|issue:45161", + "issue:43531|issue:43572", + "issue:44186|issue:44361", + "issue:43388|issue:45059", + "issue:43540|issue:43782", + "issue:41720|issue:43873", + "issue:42915|issue:43322", + "issue:36010|issue:42617", + "issue:45406|issue:45464", + "issue:36331|issue:41720", + "issue:44514|issue:44545" + ], + "prepared_review_unit_hash": "c40d944a7ebd82310614d29f462d0d5ccaa5648806c978e1b27b59e8ad14481a", + "component_budget": { + "node_count": 367, + "item_count": 367, + "soft_pair_count": 1370, + "serialized_chars": 706986, + "estimated_input_tokens": 176747, + "estimated_eval_tokens": 353750 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 35, + "serialized_chars": 38022, + "estimated_input_tokens": 9506, + "estimated_eval_tokens": 19268 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly heterogeneous, but one near-exact duplicate pair stands out: `Qwen2_5_VLProcessor.apply_chat_template` crashes on batched input when `padding=False` (issues 44514 and 44545). The rest are related at most by broad themes like loading/offloading, quantization, and model-specific regressions, but not the same underlying bug.", + "confidence": 0.77, + "canonical_issue_reason": "Issue 44545 is the clearest canonical issue: it is an exact duplicate of 44514 with a cleaner title and the same underlying batched `apply_chat_template` padding=False crash.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 44545 is the best issue to keep as the representative duplicate target because it most directly names the concrete failing code path and matches the only clear duplicate pair.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43454", + "right": "issue:43720", + "accept": false, + "reason": "Both are model-loading/weight-handling bugs, but they involve different models and different failure modes (`lm_head` tying vs packed-weight unpacking)." + }, + { + "left": "issue:41720", + "right": "issue:43322", + "accept": false, + "reason": "Both mention severe runtime failures, but one is Qwen3 auto device mapping on A800 and the other is Llava Next loading segfault; no shared concrete bug." + }, + { + "left": "issue:43792", + "right": "issue:44683", + "accept": false, + "reason": "Different subsystems and root causes: Whisper loading/run failure vs compiled flex_attention on torch >= 2.9." + }, + { + "left": "issue:41720", + "right": "issue:44977", + "accept": false, + "reason": "Both involve Qwen3-family generation/runtime issues, but one is a CUDA assert during auto device mapping and the other is flash-attention generation behavior." + }, + { + "left": "issue:43404", + "right": "issue:45357", + "accept": false, + "reason": "Both are weight-tying/serialization problems, but they affect different models and different assets (Mistral3 lm_head vs Qwen3.5 visual encoder keys)." + }, + { + "left": "issue:44423", + "right": "issue:45406", + "accept": false, + "reason": "Different crash sites: continuous-batching multimodal AttributeError vs serve crash from missing `_tokenizer` in `Gemma4Processor`." + }, + { + "left": "issue:44186", + "right": "issue:45072", + "accept": false, + "reason": "Both are dtype/input handling bugs, but one is tokenizer NER padding/truncation and the other is inference dtype mismatches in wrappers." + }, + { + "left": "issue:43526", + "right": "issue:43881", + "accept": false, + "reason": "Different components and symptoms: BeitImageProcessorFast label reduction vs glm-4v-9b loading failure." + }, + { + "left": "issue:43873", + "right": "issue:44756", + "accept": false, + "reason": "Both discuss offloading/memory behavior, but one is quantization offloading and the other is mmap/OOM on Strix Halo; not the same bug." + }, + { + "left": "issue:43577", + "right": "issue:43708", + "accept": false, + "reason": "Different trainer/config issues: BLIP2 dtype propagation vs resume_from_checkpoint max_steps calculation." + }, + { + "left": "issue:42617", + "right": "issue:42915", + "accept": false, + "reason": "Both are runtime failures, but one is running `3d_parallel.py` and the other is Qwen3Moe with FineGrainedFP8Config; no shared code path." + }, + { + "left": "issue:43550", + "right": "issue:43720", + "accept": false, + "reason": "Different issues in acceleration/compilation vs loading packed weights; too broad to be the same bug." + }, + { + "left": "issue:44610", + "right": "issue:44898", + "accept": false, + "reason": "Both concern image-size/preprocessing mismatches, but they affect different models and different expected resolutions." + }, + { + "left": "issue:44877", + "right": "issue:44991", + "accept": false, + "reason": "Both are config/tokenizer loading regressions, but one is strict config for granite_speech and the other is tokenizer loading for a specific model." + }, + { + "left": "issue:44479", + "right": "issue:44918", + "accept": false, + "reason": "Both mention Qwen3.5-family regressions, but one is video-input regression and the other is SFT unpacking of input embeddings." + }, + { + "left": "issue:44186", + "right": "issue:44610", + "accept": false, + "reason": "Different model families and bugs: LayoutLMv2Tokenizer crashes vs OmDet-Turbo processor output size mismatch." + }, + { + "left": "issue:36331", + "right": "issue:42915", + "accept": false, + "reason": "Trainer custom loss signature mismatch vs Qwen3Moe FP8 failure; unrelated failures." + }, + { + "left": "issue:36331", + "right": "issue:42617", + "accept": false, + "reason": "Unexpected `num_items_in_batch` argument is a trainer API mismatch, not related to 3d_parallel execution failure." + }, + { + "left": "issue:44671", + "right": "issue:44743", + "accept": false, + "reason": "Different Qwen3.5 areas: CamemBERT masked LM predictions vs recurrent states resetting in modular_qwen3_5.py." + }, + { + "left": "issue:43316", + "right": "issue:44913", + "accept": false, + "reason": "Both are config reload discrepancies, but they involve different fields and model families (`Gemma3TextConfig` vs GPTNeoX rotary_pct)." + }, + { + "left": "issue:43550", + "right": "issue:43606", + "accept": false, + "reason": "Both involve device/compile/offload behavior, but one is SDPA with torch.compile and the other is CPU offload device mismatch for bark-small." + }, + { + "left": "issue:42915", + "right": "issue:43317", + "accept": false, + "reason": "Both are loading/offload-related, but one is Qwen3Moe FP8 failure and the other is device_map=auto dequantized model offload failure." + }, + { + "left": "issue:41720", + "right": "issue:42617", + "accept": false, + "reason": "Both are execution failures, but one is CUDA assert on auto device mapping and the other is inability to run `3d_parallel.py`; not the same bug." + }, + { + "left": "issue:44292", + "right": "issue:44479", + "accept": false, + "reason": "Different Qwen variants and different failure modes: NVFP4 run error vs video-input regression." + }, + { + "left": "issue:41720", + "right": "issue:45161", + "accept": false, + "reason": "Both touch parallelism/device placement, but one is a Qwen3 auto device-map crash and the other is only TP not working in GPT-OSS MoE." + }, + { + "left": "issue:43531", + "right": "issue:43572", + "accept": false, + "reason": "Different model/config issues: Qwen3-MoE sliding window vs StableLmConfig missing pad_token_idx." + }, + { + "left": "issue:44186", + "right": "issue:44361", + "accept": false, + "reason": "Both are tokenizer/input-processing bugs, but they affect different tokenizers and different crash conditions." + }, + { + "left": "issue:43388", + "right": "issue:45059", + "accept": false, + "reason": "Different multimodal/data-prep issues: metric gathering drops labels vs SAM3 text/bounding-box behavior." + }, + { + "left": "issue:43540", + "right": "issue:43782", + "accept": false, + "reason": "Different Qwen3 Omni/MoE input-processing and from_pretrained weight_only error; not the same underlying bug." + }, + { + "left": "issue:41720", + "right": "issue:43873", + "accept": false, + "reason": "Both involve Qwen3 and device/offload concerns, but one is a CUDA assert on A800 and the other is quantization offloading behavior; not mergeable as one fix." + }, + { + "left": "issue:42915", + "right": "issue:43322", + "accept": false, + "reason": "Different model paths and failure types: FineGrainedFP8Config failure vs segmentation fault when loading Llava Next." + }, + { + "left": "issue:36010", + "right": "issue:42617", + "accept": false, + "reason": "Completely different problems: import path regression vs inability to run `3d_parallel.py`." + }, + { + "left": "issue:45406", + "right": "issue:45464", + "accept": false, + "reason": "Both affect serving, but one is a missing `_tokenizer` on `Gemma4Processor` and the other is streaming chat/completions on Qwen3.5-0.8B." + }, + { + "left": "issue:36331", + "right": "issue:41720", + "accept": false, + "reason": "Trainer loss-signature API mismatch is unrelated to Qwen3 auto device-map CUDA failure." + }, + { + "left": "issue:44514", + "right": "issue:44545", + "accept": true, + "reason": "Exact same crash and API path; the titles/body describe the same batched `Qwen2_5_VLProcessor.apply_chat_template` padding=False bug, with only minor wording differences." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet and stays conservative. The only accepted duplicate pair, 44514/44545, is well supported by the near-identical titles and the shared `Qwen2_5_VLProcessor.apply_chat_template` padding=False crash. The remaining soft-edge verdicts are appropriately cautious and 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": 13, + "review_unit_index": 11, + "review_unit_count": 42, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-367", + "nodes": [ + "issue:30064", + "issue:33357", + "issue:34567", + "issue:34634", + "issue:35707", + "issue:39692", + "issue:41720", + "issue:42886", + "issue:43381", + "issue:43421", + "issue:43525", + "issue:43531", + "issue:43550", + "issue:43606", + "issue:43653", + "issue:43720", + "issue:43746", + "issue:43782", + "issue:43844", + "issue:43873", + "issue:43927", + "issue:43994", + "issue:44062", + "issue:44077", + "issue:44186", + "issue:44368", + "issue:44451", + "issue:44492", + "issue:44509", + "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:45183", + "issue:45356", + "issue:45357", + "issue:45381", + "issue:45479" + ], + "soft_pairs": [ + "issue:33357|issue:35707", + "issue:43606|issue:43720", + "issue:41720|issue:43746", + "issue:30064|issue:34634", + "issue:43550|issue:43653", + "issue:44492|issue:44509", + "issue:44368|issue:45357", + "issue:44625|issue:45479", + "issue:43606|issue:43653", + "issue:43421|issue:43927", + "issue:44898|issue:45072", + "issue:43782|issue:45381", + "issue:34567|issue:35707", + "issue:34634|issue:41720", + "issue:43381|issue:43844", + "issue:42886|issue:44843", + "issue:43421|issue:43653", + "issue:44077|issue:44625", + "issue:44451|issue:44991", + "issue:44062|issue:44451", + "issue:44987|issue:45042", + "issue:43531|issue:44589", + "issue:44918|issue:44936", + "issue:43994|issue:44671", + "issue:39692|issue:44610", + "issue:43421|issue:44186", + "issue:45081|issue:45356", + "issue:43525|issue:43531", + "issue:39692|issue:43873", + "issue:45183|issue:45479", + "issue:43746|issue:45381" + ], + "prepared_review_unit_hash": "27bbba0e0dac76012521bdd0bbd6a1b16f2717e0fd4bd2e4bd4b22c7114c4e41", + "component_budget": { + "node_count": 367, + "item_count": 367, + "soft_pair_count": 1370, + "serialized_chars": 706986, + "estimated_input_tokens": 176747, + "estimated_eval_tokens": 353750 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 31, + "serialized_chars": 36694, + "estimated_input_tokens": 9174, + "estimated_eval_tokens": 18604 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly heterogeneous: tokenizer/load regressions, model-specific inference bugs, processor mismatches, and training issues. Only one soft edge looks like the same underlying bug: the offline tokenizer/cache failure tied to _patch_mistral_regex.", + "confidence": 0.67, + "canonical_issue_reason": "issue:44843 is the clearest root-cause report; it names the concrete failing code path (_patch_mistral_regex calling model_info() unconditionally), which directly explains the offline-cache symptom in issue:42886.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:44843 is the most actionable representative of the only real duplicate subcluster, because it identifies the exact offending call site and the HF_HUB_OFFLINE breakage.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:33357", + "right": "issue:35707", + "accept": false, + "reason": "Different failures in unrelated paths: MacOS bus error on a CLIP model vs. progressive generation with inputs_embeds/past_key_values." + }, + { + "left": "issue:43606", + "right": "issue:43720", + "accept": false, + "reason": "Both involve loading/offload context, but they are different bugs with different models and failure modes (CPU offload device mismatch vs. packed weights not unpacked)." + }, + { + "left": "issue:41720", + "right": "issue:43746", + "accept": false, + "reason": "Different model families and root causes: Qwen3 device-mapping cuda assert vs. GraniteSpeech PEFT local checkpoint loading." + }, + { + "left": "issue:30064", + "right": "issue:34634", + "accept": false, + "reason": "Unrelated processor bugs: void segmentation maps vs. BarkProcessor voice_preset handling." + }, + { + "left": "issue:43550", + "right": "issue:43653", + "accept": false, + "reason": "Different subsystems entirely: torch.compile/SDPA on Bamba vs. BigBirdTokenizer special-token registration." + }, + { + "left": "issue:44492", + "right": "issue:44509", + "accept": false, + "reason": "A typo in cache strategy docs is not the same issue as stale pipeline-task documentation." + }, + { + "left": "issue:44368", + "right": "issue:45357", + "accept": false, + "reason": "One is a warning about tie_word_embeddings during LoRA fine-tuning; the other is a save_pretrained regression for Qwen3.5 visual encoder keys." + }, + { + "left": "issue:44625", + "right": "issue:45479", + "accept": false, + "reason": "Both mention classification config, but one is Qwen3.5 num_labels propagation and the other is a generic zero-loss bug for num_labels=1." + }, + { + "left": "issue:43606", + "right": "issue:43653", + "accept": false, + "reason": "Different bugs: offloading/device mismatch vs. tokenizer special-token decoding." + }, + { + "left": "issue:43421", + "right": "issue:43927", + "accept": false, + "reason": "Both are tokenizer/config related, but one is runtime post-processor updates and the other is save/load losing custom token IDs." + }, + { + "left": "issue:44898", + "right": "issue:45072", + "accept": false, + "reason": "Different model/runtime problems: Perceiver image-size handling vs. dtype mismatch in bfloat16 inference." + }, + { + "left": "issue:43782", + "right": "issue:45381", + "accept": false, + "reason": "Both are Qwen/VL-related, but one is a from_pretrained weight_only error and the other is a video vision_position_ids regression." + }, + { + "left": "issue:34567", + "right": "issue:35707", + "accept": false, + "reason": "TrainerState token counting is unrelated to progressive generation with embeddings/past key values." + }, + { + "left": "issue:34634", + "right": "issue:41720", + "accept": false, + "reason": "BarkProcessor voice_preset and Qwen3 device-mapping are unrelated bugs." + }, + { + "left": "issue:43381", + "right": "issue:43844", + "accept": false, + "reason": "One is about checkpointing in eval mode; the other is a training-gradient anomaly with ZeRO-3." + }, + { + "left": "issue:42886", + "right": "issue:44843", + "accept": true, + "reason": "Same underlying offline tokenizer/cache regression: 42886 reports the symptom, while 44843 pinpoints the root cause in _patch_mistral_regex/model_info()." + }, + { + "left": "issue:43421", + "right": "issue:43653", + "accept": false, + "reason": "Different tokenizer issues: runtime special-token updates vs. mask token registration/decoding." + }, + { + "left": "issue:44077", + "right": "issue:44625", + "accept": false, + "reason": "Different problems: patchtsmixer post_init policy vs. Qwen3.5 num_labels propagation." + }, + { + "left": "issue:44451", + "right": "issue:44991", + "accept": false, + "reason": "Both are tokenizer loading failures, but they affect different models and don\u2019t show the same root cause." + }, + { + "left": "issue:44062", + "right": "issue:44451", + "accept": false, + "reason": "Different failures: AddedToken special-argument conflict vs. model loading failure for ScandiBERT." + }, + { + "left": "issue:44987", + "right": "issue:45042", + "accept": false, + "reason": "Unrelated: loading a specific model vs. PIL image processors wrongly requiring torchvision." + }, + { + "left": "issue:43531", + "right": "issue:44589", + "accept": false, + "reason": "Different issues: sliding_window behavior in Qwen3-MoE vs. missing Float8 storage type." + }, + { + "left": "issue:44918", + "right": "issue:44936", + "accept": false, + "reason": "Different trainer failures: unpacking Qwen3.5 embeddings in TRL SFT vs. evaluate() after train()." + }, + { + "left": "issue:43994", + "right": "issue:44671", + "accept": false, + "reason": "Both are wrong-model-output complaints, but they are model-specific and not the same bug." + }, + { + "left": "issue:39692", + "right": "issue:44610", + "accept": false, + "reason": "Both mention processor/model mismatch, but one is a documentation example with multiple errors and the other is a specific OmDet-Turbo image-size mismatch." + }, + { + "left": "issue:43421", + "right": "issue:44186", + "accept": false, + "reason": "Different tokenizer problems: post-processor updates vs. LayoutLMv2 NER/padding crashes." + }, + { + "left": "issue:45081", + "right": "issue:45356", + "accept": false, + "reason": "Both mention fix_mistral_regex/tokenizer regressions, but the symptoms and affected models differ enough that they do not clearly look like the same underlying bug." + }, + { + "left": "issue:43525", + "right": "issue:43531", + "accept": false, + "reason": "Llama4Config pad_token_id and Qwen3-MoE sliding_window are unrelated configuration bugs." + }, + { + "left": "issue:39692", + "right": "issue:43873", + "accept": false, + "reason": "Documentation/model-processor example errors are unrelated to offloading with quantization." + }, + { + "left": "issue:45183", + "right": "issue:45479", + "accept": false, + "reason": "Both are broad validation/config bugs, but one is about input_ids/inputs_embeds error messaging and the other is a degenerate loss issue for num_labels=1." + }, + { + "left": "issue:43746", + "right": "issue:45381", + "accept": false, + "reason": "Different Qwen/GraniteSpeech failures with no shared concrete code path." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "Grounded in the packet and conservative overall. The only accepted soft edge, issue:42886 \u2194 issue:44843, is well supported by the titles and the described offline/cache regression. The rejected edges are phrased cautiously and do not overclaim duplicate status." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 13, + "review_unit_index": 12, + "review_unit_count": 42, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-367", + "nodes": [ + "issue:30064", + "issue:36010", + "issue:40444", + "issue:43335", + "issue:43421", + "issue:43531", + "issue:43540", + "issue:43550", + "issue:43577", + "issue:43582", + "issue:43596", + "issue:43618", + "issue:43643", + "issue:43644", + "issue:43653", + "issue:43697", + "issue:43716", + "issue:43720", + "issue:43746", + "issue:43749", + "issue:43756", + "issue:43782", + "issue:43818", + "issue:43819", + "issue:43825", + "issue:43827", + "issue:43873", + "issue:43927", + "issue:43994", + "issue:44008", + "issue:44062", + "issue:44186", + "issue:44220", + "issue:44442", + "issue:44448", + "issue:44488", + "issue:44561", + "issue:44661", + "issue:44671", + "issue:44811", + "issue:44843", + "issue:44877", + "issue:44898", + "issue:44977", + "issue:45020", + "issue:45072", + "issue:45081", + "issue:45381" + ], + "soft_pairs": [ + "issue:43596|issue:44220", + "issue:43746|issue:43873", + "issue:43540|issue:44008", + "issue:43618|issue:43697", + "issue:40444|issue:43746", + "issue:44220|issue:44661", + "issue:43716|issue:44008", + "issue:43720|issue:43819", + "issue:43335|issue:43756", + "issue:43653|issue:43720", + "issue:44843|issue:45081", + "issue:43421|issue:44811", + "issue:44186|issue:44898", + "issue:30064|issue:36010", + "issue:43550|issue:45072", + "issue:43577|issue:43818", + "issue:44062|issue:44488", + "issue:43550|issue:43927", + "issue:43540|issue:45381", + "issue:43643|issue:44877", + "issue:43577|issue:43582", + "issue:43531|issue:43643", + "issue:43782|issue:44977", + "issue:43644|issue:43749", + "issue:44448|issue:44671", + "issue:43653|issue:43819", + "issue:44561|issue:45020", + "issue:43825|issue:43994", + "issue:43550|issue:43819", + "issue:44186|issue:44442", + "issue:43825|issue:43827" + ], + "prepared_review_unit_hash": "62c85041d11196b4d996beacf2ab6b3fb56dd5e40c61c64dd408d7752077b034", + "component_budget": { + "node_count": 367, + "item_count": 367, + "soft_pair_count": 1370, + "serialized_chars": 706986, + "estimated_input_tokens": 176747, + "estimated_eval_tokens": 353750 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 37219, + "estimated_input_tokens": 9305, + "estimated_eval_tokens": 18866 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is not a true duplicate set; it\u2019s a heterogeneous bucket of unrelated GitHub issues across tokenizers, multimodal models, loading/config paths, and v5 regressions. I rejected all soft edges because each pair only shares broad keywords or subsystem overlap, not the same concrete bug.", + "confidence": 0.97, + "canonical_issue_reason": "No single canonical issue stands out: the items span many unrelated failures, so there is no stable duplicate root to anchor the cluster.", + "canonical_pr_reason": null, + "best_issue_reason": "None of the issues is a good global representative; the similarity links are mostly superficial rather than evidence of one underlying defect.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43596", + "right": "issue:44220", + "accept": false, + "reason": "Both mention crashes, but one is DeepSpeed/BertModel init and the other is audio fbank feature extraction; different code paths and bugs." + }, + { + "left": "issue:43746", + "right": "issue:43873", + "accept": false, + "reason": "One is local checkpoint loading with PEFT adapters, the other is offloading under quantization; same broad loading area but not the same failure." + }, + { + "left": "issue:43540", + "right": "issue:44008", + "accept": false, + "reason": "Different multimodal models and different bugs: video input ValueError vs a Gemma3n variable-name collision in forward()." + }, + { + "left": "issue:43618", + "right": "issue:43697", + "accept": false, + "reason": "CLIP attentions assignment regression and RTDetrV2 output drift are unrelated model-specific issues." + }, + { + "left": "issue:40444", + "right": "issue:43746", + "accept": false, + "reason": "IterableDataset multi-image finetuning failure is unrelated to GraniteSpeech local checkpoint loading." + }, + { + "left": "issue:44220", + "right": "issue:44661", + "accept": false, + "reason": "Audio feature extraction failure and tokenizer-mapping/model addition failure are different subsystems." + }, + { + "left": "issue:43716", + "right": "issue:44008", + "accept": false, + "reason": "Mistral-3 image preprocessor dtype mismatch is not the same bug as Gemma3n\u2019s audio_mask attribute collision." + }, + { + "left": "issue:43720", + "right": "issue:43819", + "accept": false, + "reason": "BitNet packed-weight loading and DAC.from_latents/STE mismatch are distinct model implementations and defects." + }, + { + "left": "issue:43335", + "right": "issue:43756", + "accept": false, + "reason": "SwitchTransformers sparse-layer config bug and Smollm3 RoPE-layer dropping are unrelated configuration/model issues." + }, + { + "left": "issue:43653", + "right": "issue:43720", + "accept": false, + "reason": "Tokenizer special-token registration and BitNet accelerate loading are unrelated." + }, + { + "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 a backend_tokenizer AttributeError; different concrete failures." + }, + { + "left": "issue:43421", + "right": "issue:44811", + "accept": false, + "reason": "Runtime special-token/post-processor update behavior is not the same as Whisper batch_decode ignoring skip_special_tokens." + }, + { + "left": "issue:44186", + "right": "issue:44898", + "accept": false, + "reason": "LayoutLMv2 tokenizer padding/NER crash and Perceiver image-classification failure are unrelated." + }, + { + "left": "issue:30064", + "right": "issue:36010", + "accept": false, + "reason": "Void segmentation maps in image processing and a GenerationMixin import error have no shared code path." + }, + { + "left": "issue:43550", + "right": "issue:45072", + "accept": false, + "reason": "Bamba torch.compile with SDPA and SwitchTransformers/TimmWrapperModel dtype mismatches are different bugs." + }, + { + "left": "issue:43577", + "right": "issue:43818", + "accept": false, + "reason": "BLIP2 dtype propagation and Video-LLaVA temporal-attention/weight-sharing issues are unrelated." + }, + { + "left": "issue:44062", + "right": "issue:44488", + "accept": false, + "reason": "AddedToken duplicate special kwarg and failing to load a specific model are not the same defect." + }, + { + "left": "issue:43550", + "right": "issue:43927", + "accept": false, + "reason": "Bamba compile/SDPA failure and DiaConfig save/load token-id loss are unrelated." + }, + { + "left": "issue:43540", + "right": "issue:45381", + "accept": false, + "reason": "Both are video-related Qwen issues, but they affect different model versions and different bugs (input processing vs vision_position_ids)." + }, + { + "left": "issue:43643", + "right": "issue:44877", + "accept": false, + "reason": "AutoConfig trust_remote_code field loss and strict config blocking granite_speech loading are separate config-loading problems." + }, + { + "left": "issue:43577", + "right": "issue:43582", + "accept": false, + "reason": "BLIP2 dtype mismatch and Apple Silicon caching_allocator_warmup TypeError are unrelated." + }, + { + "left": "issue:43531", + "right": "issue:43643", + "accept": false, + "reason": "Qwen3-MoE sliding_window behavior and trust_remote_code AutoConfig field loss are different issues." + }, + { + "left": "issue:43782", + "right": "issue:44977", + "accept": false, + "reason": "Qwen3VL weight_only loading error and Qwen3.5 flash-attention generation failure are different model/path problems." + }, + { + "left": "issue:43644", + "right": "issue:43749", + "accept": false, + "reason": "Non-persistent buffer junk in Transformers 5.0 and FSDP CPU RAM efficient loading are distinct loading regressions." + }, + { + "left": "issue:44448", + "right": "issue:44671", + "accept": false, + "reason": "Pegasus v4/v5 output drift and CamemBERT masked-LM prediction regressions are unrelated." + }, + { + "left": "issue:43653", + "right": "issue:43819", + "accept": false, + "reason": "Tokenizer special-token registration and DAC STE mismatch are unrelated bugs." + }, + { + "left": "issue:44561", + "right": "issue:45020", + "accept": false, + "reason": "Both mention remote_code breakage, but one is specifically is_torch_fx_available removal and the other is a broad recent-version regression; not the same concrete bug." + }, + { + "left": "issue:43825", + "right": "issue:43994", + "accept": false, + "reason": "Pipeline error-message wording and Siglip2 nonsensical results are different problems." + }, + { + "left": "issue:43550", + "right": "issue:43819", + "accept": false, + "reason": "Bamba SDPA/torch.compile failure and DAC forward/from_latents mismatch are unrelated." + }, + { + "left": "issue:44186", + "right": "issue:44442", + "accept": false, + "reason": "LayoutLMv2 tokenizer crash and AutoTokenizer loading FastSpeech2ConformerTokenizer are different tokenizer issues." + }, + { + "left": "issue:43825", + "right": "issue:43827", + "accept": false, + "reason": "Both involve pipeline v5 messaging/docs, but one is a runtime error message and the other is stale documentation; not the same bug/change." + } + ] + }, + "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": 13, + "review_unit_index": 13, + "review_unit_count": 42, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-367", + "nodes": [ + "issue:30064", + "issue:34634", + "issue:36010", + "issue:36331", + "issue:41720", + "issue:42175", + "issue:43257", + "issue:43262", + "issue:43329", + "issue:43452", + "issue:43454", + "issue:43526", + "issue:43577", + "issue:43582", + "issue:43606", + "issue:43638", + "issue:43650", + "issue:43653", + "issue:43697", + "issue:43720", + "issue:43749", + "issue:43782", + "issue:43819", + "issue:43854", + "issue:43906", + "issue:44060", + "issue:44291", + "issue:44368", + "issue:44485", + "issue:44509", + "issue:44554", + "issue:44568", + "issue:44610", + "issue:44671", + "issue:44805", + "issue:44821", + "issue:44857", + "issue:44912", + "issue:44936", + "issue:44977", + "issue:44991", + "issue:45072", + "issue:45081", + "issue:45084", + "issue:45325", + "issue:45406", + "issue:45488", + "issue:45538" + ], + "soft_pairs": [ + "issue:44821|issue:45084", + "issue:43650|issue:43906", + "issue:41720|issue:44991", + "issue:43606|issue:45072", + "issue:44805|issue:44936", + "issue:30064|issue:36331", + "issue:44485|issue:44509", + "issue:42175|issue:45081", + "issue:44610|issue:45072", + "issue:43782|issue:45325", + "issue:34634|issue:36010", + "issue:43720|issue:45072", + "issue:44060|issue:44368", + "issue:45081|issue:45406", + "issue:43749|issue:44291", + "issue:43262|issue:43819", + "issue:43582|issue:44857", + "issue:43653|issue:44671", + "issue:43526|issue:43697", + "issue:43526|issue:44554", + "issue:43653|issue:44912", + "issue:45084|issue:45538", + "issue:43454|issue:43577", + "issue:34634|issue:36331", + "issue:44568|issue:45488", + "issue:43257|issue:43329", + "issue:43452|issue:44991", + "issue:43638|issue:44936", + "issue:43854|issue:44977" + ], + "prepared_review_unit_hash": "cafd1b8c75caa3610108444bef45fd54e7d117783b54cbc849ff18bf5d74ba44", + "component_budget": { + "node_count": 367, + "item_count": 367, + "soft_pair_count": 1370, + "serialized_chars": 706986, + "estimated_input_tokens": 176747, + "estimated_eval_tokens": 353750 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36843, + "estimated_input_tokens": 9211, + "estimated_eval_tokens": 18678 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a heterogeneous grab-bag of unrelated Transformers issues. None of the soft pairs look like the same underlying bug or change, so I would not merge any of them as duplicates. The best anchor issue is the bfloat16/device-mismatch report because it is the most central to the few loosely related loading/inference items, but it is still not a true duplicate of the others.", + "confidence": 0.43, + "canonical_issue_reason": "issue:45072 is the most cluster-central item because several other soft links loosely point at dtype/device loading/inference failures, but the cluster is too heterogeneous to justify a stronger duplicate canonical.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45072 has the broadest relevance to the few related runtime-mismatch reports (device/dtype/inference), making it the best representative anchor even though the edges are not true duplicates.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44821", + "right": "issue:45084", + "accept": false, + "reason": "AutoImageProcessor URL loading and a tokenizer template compilation error are unrelated failure modes." + }, + { + "left": "issue:43650", + "right": "issue:43906", + "accept": false, + "reason": "One is a vague data request; the other is a reproduction of a specific upstream bug. Not the same issue." + }, + { + "left": "issue:41720", + "right": "issue:44991", + "accept": false, + "reason": "Different subsystems and different failures: CUDA assert during Qwen3 device mapping vs tokenizer loading for est-roberta." + }, + { + "left": "issue:43606", + "right": "issue:45072", + "accept": false, + "reason": "Both mention mismatches, but they are different models and different concrete code paths; not a mergeable duplicate." + }, + { + "left": "issue:44805", + "right": "issue:44936", + "accept": false, + "reason": "A mask/indexing shape error is not the same as trainer.evaluate() failing after train()." + }, + { + "left": "issue:30064", + "right": "issue:36331", + "accept": false, + "reason": "Image processor segmentation-map handling and a custom trainer keyword-argument regression are unrelated." + }, + { + "left": "issue:44485", + "right": "issue:44509", + "accept": false, + "reason": "RoPE implementation discussion and stale docs about removed pipelines are unrelated." + }, + { + "left": "issue:42175", + "right": "issue:45081", + "accept": false, + "reason": "Packaging/backends availability and a Mistral regex patch crash are different problems." + }, + { + "left": "issue:44610", + "right": "issue:45072", + "accept": false, + "reason": "Processor input-size mismatch and dtype mismatch in inference are different concrete bugs." + }, + { + "left": "issue:43782", + "right": "issue:45325", + "accept": false, + "reason": "One is a from_pretrained weight_only load error; the other is a RoPE position-id scaling bug in a different Qwen VL model." + }, + { + "left": "issue:34634", + "right": "issue:36010", + "accept": false, + "reason": "Bark voice_preset behavior and a GenerationMixin import error are unrelated." + }, + { + "left": "issue:43720", + "right": "issue:45072", + "accept": false, + "reason": "Packed-weight unpacking during accelerate loading is a different failure from dtype mismatch in bfloat16 inference." + }, + { + "left": "issue:44060", + "right": "issue:44368", + "accept": false, + "reason": "Both mention tied-weights warnings, but one is an incorrect tying bug and the other is a config warning in fine-tuning." + }, + { + "left": "issue:45081", + "right": "issue:45406", + "accept": false, + "reason": "Mistral regex patching and Gemma4Processor missing _tokenizer are unrelated tokenizer/serve failures." + }, + { + "left": "issue:43749", + "right": "issue:44291", + "accept": false, + "reason": "Both concern low-memory loading, but one is broken FSDP CPU RAM efficient loading and the other is an unexpected constructor arg regression." + }, + { + "left": "issue:43262", + "right": "issue:43819", + "accept": false, + "reason": "Audio sampling-rate defaults and DAC latent reconstruction mismatch are unrelated." + }, + { + "left": "issue:43582", + "right": "issue:44857", + "accept": false, + "reason": "Apple Silicon warmup TypeError and CUDA AMP loss crash are different platform/code-path issues." + }, + { + "left": "issue:43653", + "right": "issue:44671", + "accept": false, + "reason": "Different tokenizer/model bugs: BigBird special-token registration vs CamemBERT masked LM predictions." + }, + { + "left": "issue:43526", + "right": "issue:43697", + "accept": false, + "reason": "Label reduction in BeitImageProcessorFast is not the same as RTDetrV2 output drift in v5." + }, + { + "left": "issue:43526", + "right": "issue:44554", + "accept": false, + "reason": "Image processor label handling and an MPS attention correctness issue are unrelated." + }, + { + "left": "issue:43653", + "right": "issue:44912", + "accept": false, + "reason": "Tokenizer special-token registration and MXFP4 quantization fallback are different bugs." + }, + { + "left": "issue:45084", + "right": "issue:45538", + "accept": false, + "reason": "Template-compilation failure and CLIPTokenizer max-length misconfiguration are not the same problem." + }, + { + "left": "issue:43454", + "right": "issue:43577", + "accept": false, + "reason": "AyaVision weight tying and Blip2 dtype casting are different multimodal loading bugs." + }, + { + "left": "issue:34634", + "right": "issue:36331", + "accept": false, + "reason": "BarkProcessor voice preset and CustomTrainer compute_loss signature are unrelated." + }, + { + "left": "issue:44568", + "right": "issue:45488", + "accept": false, + "reason": "Different tokenizer regressions affecting different models and behaviors; not the same code path." + }, + { + "left": "issue:43257", + "right": "issue:43329", + "accept": false, + "reason": "Qwen3 MOE loading with accelerate/deepspeed and multimodal token counting/video-branch undefined names are unrelated." + }, + { + "left": "issue:43452", + "right": "issue:44991", + "accept": false, + "reason": "gguf_file loading support and est-roberta tokenizer loading failure are different issues." + }, + { + "left": "issue:43638", + "right": "issue:44936", + "accept": false, + "reason": "Zero3/Bert training index error and post-train evaluation failure are unrelated." + }, + { + "left": "issue:43854", + "right": "issue:44977", + "accept": false, + "reason": "GLM-4.7 unit-test loading and Qwen3.5 flash-attention generation issues are different model-specific 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": 13, + "review_unit_index": 14, + "review_unit_count": 42, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-367", + "nodes": [ + "issue:33357", + "issue:34567", + "issue:36010", + "issue:36331", + "issue:38175", + "issue:39692", + "issue:40990", + "issue:41720", + "issue:42175", + "issue:42915", + "issue:43257", + "issue:43262", + "issue:43317", + "issue:43329", + "issue:43450", + "issue:43452", + "issue:43531", + "issue:43540", + "issue:43550", + "issue:43577", + "issue:43582", + "issue:43645", + "issue:43696", + "issue:43697", + "issue:43720", + "issue:43742", + "issue:43746", + "issue:43749", + "issue:43756", + "issue:43819", + "issue:43927", + "issue:43931", + "issue:44291", + "issue:44479", + "issue:44568", + "issue:44589", + "issue:44610", + "issue:44756", + "issue:44811", + "issue:44912", + "issue:44938", + "issue:44987", + "issue:45042", + "issue:45072", + "issue:45356", + "issue:45357", + "issue:45381", + "issue:45538" + ], + "soft_pairs": [ + "issue:43746|issue:44811", + "issue:44589|issue:45042", + "issue:38175|issue:44610", + "issue:39692|issue:44912", + "issue:43257|issue:43696", + "issue:43819|issue:43927", + "issue:43720|issue:44610", + "issue:43550|issue:44912", + "issue:36010|issue:43645", + "issue:44568|issue:45356", + "issue:43262|issue:44811", + "issue:43749|issue:44756", + "issue:43329|issue:44479", + "issue:43531|issue:45042", + "issue:43257|issue:43931", + "issue:36010|issue:36331", + "issue:43257|issue:43756", + "issue:43329|issue:43450", + "issue:34567|issue:40990", + "issue:43577|issue:45072", + "issue:43697|issue:45357", + "issue:42175|issue:44938", + "issue:43452|issue:43582", + "issue:41720|issue:43540", + "issue:44479|issue:45381", + "issue:43927|issue:44912", + "issue:44987|issue:45538", + "issue:42915|issue:43742", + "issue:44291|issue:45042", + "issue:43317|issue:45538", + "issue:33357|issue:34567" + ], + "prepared_review_unit_hash": "1177635d0dbb40e6d3b10aca178c2aade811f200e3da111666a61ee846954453", + "component_budget": { + "node_count": 367, + "item_count": 367, + "soft_pair_count": 1370, + "serialized_chars": 706986, + "estimated_input_tokens": 176747, + "estimated_eval_tokens": 353750 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 37117, + "estimated_input_tokens": 9280, + "estimated_eval_tokens": 18816 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These are mostly unrelated issues grouped by broad Transformers subsystems. I do not see a single true duplicate cluster, and none of the soft pairs are strong enough to merge as the same underlying bug/change.", + "confidence": 0.88, + "canonical_issue_reason": "No canonical issue: the set spans many distinct failures (tokenizers, multimodal/video processors, loading/quantization, hardware-specific crashes, and trainer behavior) rather than one shared root cause.", + "canonical_pr_reason": null, + "best_issue_reason": "No global best issue: there is no one issue that represents the whole set, and the pairwise candidates are not close enough to be treated as duplicates.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43746", + "right": "issue:44811", + "accept": false, + "reason": "Different surfaces and root causes: local checkpoint loading for GraniteSpeech adapters vs Whisper processor decoding behavior." + }, + { + "left": "issue:44589", + "right": "issue:45042", + "accept": false, + "reason": "Unrelated bugs: Float8 storage lookup failure vs PIL image processors incorrectly requiring torchvision." + }, + { + "left": "issue:38175", + "right": "issue:44610", + "accept": false, + "reason": "Both are model/processor complaints, but they concern different models and different failures (zero probabilities vs input-size mismatch)." + }, + { + "left": "issue:39692", + "right": "issue:44912", + "accept": false, + "reason": "Different models and different bugs: SigLIP2 doc example errors vs GPT-OSS MXFP4 quantization fallback/loading." + }, + { + "left": "issue:43257", + "right": "issue:43696", + "accept": false, + "reason": "Accelerate/DeepSpeed MoE weight conversion is not the same problem as a CUDA OOM on GPT-oss-20b." + }, + { + "left": "issue:43819", + "right": "issue:43927", + "accept": false, + "reason": "Different code paths: DAC latent reconstruction/STE mismatch vs DiaConfig losing custom token IDs on save/load." + }, + { + "left": "issue:43720", + "right": "issue:44610", + "accept": false, + "reason": "Packed-weight loading in BitNet is unrelated to OmDet-Turbo processor image-size output mismatch." + }, + { + "left": "issue:43550", + "right": "issue:44912", + "accept": false, + "reason": "torch.compile + SDPA failure in Bamba-9B-v2 is a different bug from MXFP4 quantization loading/fallback." + }, + { + "left": "issue:36010", + "right": "issue:43645", + "accept": false, + "reason": "Import path breakage for GenerationMixin is not the same as Jupyter custom-model initialization regressions." + }, + { + "left": "issue:44568", + "right": "issue:45356", + "accept": false, + "reason": "Both are tokenizer regressions, but they affect different models and different behaviors, so they are not the same bug." + }, + { + "left": "issue:43262", + "right": "issue:44811", + "accept": false, + "reason": "Audio chat-template sampling-rate defaulting is unrelated to Whisper batch_decode skip_special_tokens handling." + }, + { + "left": "issue:43749", + "right": "issue:44756", + "accept": false, + "reason": "FSDP CPU-RAM-efficient loading and Strix Halo mmap OOM are different loading/memory issues with no shared concrete root cause." + }, + { + "left": "issue:43329", + "right": "issue:44479", + "accept": false, + "reason": "Both involve video VLMs, but one is undefined names in a helper and the other is a broader regression in video input handling." + }, + { + "left": "issue:43531", + "right": "issue:45042", + "accept": false, + "reason": "Qwen3-MoE sliding_window behavior is unrelated to PIL backend image processors requiring torchvision." + }, + { + "left": "issue:43257", + "right": "issue:43931", + "accept": false, + "reason": "MoE weight conversion during accelerate/deepspeed loading is not the same as Qwen3-VL weight-shape mismatch on model load." + }, + { + "left": "issue:36010", + "right": "issue:36331", + "accept": false, + "reason": "GenerationMixin import failure and CustomTrainer.compute_loss signature mismatch are separate API breakages." + }, + { + "left": "issue:43257", + "right": "issue:43756", + "accept": false, + "reason": "Qwen3 MoE weight conversion and Smollm3 RoPE-layer count regression are unrelated model-specific issues." + }, + { + "left": "issue:43329", + "right": "issue:43450", + "accept": false, + "reason": "Undefined video-token helper names are not the same as batched video processor shape errors." + }, + { + "left": "issue:34567", + "right": "issue:40990", + "accept": false, + "reason": "TrainerState token-count tracking and GPT-oss perplexity are unrelated training/evaluation issues." + }, + { + "left": "issue:43577", + "right": "issue:45072", + "accept": false, + "reason": "BLIP2 dtype propagation and SwitchTransformers/TimmWrapper bfloat16 mismatches are different model-loading dtype bugs." + }, + { + "left": "issue:43697", + "right": "issue:45357", + "accept": false, + "reason": "RTDetrV2 output drift in v5 is unrelated to Qwen3.5 save_pretrained visual-encoder key regression." + }, + { + "left": "issue:42175", + "right": "issue:44938", + "accept": false, + "reason": "Missing TensorFlow in torch extra install and Python 3.14 import failure are different packaging/runtime problems." + }, + { + "left": "issue:43452", + "right": "issue:43582", + "accept": false, + "reason": "gguf_file tokenizer/model loading breakage is unrelated to Apple Silicon caching_allocator_warmup TypeError." + }, + { + "left": "issue:41720", + "right": "issue:43540", + "accept": false, + "reason": "Qwen3 auto device-map CUDA assert and Qwen3OmniMoe video ValueError are different failures and likely different code paths." + }, + { + "left": "issue:44479", + "right": "issue:45381", + "accept": false, + "reason": "Both are video-input regressions, but they mention different models and different symptoms, so they are not clearly the same bug." + }, + { + "left": "issue:43927", + "right": "issue:44912", + "accept": false, + "reason": "Config token-ID persistence and MXFP4 loading fallback are unrelated issues." + }, + { + "left": "issue:44987", + "right": "issue:45538", + "accept": false, + "reason": "Loading physical-intelligence/fast on v5.1.0+ is not the same as CLIPTokenizer model_max_length behavior." + }, + { + "left": "issue:42915", + "right": "issue:43742", + "accept": false, + "reason": "FineGrainedFP8Config failure for Qwen3Moe and a key error loading MobileLLM-125M are different model-load bugs." + }, + { + "left": "issue:44291", + "right": "issue:45042", + "accept": false, + "reason": "init_empty_weights/_is_hf_initialized TypeError is unrelated to the torchvision requirement regression." + }, + { + "left": "issue:43317", + "right": "issue:45538", + "accept": false, + "reason": "GPU+CPU offload loading failure is unrelated to CLIPTokenizer max-length semantics." + }, + { + "left": "issue:33357", + "right": "issue:34567", + "accept": false, + "reason": "MacOS bus error on CLIP model load and TrainerState token counting are completely different 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": 13, + "review_unit_index": 15, + "review_unit_count": 42, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-367", + "nodes": [ + "issue:28282", + "issue:29127", + "issue:34567", + "issue:34689", + "issue:36010", + "issue:36331", + "issue:38175", + "issue:42175", + "issue:42915", + "issue:43452", + "issue:43493", + "issue:43526", + "issue:43540", + "issue:43577", + "issue:43582", + "issue:43643", + "issue:43646", + "issue:43688", + "issue:43742", + "issue:43749", + "issue:43756", + "issue:43782", + "issue:43824", + "issue:43906", + "issue:43994", + "issue:44062", + "issue:44291", + "issue:44479", + "issue:44484", + "issue:44488", + "issue:44514", + "issue:44560", + "issue:44561", + "issue:44704", + "issue:44792", + "issue:44821", + "issue:44898", + "issue:44964", + "issue:44991", + "issue:45081", + "issue:45137", + "issue:45290", + "issue:45325", + "issue:45375", + "issue:45381", + "issue:45405", + "issue:45479", + "issue:45538" + ], + "soft_pairs": [ + "issue:43540|issue:44560", + "issue:43906|issue:45381", + "issue:43540|issue:44479", + "issue:45137|issue:45405", + "issue:42915|issue:45538", + "issue:42175|issue:43824", + "issue:42915|issue:44821", + "issue:43526|issue:45479", + "issue:42175|issue:44488", + "issue:44062|issue:44991", + "issue:38175|issue:44898", + "issue:28282|issue:34689", + "issue:43646|issue:43749", + "issue:43582|issue:43782", + "issue:43643|issue:45375", + "issue:43742|issue:44291", + "issue:43756|issue:45325", + "issue:43493|issue:43994", + "issue:28282|issue:29127", + "issue:36010|issue:42175", + "issue:44484|issue:44704", + "issue:34567|issue:36331", + "issue:44291|issue:44821", + "issue:43452|issue:44488", + "issue:44062|issue:45081", + "issue:44792|issue:44964", + "issue:43688|issue:43749", + "issue:44514|issue:45290", + "issue:43452|issue:43577", + "issue:43643|issue:44561" + ], + "prepared_review_unit_hash": "42e8fc472bfa84066a4901b9b63b663c1290e95776102c46d2c94ee706323b48", + "component_budget": { + "node_count": 367, + "item_count": 367, + "soft_pair_count": 1370, + "serialized_chars": 706986, + "estimated_input_tokens": 176747, + "estimated_eval_tokens": 353750 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 36890, + "estimated_input_tokens": 9223, + "estimated_eval_tokens": 18702 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Mostly unrelated issue reports; the only clear duplicate pair is the `apply_chat_template` crash with assistant tool calls and empty content. The rest are at best broad topical similarities, not the same concrete bug.", + "confidence": 0.58, + "canonical_issue_reason": "`issue:45290` is the cleanest canonical report: it states the core `apply_chat_template(tokenize=True)` crash with assistant tool calls and no content, while `issue:44514` looks like a narrower variant of the same failure.", + "canonical_pr_reason": null, + "best_issue_reason": "`issue:45290` is the best representative of the only strong duplicate subgroup in this set because it describes the underlying code-path bug most directly and broadly.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43540", + "right": "issue:44560", + "accept": false, + "reason": "Same broad Qwen video area, but different models and error shapes; not enough to treat as the same concrete bug." + }, + { + "left": "issue:43906", + "right": "issue:45381", + "accept": false, + "reason": "Different problems: isolated reproduction of an older issue versus a Qwen2.5-VL video position-id regression." + }, + { + "left": "issue:43540", + "right": "issue:44479", + "accept": false, + "reason": "Both mention video inputs, but one is Qwen3OmniMoe-specific and the other is a broader Qwen video regression; not clearly the same fixable bug." + }, + { + "left": "issue:45137", + "right": "issue:45405", + "accept": false, + "reason": "Completely different topics: DeepSpeed ZeRO3 deque error vs PEFT version pin bump." + }, + { + "left": "issue:42915", + "right": "issue:45538", + "accept": false, + "reason": "Unrelated: Qwen3Moe FP8 loading failure vs CLIPTokenizer max length behavior." + }, + { + "left": "issue:42175", + "right": "issue:43824", + "accept": false, + "reason": "Both are import/backend-related, but one is TensorFlow backend packaging and the other is a missing Qwen2_5_VL export." + }, + { + "left": "issue:42915", + "right": "issue:44821", + "accept": false, + "reason": "Different failures in different components; no shared code-path evidence." + }, + { + "left": "issue:43526", + "right": "issue:45479", + "accept": false, + "reason": "Different model families and bugs: BeitImageProcessorFast label reduction vs sequence-classification zero-loss behavior." + }, + { + "left": "issue:42175", + "right": "issue:44488", + "accept": false, + "reason": "Both are loading-related, but the specific failures and targets are unrelated." + }, + { + "left": "issue:44062", + "right": "issue:44991", + "accept": false, + "reason": "Both involve tokenizer loading, but one is a generic AddedToken TypeError and the other is a model-specific tokenizer load failure." + }, + { + "left": "issue:38175", + "right": "issue:44898", + "accept": false, + "reason": "Different model behaviors: SigLIP2 zero probabilities vs Perceiver non-default resolution failure." + }, + { + "left": "issue:28282", + "right": "issue:34689", + "accept": false, + "reason": "Both are import/load complaints, but about different models and different underlying causes." + }, + { + "left": "issue:43646", + "right": "issue:43749", + "accept": false, + "reason": "Different regression classes: custom model init breakage vs FSDP CPU RAM-efficient loading." + }, + { + "left": "issue:43582", + "right": "issue:43782", + "accept": false, + "reason": "Apple Silicon allocator TypeError and Qwen3VL weight_only loading error are unrelated." + }, + { + "left": "issue:43643", + "right": "issue:45375", + "accept": false, + "reason": "Both touch config handling, but one is generic remote-code field loss and the other is a specific missing config field." + }, + { + "left": "issue:43742", + "right": "issue:44291", + "accept": false, + "reason": "Different loading bugs: MobileLLM key error vs init_empty_weights/_is_hf_initialized TypeError." + }, + { + "left": "issue:43756", + "right": "issue:45325", + "accept": false, + "reason": "Both mention RoPE/position handling, but they concern different models and different bugs." + }, + { + "left": "issue:43493", + "right": "issue:43994", + "accept": false, + "reason": "Same model family, but one is an implementation discrepancy and the other is a bad AutoModel/pipeline result; too broad to merge." + }, + { + "left": "issue:28282", + "right": "issue:29127", + "accept": false, + "reason": "Both are error-message issues, but one is missing PyTorch import dependency and the other is LayoutLMv3 box-content validation." + }, + { + "left": "issue:36010", + "right": "issue:42175", + "accept": false, + "reason": "Both are import/version issues, but they concern different APIs and failure modes." + }, + { + "left": "issue:44484", + "right": "issue:44704", + "accept": false, + "reason": "Different API behaviors: save_pretrained shard sizing vs AutoProcessor kwargs passthrough." + }, + { + "left": "issue:34567", + "right": "issue:36331", + "accept": false, + "reason": "TrainerState token counting and custom trainer compute_loss signature are unrelated trainer issues." + }, + { + "left": "issue:44291", + "right": "issue:44821", + "accept": false, + "reason": "Different subsystems and failures: empty-weights init vs loading AutoImageProcessor from URL." + }, + { + "left": "issue:43452", + "right": "issue:44488", + "accept": false, + "reason": "Both are model-loading complaints, but one is a gguf_file/from_pretrained regression and the other is a specific model repo load failure." + }, + { + "left": "issue:44062", + "right": "issue:45081", + "accept": false, + "reason": "Tokenizer errors in both, but the root causes and affected paths differ." + }, + { + "left": "issue:44792", + "right": "issue:44964", + "accept": false, + "reason": "Different multimodal model failures; one is a failing janus test, the other is a Phi-4 multimodal load issue." + }, + { + "left": "issue:43688", + "right": "issue:43749", + "accept": false, + "reason": "Aux-loss normalization bug and FSDP loading bug are unrelated." + }, + { + "left": "issue:44514", + "right": "issue:45290", + "accept": true, + "reason": "Same underlying `apply_chat_template` crash with assistant tool calls and empty content; one is just a batched/padding variant." + }, + { + "left": "issue:43452", + "right": "issue:43577", + "accept": false, + "reason": "Both involve model loading, but the specific models and failure causes differ." + }, + { + "left": "issue:43643", + "right": "issue:44561", + "accept": false, + "reason": "Both are about `trust_remote_code`-adjacent loading, but one is missing fields in AutoConfig and the other is a removed helper breaking remote-code models." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is conservative and supported by the packet. The one accepted soft pair is the only clear duplicate subgroup indicated, and the other pairwise rejections stay within the evidence available from titles/metadata without overstating sameness." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 13, + "review_unit_index": 16, + "review_unit_count": 42, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-367", + "nodes": [ + "issue:24643", + "issue:28282", + "issue:29127", + "issue:32090", + "issue:33357", + "issue:35141", + "issue:38175", + "issue:41720", + "issue:41762", + "issue:43122", + "issue:43295", + "issue:43452", + "issue:43525", + "issue:43531", + "issue:43577", + "issue:43643", + "issue:43653", + "issue:43688", + "issue:43697", + "issue:43720", + "issue:43742", + "issue:43819", + "issue:43906", + "issue:44361", + "issue:44442", + "issue:44464", + "issue:44466", + "issue:44479", + "issue:44514", + "issue:44568", + "issue:44589", + "issue:44625", + "issue:44743", + "issue:44811", + "issue:44821", + "issue:44843", + "issue:44877", + "issue:44898", + "issue:44936", + "issue:44977", + "issue:44991", + "issue:45005", + "issue:45042", + "issue:45081", + "issue:45216", + "issue:45356", + "issue:45381", + "issue:45440" + ], + "soft_pairs": [ + "issue:43742|issue:44589", + "issue:44514|issue:45381", + "issue:43697|issue:45216", + "issue:45042|issue:45081", + "issue:43531|issue:44821", + "issue:44625|issue:44877", + "issue:43688|issue:45440", + "issue:43720|issue:44898", + "issue:43452|issue:44843", + "issue:43819|issue:44811", + "issue:43653|issue:44568", + "issue:44466|issue:45005", + "issue:43295|issue:45356", + "issue:44464|issue:44743", + "issue:41762|issue:44936", + "issue:24643|issue:29127", + "issue:41720|issue:43906", + "issue:43525|issue:43643", + "issue:44361|issue:44442", + "issue:43577|issue:43697", + "issue:38175|issue:43295", + "issue:43653|issue:44361", + "issue:44843|issue:44991", + "issue:28282|issue:32090", + "issue:28282|issue:35141", + "issue:44479|issue:44936", + "issue:28282|issue:33357", + "issue:43122|issue:43906", + "issue:44977|issue:45356" + ], + "prepared_review_unit_hash": "e1d30d83ca0947ed1fae1660f69c0d19a24ac41d938e48d521623eaa92f3df00", + "component_budget": { + "node_count": 367, + "item_count": 367, + "soft_pair_count": 1370, + "serialized_chars": 706986, + "estimated_input_tokens": 176747, + "estimated_eval_tokens": 353750 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36769, + "estimated_input_tokens": 9193, + "estimated_eval_tokens": 18642 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This batch is mostly a grab-bag of unrelated issue reports. I only see one plausible duplicate pair around tied-weights serialization in v5; the rest are different model families, different code paths, or different failure modes.", + "confidence": 0.83, + "canonical_issue_reason": "If a canonical issue must be chosen, issue 45005 is the best representative because it describes the broader tied-weights regression that also covers the more device-specific symptom in 44466.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 45005 is the strongest global representative in this set: it is the clearest, broadest description of the only likely duplicate cluster (tied weights in v5).", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43742", + "right": "issue:44589", + "accept": false, + "reason": "Different load failures and likely different root causes: MobileLLM key error vs Float8 storage lookup error." + }, + { + "left": "issue:44514", + "right": "issue:45381", + "accept": false, + "reason": "Both involve Qwen2.5-VL, but one is chat-template batching/padding and the other is video vision position IDs." + }, + { + "left": "issue:43697", + "right": "issue:45216", + "accept": false, + "reason": "Unrelated regressions: RTDetrV2 output mismatch vs Qwen3.5 checkpoint save_pretrained corruption." + }, + { + "left": "issue:45042", + "right": "issue:45081", + "accept": false, + "reason": "Different subsystems entirely: image processor torchvision dependency vs Mistral regex patch crash." + }, + { + "left": "issue:43531", + "right": "issue:44821", + "accept": false, + "reason": "Qwen3-MoE sliding-window bug is unrelated to AutoImageProcessor URL loading." + }, + { + "left": "issue:44625", + "right": "issue:44877", + "accept": false, + "reason": "Both are config-related, but they affect different models and different failures." + }, + { + "left": "issue:43688", + "right": "issue:45440", + "accept": false, + "reason": "Both mention MoE models, but one is auxiliary-loss normalization and the other is DeepSeekV3 implementation divergence." + }, + { + "left": "issue:43720", + "right": "issue:44898", + "accept": false, + "reason": "Different bug classes: unpacking packed weights during accelerate load vs Perceiver image classification at non-default resolution." + }, + { + "left": "issue:43452", + "right": "issue:44843", + "accept": false, + "reason": "Both are from tokenizer/model loading, but one is gguf_file handling and the other is an offline model_info() call in mistral regex patching." + }, + { + "left": "issue:43819", + "right": "issue:44811", + "accept": false, + "reason": "DAC latent reconstruction mismatch is unrelated to Whisper batch_decode skip_special_tokens handling." + }, + { + "left": "issue:43653", + "right": "issue:44568", + "accept": false, + "reason": "Both concern special tokens, but they are different tokenizers and different failures." + }, + { + "left": "issue:44466", + "right": "issue:45005", + "accept": true, + "reason": "Same underlying v5 tied-weights regression: 44466 is the device-dependent serialization symptom, while 45005 is the broader tied-weights report." + }, + { + "left": "issue:43295", + "right": "issue:45356", + "accept": false, + "reason": "Different regressions affecting different tokenizer/processor paths and different versions." + }, + { + "left": "issue:44464", + "right": "issue:44743", + "accept": false, + "reason": "Both mention generation/state behavior, but one is compiled-forward chunked generation and the other is recurrent-state reset in Qwen3.5 modular code." + }, + { + "left": "issue:41762", + "right": "issue:44936", + "accept": false, + "reason": "DeepSpeed ZeRO-3 load error is unrelated to trainer.evaluate() failing after trainer.train()." + }, + { + "left": "issue:24643", + "right": "issue:29127", + "accept": false, + "reason": "Different errors and code paths: DeepSpeed training weight shape vs LayoutLMV3 box-validation message clarity." + }, + { + "left": "issue:41720", + "right": "issue:43906", + "accept": false, + "reason": "No evidence of the same bug; the second is only an isolated reproduction of another issue." + }, + { + "left": "issue:43525", + "right": "issue:43643", + "accept": false, + "reason": "Related thematically, but one is a specific missing pad_token_id on Llama4Config and the other is a broader AutoConfig trust_remote_code field-loss bug." + }, + { + "left": "issue:44361", + "right": "issue:44442", + "accept": false, + "reason": "Different tokenizer failures with different models and symptoms." + }, + { + "left": "issue:43577", + "right": "issue:43697", + "accept": false, + "reason": "Blip2 dtype propagation is unrelated to RTDetrV2 output differences." + }, + { + "left": "issue:38175", + "right": "issue:43295", + "accept": false, + "reason": "Siglip2 zero probabilities are unrelated to the processor.tokenizer regression." + }, + { + "left": "issue:43653", + "right": "issue:44361", + "accept": false, + "reason": "Different tokenizer bugs affecting different tokenizers and failure modes." + }, + { + "left": "issue:44843", + "right": "issue:44991", + "accept": false, + "reason": "Both are tokenizer-loading regressions, but they appear to be separate root causes." + }, + { + "left": "issue:28282", + "right": "issue:32090", + "accept": false, + "reason": "ImportError for missing PyTorch and NoneType broadcast error are unrelated." + }, + { + "left": "issue:28282", + "right": "issue:35141", + "accept": false, + "reason": "No shared underlying bug: missing PyTorch vs tied embedding resize reinitialization." + }, + { + "left": "issue:28282", + "right": "issue:33357", + "accept": false, + "reason": "Completely different failures: missing backend dependency vs MacOS bus error on CLIP." + }, + { + "left": "issue:44479", + "right": "issue:44936", + "accept": false, + "reason": "Video-input regression in Qwen models is unrelated to trainer.evaluate() post-train failure." + }, + { + "left": "issue:43122", + "right": "issue:43906", + "accept": false, + "reason": "The isolated reproduction reference is too vague to establish the same underlying tokenization bug." + }, + { + "left": "issue:44977", + "right": "issue:45356", + "accept": false, + "reason": "Qwen3.5 flash-attention generation bug is unrelated to Kimi-K2.5 tokenizer codec handling." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet and stays conservative: it identifies one plausible duplicate pair (44466/45005) based on tightly related tied-weights v5 titles, and the other soft-edge rejections are appropriately cautious given only title-level evidence. No overstatement stands out." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 13, + "review_unit_index": 17, + "review_unit_count": 42, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-367", + "nodes": [ + "issue:30064", + "issue:32090", + "issue:33666", + "issue:34567", + "issue:36010", + "issue:41720", + "issue:43012", + "issue:43262", + "issue:43278", + "issue:43329", + "issue:43334", + "issue:43452", + "issue:43493", + "issue:43504", + "issue:43525", + "issue:43531", + "issue:43540", + "issue:43550", + "issue:43572", + "issue:43577", + "issue:43606", + "issue:43653", + "issue:43688", + "issue:43716", + "issue:43720", + "issue:43742", + "issue:43756", + "issue:43825", + "issue:43844", + "issue:43873", + "issue:43927", + "issue:44361", + "issue:44451", + "issue:44464", + "issue:44479", + "issue:44554", + "issue:44610", + "issue:44625", + "issue:44821", + "issue:44849", + "issue:44913", + "issue:44977", + "issue:45084", + "issue:45200", + "issue:45229", + "issue:45325", + "issue:45357" + ], + "soft_pairs": [ + "issue:30064|issue:33666", + "issue:43653|issue:43927", + "issue:44464|issue:44977", + "issue:43329|issue:45325", + "issue:43531|issue:43742", + "issue:43577|issue:44554", + "issue:44479|issue:44625", + "issue:43550|issue:44361", + "issue:43012|issue:43577", + "issue:43504|issue:43720", + "issue:41720|issue:44464", + "issue:43504|issue:43550", + "issue:32090|issue:33666", + "issue:32090|issue:34567", + "issue:43452|issue:44451", + "issue:43262|issue:45084", + "issue:43278|issue:43688", + "issue:43504|issue:44610", + "issue:45200|issue:45229", + "issue:44849|issue:45357", + "issue:43334|issue:43572", + "issue:43756|issue:44913", + "issue:43493|issue:43825", + "issue:43540|issue:43716", + "issue:43742|issue:44821", + "issue:43525|issue:43572", + "issue:33666|issue:36010", + "issue:43720|issue:43927", + "issue:43688|issue:43873", + "issue:43606|issue:43927", + "issue:43278|issue:43844" + ], + "prepared_review_unit_hash": "1d3922b93ccb8be819b125c6018785d50e044f1f45cd62514f578bfe41615d57", + "component_budget": { + "node_count": 367, + "item_count": 367, + "soft_pair_count": 1370, + "serialized_chars": 706986, + "estimated_input_tokens": 176747, + "estimated_eval_tokens": 353750 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 31, + "serialized_chars": 36418, + "estimated_input_tokens": 9105, + "estimated_eval_tokens": 18466 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These items are largely unrelated issue reports across different models, processors, loading paths, and training/runtime bugs; the soft pairs share only superficial thematic overlap, not the same underlying defect.", + "confidence": 0.96, + "canonical_issue_reason": "No single canonical duplicate stands out: the set is heterogeneous, spanning unrelated bugs in tokenizer/config loading, multimodal processing, training, quantization, and generation.", + "canonical_pr_reason": null, + "best_issue_reason": "If one representative issue must be chosen, issue 43873 is the broadest/high-activity report, but it is not a duplicate hub and should not be treated as the canonical merge target.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:30064", + "right": "issue:33666", + "accept": false, + "reason": "Different problems: void segmentation map handling vs Qwen2-VL multi-GPU training." + }, + { + "left": "issue:43653", + "right": "issue:43927", + "accept": false, + "reason": "Both involve token-related behavior, but one is BigBird special-token decode and the other is DiaConfig save/load token IDs; different bug." + }, + { + "left": "issue:44464", + "right": "issue:44977", + "accept": false, + "reason": "Both affect generation, but one is compiled-forward inconsistency and the other is flash-attention behavior in Qwen3.5." + }, + { + "left": "issue:43329", + "right": "issue:45325", + "accept": false, + "reason": "Both touch VLM multimodal internals, but one is an undefined video token-count path and the other is still-image rope indexing." + }, + { + "left": "issue:43531", + "right": "issue:43742", + "accept": false, + "reason": "Sliding-window MoE bug vs a model-loading key error; not the same code path." + }, + { + "left": "issue:43577", + "right": "issue:44554", + "accept": false, + "reason": "One is dtype propagation on load, the other is an MPS attention correctness issue." + }, + { + "left": "issue:44479", + "right": "issue:44625", + "accept": false, + "reason": "Video-input regression in multiple Qwen VL variants vs missing num_labels propagation in Qwen3.5." + }, + { + "left": "issue:43550", + "right": "issue:44361", + "accept": false, + "reason": "torch.compile + SDPA failure in Bamba is unrelated to MLukeTokenizer task AttributeError." + }, + { + "left": "issue:43012", + "right": "issue:43577", + "accept": false, + "reason": "Both concern precision/dtype, but one is a warning during bfloat16 compilation and the other is incorrect model dtype on load." + }, + { + "left": "issue:43504", + "right": "issue:43720", + "accept": false, + "reason": "Legacy-field loading bug in BEiT vs packed-weight unpacking bug in BitNet accelerate loading." + }, + { + "left": "issue:41720", + "right": "issue:44464", + "accept": false, + "reason": "CUDA assert under auto device mapping is unrelated to chunked generation divergence with compiled forward." + }, + { + "left": "issue:43504", + "right": "issue:43550", + "accept": false, + "reason": "Different models and failures: BEiT config loading vs Bamba torch.compile/SDPA." + }, + { + "left": "issue:32090", + "right": "issue:33666", + "accept": false, + "reason": "Trainer NoneType broadcast error is not the same as Qwen2-VL multi-GPU training." + }, + { + "left": "issue:32090", + "right": "issue:34567", + "accept": false, + "reason": "One is a distributed broadcast TypeError; the other is a TrainerState counter not updating." + }, + { + "left": "issue:43452", + "right": "issue:44451", + "accept": false, + "reason": "gguf_file path handling issue vs a generic inability to load a specific model." + }, + { + "left": "issue:43262", + "right": "issue:45084", + "accept": false, + "reason": "Audio sampling-rate default bug is unrelated to template-node compilation errors." + }, + { + "left": "issue:43278", + "right": "issue:43688", + "accept": false, + "reason": "Both mention training behavior, but one is dtype drift between train/eval and the other is auxiliary-loss normalization." + }, + { + "left": "issue:43504", + "right": "issue:44610", + "accept": false, + "reason": "Pretrained preset loading bug vs processor/model image-size mismatch." + }, + { + "left": "issue:45200", + "right": "issue:45229", + "accept": false, + "reason": "mm_token_type_ids defaulting issue is unrelated to multi-GPU inference OOM." + }, + { + "left": "issue:44849", + "right": "issue:45357", + "accept": false, + "reason": "Hidden-states bug in Qwen3.5 vs incorrect visual encoder keys saved on pretrained serialization." + }, + { + "left": "issue:43334", + "right": "issue:43572", + "accept": false, + "reason": "Both are missing-config-field regressions, but for different models, different fields, and different failure modes." + }, + { + "left": "issue:43756", + "right": "issue:44913", + "accept": false, + "reason": "RoPE-layer configuration bug vs rotary_pct not persisting on reload; similar theme, different defect." + }, + { + "left": "issue:43493", + "right": "issue:43825", + "accept": false, + "reason": "SigLIP2 implementation discrepancy is unrelated to pipeline task support messaging." + }, + { + "left": "issue:43540", + "right": "issue:43716", + "accept": false, + "reason": "Video-processing ValueError in Qwen3OmniMoe vs image-preprocessor/model dtype mismatch in Mistral-3." + }, + { + "left": "issue:43742", + "right": "issue:44821", + "accept": false, + "reason": "Model key error on load vs AutoImageProcessor URL loading; different subsystems." + }, + { + "left": "issue:43525", + "right": "issue:43572", + "accept": false, + "reason": "Both mention missing pad token fields, but they affect different configs and attributes." + }, + { + "left": "issue:33666", + "right": "issue:36010", + "accept": false, + "reason": "Qwen2-VL multi-GPU training is unrelated to GenerationMixin import failure." + }, + { + "left": "issue:43720", + "right": "issue:43927", + "accept": false, + "reason": "Packed-weight loading bug vs config token-ID loss on save/load." + }, + { + "left": "issue:43688", + "right": "issue:43873", + "accept": false, + "reason": "Aux-loss normalization is unrelated to offloading with quantization." + }, + { + "left": "issue:43606", + "right": "issue:43927", + "accept": false, + "reason": "CPU offload device mismatch is unrelated to DiaConfig token-ID persistence." + }, + { + "left": "issue:43278", + "right": "issue:43844", + "accept": false, + "reason": "Embedding dtype drift vs abnormal gradient growth under DeepSpeed ZeRO-3 are different 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": 13, + "review_unit_index": 18, + "review_unit_count": 42, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-367", + "nodes": [ + "issue:33357", + "issue:33666", + "issue:36331", + "issue:39692", + "issue:43257", + "issue:43278", + "issue:43295", + "issue:43299", + "issue:43329", + "issue:43493", + "issue:43504", + "issue:43540", + "issue:43572", + "issue:43575", + "issue:43606", + "issue:43646", + "issue:43653", + "issue:43688", + "issue:43701", + "issue:43720", + "issue:43746", + "issue:43756", + "issue:43824", + "issue:43844", + "issue:43927", + "issue:43994", + "issue:44060", + "issue:44112", + "issue:44291", + "issue:44361", + "issue:44423", + "issue:44464", + "issue:44466", + "issue:44488", + "issue:44493", + "issue:44521", + "issue:44589", + "issue:44625", + "issue:44734", + "issue:44792", + "issue:44843", + "issue:44964", + "issue:44977", + "issue:45290", + "issue:45405", + "issue:45520", + "issue:45584" + ], + "soft_pairs": [ + "issue:43257|issue:43299", + "issue:43720|issue:44361", + "issue:39692|issue:43493", + "issue:44964|issue:45405", + "issue:43540|issue:44977", + "issue:43329|issue:43606", + "issue:44488|issue:44843", + "issue:33666|issue:36331", + "issue:43493|issue:43844", + "issue:43756|issue:44060", + "issue:44589|issue:45520", + "issue:44625|issue:45584", + "issue:43646|issue:44466", + "issue:43688|issue:43746", + "issue:44521|issue:45290", + "issue:44423|issue:44734", + "issue:43824|issue:44589", + "issue:33357|issue:33666", + "issue:39692|issue:43295", + "issue:43504|issue:43606", + "issue:43994|issue:44493", + "issue:44464|issue:45584", + "issue:43927|issue:44361", + "issue:43572|issue:44291", + "issue:43504|issue:43653", + "issue:43575|issue:45405", + "issue:43278|issue:43701", + "issue:44792|issue:45405", + "issue:43257|issue:45405", + "issue:43606|issue:44112" + ], + "prepared_review_unit_hash": "01bf351f947c48ffb3ef4dfa6fcf840806a16ec36dc1b01d62ade80582030c3b", + "component_budget": { + "node_count": 367, + "item_count": 367, + "soft_pair_count": 1370, + "serialized_chars": 706986, + "estimated_input_tokens": 176747, + "estimated_eval_tokens": 353750 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 30, + "serialized_chars": 36407, + "estimated_input_tokens": 9102, + "estimated_eval_tokens": 18460 + }, + "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-loading, multimodal, tokenizer, training, and CI regressions. The soft pairs share broad subsystem keywords at best, but not the same concrete bug/change, so none should be merged as duplicates.", + "confidence": 0.19, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43257", + "right": "issue:43299", + "accept": false, + "reason": "Both involve Qwen3 MOE loading, but one is about missing weight conversion under accelerate/deepspeed and the other is a version regression in Qwen3VL model loading; different failure modes and code paths." + }, + { + "left": "issue:43720", + "right": "issue:44361", + "accept": false, + "reason": "One is a packed-weight loading bug in BitNet/accelerate; the other is an AttributeError in MLukeTokenizer task handling. No shared concrete defect." + }, + { + "left": "issue:39692", + "right": "issue:43493", + "accept": false, + "reason": "Both mention SigLIP2, but one is a documentation/example mismatch while the other is an implementation-vs-JAX discrepancy in the model itself." + }, + { + "left": "issue:44964", + "right": "issue:45405", + "accept": false, + "reason": "A model loading regression is not the same as a dependency version bump issue; one is about loading Phi-4 multimodal, the other about an unreleased MIN_PEFT_VERSION." + }, + { + "left": "issue:43540", + "right": "issue:44977", + "accept": false, + "reason": "Both are Qwen3.5-related, but one is video-input processing in Qwen3OmniMoe and the other is flash-attention generation behavior; distinct bugs." + }, + { + "left": "issue:43329", + "right": "issue:43606", + "accept": false, + "reason": "Undefined variables in multimodal token counting is unrelated to a CPU-offload device mismatch in bark-small." + }, + { + "left": "issue:44488", + "right": "issue:44843", + "accept": false, + "reason": "One is a model loading failure for a specific checkpoint, the other is offline-mode breakage from an unconditional hub metadata call; different root causes." + }, + { + "left": "issue:33666", + "right": "issue:36331", + "accept": false, + "reason": "Multi-GPU training support and a custom trainer compute_loss signature mismatch are separate training issues." + }, + { + "left": "issue:43493", + "right": "issue:43844", + "accept": false, + "reason": "Both are correctness bugs, but they affect different models and mechanisms: SigLIP2 implementation mismatch versus gradient inflation under ZeRO-3." + }, + { + "left": "issue:43756", + "right": "issue:44060", + "accept": false, + "reason": "Smollm3 RoPE-layer truncation and Qwen3-Next tied-weights warnings are unrelated model-specific configuration bugs." + }, + { + "left": "issue:44589", + "right": "issue:45520", + "accept": false, + "reason": "Both mention missing storage/dispatch paths, but one is a Float8 storage lookup failure and the other is a flash_attn import_utils KeyError on Python 3.13." + }, + { + "left": "issue:44625", + "right": "issue:45584", + "accept": false, + "reason": "num_labels propagation in Qwen3.5 is unrelated to Whisper empty-transcription generation after align_special_tokens." + }, + { + "left": "issue:43646", + "right": "issue:44466", + "accept": false, + "reason": "Both are v5 regressions, but one is custom model initialization and the other is inconsistent lm_head serialization; different code paths." + }, + { + "left": "issue:43688", + "right": "issue:43746", + "accept": false, + "reason": "Auxiliary-loss normalization in MoE models is not the same as GraniteSpeech PEFT adapter loading from local checkpoints." + }, + { + "left": "issue:44521", + "right": "issue:45290", + "accept": false, + "reason": "Both concern apply_chat_template, but one is all-zero assistant masks for multimodal inputs and the other is tool-call messages with no content. Different bugs." + }, + { + "left": "issue:44423", + "right": "issue:44734", + "accept": false, + "reason": "Both hit serve/runtime multimodal paths, but one crashes on string-to-.to() handling and the other on KV-cache continuation tensor indexing." + }, + { + "left": "issue:43824", + "right": "issue:44589", + "accept": false, + "reason": "ImportError for a renamed model class is unrelated to a missing Float8 storage object during serialization/loading." + }, + { + "left": "issue:33357", + "right": "issue:33666", + "accept": false, + "reason": "A MacOS bus error loading CLIP is unrelated to Qwen2-VL multi-GPU training." + }, + { + "left": "issue:39692", + "right": "issue:43295", + "accept": false, + "reason": "A broken SigLIP2 example in docs is not the same as a regression in processor.tokenizer access and image-to-tokenizer passing." + }, + { + "left": "issue:43504", + "right": "issue:43606", + "accept": false, + "reason": "A legacy field preventing BEiT semantic-segmentation loading is unrelated to a bark-small CPU offload device mismatch." + }, + { + "left": "issue:43994", + "right": "issue:44493", + "accept": false, + "reason": "Both are model-quality/loading complaints, but one is SigLIP2 nonsensical outputs and the other is unexpected position-id keys across many models." + }, + { + "left": "issue:44464", + "right": "issue:45584", + "accept": false, + "reason": "Chunked generation inconsistency with compiled forward is unrelated to Whisper empty-transcription failure after token alignment." + }, + { + "left": "issue:43927", + "right": "issue:44361", + "accept": false, + "reason": "DiaConfig save/load token ID loss and MLukeTokenizer task AttributeError are different layers and failure modes." + }, + { + "left": "issue:43572", + "right": "issue:44291", + "accept": false, + "reason": "Both involve initialization/loading changes in v5, but missing pad_token_idx in StableLmConfig is not the same as an unexpected _is_hf_initialized argument in init_empty_weights." + }, + { + "left": "issue:43504", + "right": "issue:43653", + "accept": false, + "reason": "BEiT legacy-field loading and BigBirdTokenizer special-token registration are unrelated bugs." + }, + { + "left": "issue:43575", + "right": "issue:45405", + "accept": false, + "reason": "TP OOM for a large Qwen2 model is unrelated to a dependency version bump for PEFT." + }, + { + "left": "issue:43278", + "right": "issue:43701", + "accept": false, + "reason": "Embedding dtype drift between train/eval is unrelated to resume_from_checkpoint key mismatches." + }, + { + "left": "issue:44792", + "right": "issue:45405", + "accept": false, + "reason": "A failing janus image-generation test does not match a PEFT minimum-version issue." + }, + { + "left": "issue:43257", + "right": "issue:45405", + "accept": false, + "reason": "Qwen3 MOE weight conversion/loading and a PEFT version requirement are unrelated." + }, + { + "left": "issue:43606", + "right": "issue:44112", + "accept": false, + "reason": "Both touch GraniteSpeech/device handling, but one is bark-small CPU offload mismatch and the other is a stale CI test override. Not the same defect." + } + ] + }, + "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": 13, + "review_unit_index": 19, + "review_unit_count": 42, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-367", + "nodes": [ + "issue:15354", + "issue:37428", + "issue:38175", + "issue:39401", + "issue:43232", + "issue:43295", + "issue:43334", + "issue:43388", + "issue:43450", + "issue:43493", + "issue:43525", + "issue:43526", + "issue:43540", + "issue:43572", + "issue:43582", + "issue:43701", + "issue:43824", + "issue:43844", + "issue:43873", + "issue:43957", + "issue:43975", + "issue:44315", + "issue:44464", + "issue:44466", + "issue:44514", + "issue:44521", + "issue:44554", + "issue:44568", + "issue:44589", + "issue:44610", + "issue:44792", + "issue:44849", + "issue:44898", + "issue:44936", + "issue:44964", + "issue:44977", + "issue:44987", + "issue:45072", + "issue:45084", + "issue:45127", + "issue:45198", + "issue:45200", + "issue:45325", + "issue:45356", + "issue:45412", + "issue:45446", + "issue:45520", + "issue:45563" + ], + "soft_pairs": [ + "issue:15354|issue:38175", + "issue:44554|issue:45446", + "issue:45084|issue:45520", + "issue:43582|issue:44977", + "issue:43540|issue:45200", + "issue:43844|issue:44964", + "issue:45072|issue:45198", + "issue:43232|issue:44849", + "issue:39401|issue:44568", + "issue:37428|issue:43957", + "issue:44315|issue:44849", + "issue:43334|issue:43525", + "issue:44554|issue:45356", + "issue:43295|issue:44610", + "issue:43493|issue:43701", + "issue:45200|issue:45325", + "issue:44466|issue:45127", + "issue:44898|issue:45412", + "issue:43873|issue:44464", + "issue:44792|issue:45563", + "issue:43388|issue:43526", + "issue:43844|issue:44936", + "issue:44514|issue:44521", + "issue:37428|issue:43824", + "issue:43572|issue:44987", + "issue:43295|issue:43450", + "issue:43572|issue:44589", + "issue:43824|issue:43975", + "issue:43295|issue:43873" + ], + "prepared_review_unit_hash": "0a150c296e824a0b32e894eabaf61d94e42bc53278af1318b794133910c7a4e7", + "component_budget": { + "node_count": 367, + "item_count": 367, + "soft_pair_count": 1370, + "serialized_chars": 706986, + "estimated_input_tokens": 176747, + "estimated_eval_tokens": 353750 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36626, + "estimated_input_tokens": 9157, + "estimated_eval_tokens": 18570 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly heterogeneous: several unrelated Transformers regressions are mixed together. The only clear duplicate-style pair is the missing `pad_token_id` config AttributeError; most other soft edges are similar only at a broad subsystem level and should stay separate.", + "confidence": 0.38, + "canonical_issue_reason": "#43334 is the clearest representative of the `pad_token_id`-missing config regression and is earlier/more discussed than its sibling.", + "canonical_pr_reason": null, + "best_issue_reason": "#43334 is the best overall issue to anchor the small `pad_token_id` subcluster; it states the failure mode cleanly and matches the only strong duplicate candidate.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:15354", + "right": "issue:38175", + "accept": false, + "reason": "Different models and failures: TorchScript GeneratorExp export vs SigLIP2 zero-probability output." + }, + { + "left": "issue:44554", + "right": "issue:45446", + "accept": false, + "reason": "Both touch attention/flex attention areas, but one is an MPS correctness bug and the other is a version-check import bug." + }, + { + "left": "issue:45084", + "right": "issue:45520", + "accept": false, + "reason": "Unrelated import/compile failures with no shared concrete code path." + }, + { + "left": "issue:43582", + "right": "issue:44977", + "accept": false, + "reason": "AppleSilicon allocator warmup TypeError and Qwen3.5 flash-attention generation issue are different bugs." + }, + { + "left": "issue:43540", + "right": "issue:45200", + "accept": false, + "reason": "Both are multimodal/model-processing issues, but one is video input handling and the other is Gemma 4 token-type defaults." + }, + { + "left": "issue:43844", + "right": "issue:44964", + "accept": false, + "reason": "DeepSpeed/ZeRO gradient instability is unrelated to Phi-4 multimodal model loading." + }, + { + "left": "issue:45072", + "right": "issue:45198", + "accept": false, + "reason": "Different subsystems: dtype mismatch in inference vs Wav2Vec2 save/tokenization failure." + }, + { + "left": "issue:43232", + "right": "issue:44849", + "accept": false, + "reason": "Generation kwargs syncing and Qwen3.5 hidden-state output are separate generation behaviors." + }, + { + "left": "issue:39401", + "right": "issue:44568", + "accept": false, + "reason": "Both are tokenizer bugs, but they affect different tokenizers and different behavior (`offset_mapping` vs special-token insertion)." + }, + { + "left": "issue:37428", + "right": "issue:43957", + "accept": false, + "reason": "Flash-attention import failure and meta-device model loading breakage are distinct issues." + }, + { + "left": "issue:44315", + "right": "issue:44849", + "accept": false, + "reason": "Liger Kernel application in `model_init` is unrelated to Qwen3.5 hidden-state output behavior." + }, + { + "left": "issue:43334", + "right": "issue:43525", + "accept": true, + "reason": "Both report the same `AttributeError` caused by config objects missing `pad_token_id`; likely one fix addresses both." + }, + { + "left": "issue:44554", + "right": "issue:45356", + "accept": false, + "reason": "Different problems: MPS attention correctness vs Kimi-K2.5 tokenizer codec/regex regression." + }, + { + "left": "issue:43295", + "right": "issue:44610", + "accept": false, + "reason": "Processor/tokenizer API regression and model loading with `torch.device(\"meta\")` are not the same bug." + }, + { + "left": "issue:43493", + "right": "issue:43701", + "accept": false, + "reason": "SigLIP2 implementation discrepancy and resume checkpoint key mismatch do not share a concrete failure path." + }, + { + "left": "issue:45200", + "right": "issue:45325", + "accept": false, + "reason": "Both are Qwen2.5-VL issues, but one is token-type defaults and the other is temporal position-id scaling." + }, + { + "left": "issue:44466", + "right": "issue:45127", + "accept": false, + "reason": "Serialization of tied weights and LoRA merge collapse are related only broadly; the concrete bugs differ." + }, + { + "left": "issue:44898", + "right": "issue:45412", + "accept": false, + "reason": "Perceiver resolution handling and RT-DETR memory release are unrelated." + }, + { + "left": "issue:43873", + "right": "issue:44464", + "accept": false, + "reason": "Quantized offloading and compiled-forward chunked generation are different behaviors." + }, + { + "left": "issue:44792", + "right": "issue:45563", + "accept": false, + "reason": "Janus image-generation test failure and stale paged-generate warning are not the same bug." + }, + { + "left": "issue:43388", + "right": "issue:43526", + "accept": false, + "reason": "Both involve labels, but one is `gather_for_metrics` dropping tuple elements and the other is `BeitImageProcessorFast` reduce_labels truncation." + }, + { + "left": "issue:43844", + "right": "issue:44936", + "accept": false, + "reason": "DeepSpeed gradient growth and trainer evaluate-after-train failure are separate training issues." + }, + { + "left": "issue:44514", + "right": "issue:44521", + "accept": false, + "reason": "Both involve `apply_chat_template`, but one is a batched padding=False crash and the other is incorrect assistant masks for multimodal inputs." + }, + { + "left": "issue:37428", + "right": "issue:43824", + "accept": false, + "reason": "Both are import errors, but they concern different missing symbols and different code paths." + }, + { + "left": "issue:43572", + "right": "issue:44987", + "accept": false, + "reason": "StableLm config `pad_token_idx` regression and loading `physical-intelligence/fast` under v5.1 are not the same underlying bug." + }, + { + "left": "issue:43295", + "right": "issue:43873", + "accept": false, + "reason": "Processor/tokenizer regression and quantized offloading failure do not look like one fixable code path." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The assessment is grounded in the packet and stays conservative. The pad_token_id pair (43334/43525) is a defensible duplicate-style match, and the remaining soft-edge rejections do not overclaim shared evidence. No PR-pair mergeability issue is present here." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 13, + "review_unit_index": 20, + "review_unit_count": 42, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-367", + "nodes": [ + "issue:29942", + "issue:32090", + "issue:37428", + "issue:38175", + "issue:42915", + "issue:43262", + "issue:43278", + "issue:43295", + "issue:43425", + "issue:43450", + "issue:43493", + "issue:43575", + "issue:43606", + "issue:43611", + "issue:43653", + "issue:43819", + "issue:43824", + "issue:43825", + "issue:43827", + "issue:43856", + "issue:43901", + "issue:43931", + "issue:43994", + "issue:44112", + "issue:44206", + "issue:44360", + "issue:44442", + "issue:44448", + "issue:44462", + "issue:44561", + "issue:44568", + "issue:44610", + "issue:44743", + "issue:44779", + "issue:44849", + "issue:44857", + "issue:44991", + "issue:45003", + "issue:45083", + "issue:45127", + "issue:45198", + "issue:45245", + "issue:45290", + "issue:45310", + "issue:45381", + "issue:45412", + "issue:45478", + "issue:45584" + ], + "soft_pairs": [ + "issue:44442|issue:45198", + "issue:43278|issue:43856", + "issue:44360|issue:44561", + "issue:38175|issue:45412", + "issue:43295|issue:43994", + "issue:44206|issue:45584", + "issue:38175|issue:43493", + "issue:43827|issue:43901", + "issue:44849|issue:45003", + "issue:42915|issue:43824", + "issue:43653|issue:44112", + "issue:43450|issue:43493", + "issue:44568|issue:44991", + "issue:45290|issue:45381", + "issue:45127|issue:45245", + "issue:43819|issue:45083", + "issue:43450|issue:43575", + "issue:44462|issue:44561", + "issue:44561|issue:44779", + "issue:44610|issue:44857", + "issue:43262|issue:43450", + "issue:43819|issue:44112", + "issue:44448|issue:44743", + "issue:43606|issue:45083", + "issue:43278|issue:43825", + "issue:29942|issue:32090", + "issue:37428|issue:43425", + "issue:45310|issue:45478", + "issue:43611|issue:43931" + ], + "prepared_review_unit_hash": "a42d36626f56171845c0af13dba9ccba7ec9265319ea73dabb85ca7ff1916be2", + "component_budget": { + "node_count": 367, + "item_count": 367, + "soft_pair_count": 1370, + "serialized_chars": 706986, + "estimated_input_tokens": 176747, + "estimated_eval_tokens": 353750 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36310, + "estimated_input_tokens": 9078, + "estimated_eval_tokens": 18412 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a grab-bag of unrelated transformer/model issues. The only clear exact duplicate pair is the Qwen3.5 MoE from_pretrained regression, and the SigLIP2 pair is the only other plausible same-bug match.", + "confidence": 0.71, + "canonical_issue_reason": "issue:45310 is the best canonical issue: it is the earliest report of the exact Qwen3.5 MoE from_pretrained regression on transformers>=5.4.0, and issue:45478 reads like a later duplicate/follow-up.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45310 is the strongest global issue anchor because it names a concrete version-triggered regression, has the same failure mode as its duplicate, and is the most representative dedupe target in this set.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44442", + "right": "issue:45198", + "accept": false, + "reason": "Both are tokenizer/processor bugs, but they affect different models and different failure modes; not the same underlying defect." + }, + { + "left": "issue:43278", + "right": "issue:43856", + "accept": false, + "reason": "Training dtype mismatch and MoE memory inefficiency are different bugs with no shared concrete code-path." + }, + { + "left": "issue:44360", + "right": "issue:44561", + "accept": false, + "reason": "DSA indexer/ReLU discussion is unrelated to the v5 trust_remote_code import breakage." + }, + { + "left": "issue:38175", + "right": "issue:45412", + "accept": false, + "reason": "SigLIP2 zero-probability output and RT-DETR memory leak are unrelated model behaviors." + }, + { + "left": "issue:43295", + "right": "issue:43994", + "accept": false, + "reason": "Both mention processor/model usage, but they are different regressions on different models and APIs." + }, + { + "left": "issue:44206", + "right": "issue:45584", + "accept": false, + "reason": "Audio feature-extractor crash on unsupported args is not the same bug as Whisper empty-transcription generation failure." + }, + { + "left": "issue:38175", + "right": "issue:43493", + "accept": true, + "reason": "Both are about SigLIP2 producing wrong outputs versus expected behavior/reference, so they look like the same underlying model implementation bug." + }, + { + "left": "issue:43827", + "right": "issue:43901", + "accept": false, + "reason": "Both are docs-related, but they refer to different deprecated/changed pipeline behaviors and are not the same concrete issue." + }, + { + "left": "issue:44849", + "right": "issue:45003", + "accept": false, + "reason": "Qwen3.5 hidden-states behavior and unsafe sys.modules access are different code paths and symptoms." + }, + { + "left": "issue:42915", + "right": "issue:43824", + "accept": false, + "reason": "Qwen3 MoE FP8 training failure and Qwen2.5-VL import error are unrelated." + }, + { + "left": "issue:43653", + "right": "issue:44112", + "accept": false, + "reason": "Tokenizer special-token registration and GraniteSpeech CI device-override failure are different bugs." + }, + { + "left": "issue:43450", + "right": "issue:43493", + "accept": false, + "reason": "Video batch-shape bug is unrelated to the SigLIP2 output discrepancy." + }, + { + "left": "issue:44568", + "right": "issue:44991", + "accept": false, + "reason": "Different tokenizer regressions affecting different models; no evidence they share the same root cause." + }, + { + "left": "issue:45290", + "right": "issue:45381", + "accept": false, + "reason": "Chat-template tool-call crash and Qwen2.5-VL video position-id bug are distinct failures." + }, + { + "left": "issue:45127", + "right": "issue:45245", + "accept": false, + "reason": "LoRA/tied-embeddings collapse and category-count overflow are unrelated." + }, + { + "left": "issue:43819", + "right": "issue:45083", + "accept": false, + "reason": "DAC latent/forward mismatch and Qwen3 Omni feature-length helper behavior are different defects." + }, + { + "left": "issue:43450", + "right": "issue:43575", + "accept": false, + "reason": "Video processor shape handling and tp OOM on Qwen2-57B are unrelated issues." + }, + { + "left": "issue:44462", + "right": "issue:44561", + "accept": false, + "reason": "Tokenizer.json being ignored is a tokenizer-loading bug, while is_torch_fx_available removal affects trust_remote_code imports; different root causes." + }, + { + "left": "issue:44561", + "right": "issue:44779", + "accept": false, + "reason": "Both involve v5 regressions, but one is an API removal/import break and the other is tokenizer output correctness; not the same bug." + }, + { + "left": "issue:44610", + "right": "issue:44857", + "accept": false, + "reason": "Processor input-size mismatch and AMP/CUDA loss crash are different model/training paths." + }, + { + "left": "issue:43262", + "right": "issue:43450", + "accept": false, + "reason": "Audio chat-template sample-rate defaulting and batched video processor shape handling are unrelated modality bugs." + }, + { + "left": "issue:43819", + "right": "issue:44112", + "accept": false, + "reason": "DAC latent mismatch and GraniteSpeech CI device override failure are not the same underlying issue." + }, + { + "left": "issue:44448", + "right": "issue:44743", + "accept": false, + "reason": "Pegasus v4/v5 output difference and Qwen3.5 recurrent-state reset bug are different model regressions." + }, + { + "left": "issue:43606", + "right": "issue:45083", + "accept": false, + "reason": "CPU offload device mismatch and feature-length helper behavior do not describe the same failure." + }, + { + "left": "issue:43278", + "right": "issue:43825", + "accept": false, + "reason": "Training/eval dtype change is unrelated to a pipeline error-message regression." + }, + { + "left": "issue:29942", + "right": "issue:32090", + "accept": false, + "reason": "Flash Attention 2 test failures and _gpu_broadcast_one NoneType errors are separate CI/runtime problems." + }, + { + "left": "issue:37428", + "right": "issue:43425", + "accept": false, + "reason": "Flash-attention import failure and Torch 2.10 compatibility are too broad and not clearly the same concrete bug." + }, + { + "left": "issue:45310", + "right": "issue:45478", + "accept": true, + "reason": "These are the same Qwen3.5 MoE from_pretrained regression on transformers>=5.4.0; 45478 is a near-verbatim duplicate of 45310." + }, + { + "left": "issue:43611", + "right": "issue:43931", + "accept": false, + "reason": "Base_model_prefix loading breakage and Qwen3-VL weight-shape mismatch are different model-loading errors." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "Grounded overall. The Qwen3.5 MoE pair is clearly the same issue by title and chronology, and the rejected pairs stay conservative without overclaiming shared root cause. The summary appropriately keeps the cluster characterization modest and avoids broad dedupe claims beyond the evidence." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 13, + "review_unit_index": 21, + "review_unit_count": 42, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-367", + "nodes": [ + "issue:29127", + "issue:29942", + "issue:32090", + "issue:34689", + "issue:35141", + "issue:37428", + "issue:38175", + "issue:39692", + "issue:40444", + "issue:43066", + "issue:43278", + "issue:43329", + "issue:43335", + "issue:43404", + "issue:43450", + "issue:43504", + "issue:43526", + "issue:43550", + "issue:43606", + "issue:43653", + "issue:43701", + "issue:43720", + "issue:43723", + "issue:43819", + "issue:43873", + "issue:44112", + "issue:44292", + "issue:44351", + "issue:44448", + "issue:44462", + "issue:44493", + "issue:44554", + "issue:44568", + "issue:44779", + "issue:44855", + "issue:44857", + "issue:44977", + "issue:44987", + "issue:45072", + "issue:45083", + "issue:45200", + "issue:45229", + "issue:45372", + "issue:45381", + "issue:45406", + "issue:45491", + "issue:45520", + "issue:45563" + ], + "soft_pairs": [ + "issue:43335|issue:43550", + "issue:43819|issue:44493", + "issue:39692|issue:43526", + "issue:37428|issue:44351", + "issue:43450|issue:43526", + "issue:45200|issue:45381", + "issue:44855|issue:45083", + "issue:43404|issue:45200", + "issue:39692|issue:44493", + "issue:43335|issue:43653", + "issue:29127|issue:32090", + "issue:44448|issue:44779", + "issue:38175|issue:44493", + "issue:44292|issue:45520", + "issue:44977|issue:45229", + "issue:43335|issue:45083", + "issue:43335|issue:43504", + "issue:43550|issue:44112", + "issue:43335|issue:45072", + "issue:44568|issue:45491", + "issue:43720|issue:44112", + "issue:44448|issue:44977", + "issue:40444|issue:43526", + "issue:43335|issue:44112", + "issue:43723|issue:44554", + "issue:34689|issue:35141", + "issue:44857|issue:45072", + "issue:43606|issue:44857", + "issue:43066|issue:44462", + "issue:45372|issue:45406", + "issue:43873|issue:44493", + "issue:44977|issue:45563", + "issue:43873|issue:45229", + "issue:43278|issue:43329", + "issue:43701|issue:45083", + "issue:29942|issue:35141", + "issue:43550|issue:45083", + "issue:37428|issue:44987" + ], + "prepared_review_unit_hash": "75c19cf1a613041fa746f7802f80b2e451fb3f21caad9d67b908d69d09ff0bd8", + "component_budget": { + "node_count": 367, + "item_count": 367, + "soft_pair_count": 1370, + "serialized_chars": 706986, + "estimated_input_tokens": 176747, + "estimated_eval_tokens": 353750 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 38, + "serialized_chars": 39326, + "estimated_input_tokens": 9832, + "estimated_eval_tokens": 19920 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "All soft-similarity pairs look like false positives: they span different models, different failure modes, and different code paths, so there is no true duplicate cluster here.", + "confidence": 0.97, + "canonical_issue_reason": "No canonical issue: the items are heterogeneous (loading/import errors, tokenizer behavior, vision/video processing, training/dispatch, CI-only failures) and do not point to one underlying bug.", + "canonical_pr_reason": null, + "best_issue_reason": "No global best issue to anchor duplicates; none of the issues is a clear representative of the others because they are not the same defect.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43335", + "right": "issue:43550", + "accept": false, + "reason": "Sparse-layer config creation and torch.compile/SDPA model failures are different bugs in different code paths." + }, + { + "left": "issue:43819", + "right": "issue:44493", + "accept": false, + "reason": "DAC latent/forward mismatch is unrelated to unexpected position-id key warnings." + }, + { + "left": "issue:39692", + "right": "issue:43526", + "accept": false, + "reason": "Doc example/model-processor mismatch and quantization problems are not the same as BeitImageProcessorFast reduce_labels behavior." + }, + { + "left": "issue:37428", + "right": "issue:44351", + "accept": false, + "reason": "Missing flash-attention helper import and missing HybridCache import are separate import/API regressions." + }, + { + "left": "issue:43450", + "right": "issue:43526", + "accept": false, + "reason": "Batched video processor shape handling is unrelated to reduce_labels returning one label in BEiT fast processing." + }, + { + "left": "issue:45200", + "right": "issue:45381", + "accept": false, + "reason": "Gemma4 mm_token_type_ids defaults and Qwen2.5-VL video vision_position_ids are different multimodal bugs." + }, + { + "left": "issue:44855", + "right": "issue:45083", + "accept": false, + "reason": "Python 3.13 torch.jit parsing/import error is unrelated to qwen3_omni_moe feature-length helper behavior." + }, + { + "left": "issue:43404", + "right": "issue:45200", + "accept": false, + "reason": "lm_head weight tying in Mistral3 is a different issue from Gemma 4 token-type defaults." + }, + { + "left": "issue:39692", + "right": "issue:44493", + "accept": false, + "reason": "SigLIP2 example/quantization issues do not match the unexpected position-id-key warning regression." + }, + { + "left": "issue:43335", + "right": "issue:43653", + "accept": false, + "reason": "Sparse-layer configuration creation and tokenizer special-token registration are unrelated defects." + }, + { + "left": "issue:29127", + "right": "issue:32090", + "accept": false, + "reason": "LayoutLMv3 error-message clarity and Trainer NoneType broadcast failures are not the same bug." + }, + { + "left": "issue:44448", + "right": "issue:44779", + "accept": false, + "reason": "Pegasus v4/v5 output differences and Deepseek tokenizer regressions affect different models and code paths." + }, + { + "left": "issue:38175", + "right": "issue:44493", + "accept": false, + "reason": "SigLIP2 zero-probability outputs are not the same as unexpected position-id key warnings." + }, + { + "left": "issue:44292", + "right": "issue:45520", + "accept": false, + "reason": "Qwen-3-8B-NVFP4 runtime failure and flash_attn import KeyError are separate issues." + }, + { + "left": "issue:44977", + "right": "issue:45229", + "accept": false, + "reason": "Qwen3.5 flash-attention generation problems are unrelated to Gemma4 multi-GPU CUDA OOM." + }, + { + "left": "issue:43335", + "right": "issue:45083", + "accept": false, + "reason": "Sparse-layer config bug and qwen3_omni_moe length-helper behavior do not share the same failure mode." + }, + { + "left": "issue:43335", + "right": "issue:43504", + "accept": false, + "reason": "Sparse-layer creation and BEiT legacy-field loading failure are unrelated." + }, + { + "left": "issue:43550", + "right": "issue:44112", + "accept": false, + "reason": "torch.compile/SDPA model failure and a stale CI device-override test are not the same defect." + }, + { + "left": "issue:43335", + "right": "issue:45072", + "accept": false, + "reason": "Sparse-layer config creation is unrelated to dtype mismatches in bfloat16 inference." + }, + { + "left": "issue:44568", + "right": "issue:45491", + "accept": false, + "reason": "Tokenizer add_special_tokens behavior and Gemma3 NaN embeddings arise from different subsystems." + }, + { + "left": "issue:43720", + "right": "issue:44112", + "accept": false, + "reason": "Packed-weight unpacking during accelerate loading is unrelated to a stale CI device override." + }, + { + "left": "issue:44448", + "right": "issue:44977", + "accept": false, + "reason": "Pegasus output drift and Qwen3.5 flash-attention generation bugs are not the same underlying problem." + }, + { + "left": "issue:40444", + "right": "issue:43526", + "accept": false, + "reason": "IterableDataset multi-image finetuning failure is unrelated to BEiT reduce_labels output." + }, + { + "left": "issue:43335", + "right": "issue:44112", + "accept": false, + "reason": "Config-induced sparse-layer creation is not the same as a CI-only stale device override failure." + }, + { + "left": "issue:43723", + "right": "issue:44554", + "accept": false, + "reason": "AutoTokenizer loading regression and an MPS attention correctness issue are different problems." + }, + { + "left": "issue:34689", + "right": "issue:35141", + "accept": false, + "reason": "Llama 3.2 model loading break and post_init reinitializing embeddings are different bugs." + }, + { + "left": "issue:44857", + "right": "issue:45072", + "accept": false, + "reason": "AMP/CUDA crash in LwDetrImageLoss is unrelated to bfloat16 dtype mismatches in inference." + }, + { + "left": "issue:43606", + "right": "issue:44857", + "accept": false, + "reason": "CPU offload device mismatch and AMP/CUDA loss crash are different runtime failures." + }, + { + "left": "issue:43066", + "right": "issue:44462", + "accept": false, + "reason": "Wrong tokenizer decoder type and ignoring tokenizer.json are separate tokenizer-loading regressions." + }, + { + "left": "issue:45372", + "right": "issue:45406", + "accept": false, + "reason": "Gemma4 processor import dependency failure and missing _tokenizer attribute in serve are not the same issue." + }, + { + "left": "issue:43873", + "right": "issue:44493", + "accept": false, + "reason": "Quantization/offloading behavior is unrelated to unexpected position-id key warnings." + }, + { + "left": "issue:44977", + "right": "issue:45563", + "accept": false, + "reason": "Flash-attention generation failure and a stale num_return_sequences warning are different concerns." + }, + { + "left": "issue:43873", + "right": "issue:45229", + "accept": false, + "reason": "Quantized offloading issues and Gemma4 multi-GPU OOM are distinct failure modes." + }, + { + "left": "issue:43278", + "right": "issue:43329", + "accept": false, + "reason": "Embedding dtype drift during eval and undefined helper names in multimodal token counting are unrelated." + }, + { + "left": "issue:43701", + "right": "issue:45083", + "accept": false, + "reason": "resume_from_checkpoint key mismatch is unrelated to qwen3_omni_moe helper behavior." + }, + { + "left": "issue:29942", + "right": "issue:35141", + "accept": false, + "reason": "Flash Attention 2 test failures and embedding reinitialization in post_init do not describe the same bug." + }, + { + "left": "issue:43550", + "right": "issue:45083", + "accept": false, + "reason": "Bamba torch.compile/SDPA failure is unrelated to qwen3_omni_moe feature-length helper issues." + }, + { + "left": "issue:37428", + "right": "issue:44987", + "accept": false, + "reason": "Missing flash-support import and loading physical-intelligence/fast on transformers>=5.1.0 are different regressions." + } + ] + }, + "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": 13, + "review_unit_index": 22, + "review_unit_count": 42, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-367", + "nodes": [ + "issue:25251", + "issue:29127", + "issue:29942", + "issue:30064", + "issue:32090", + "issue:33290", + "issue:33357", + "issue:34567", + "issue:39692", + "issue:41720", + "issue:42915", + "issue:42994", + "issue:43065", + "issue:43257", + "issue:43278", + "issue:43381", + "issue:43421", + "issue:43504", + "issue:43653", + "issue:43756", + "issue:43976", + "issue:43994", + "issue:44186", + "issue:44206", + "issue:44265", + "issue:44361", + "issue:44448", + "issue:44451", + "issue:44466", + "issue:44479", + "issue:44610", + "issue:44743", + "issue:44857", + "issue:44871", + "issue:44977", + "issue:45072", + "issue:45103", + "issue:45198", + "issue:45229", + "issue:45245", + "issue:45372", + "issue:45412", + "issue:45440", + "issue:45459", + "issue:45491", + "issue:45520", + "issue:45563" + ], + "soft_pairs": [ + "issue:41720|issue:45229", + "issue:45245|issue:45563", + "issue:32090|issue:33357", + "issue:43976|issue:45440", + "issue:45372|issue:45459", + "issue:44451|issue:45520", + "issue:44448|issue:44466", + "issue:25251|issue:32090", + "issue:43065|issue:45412", + "issue:43065|issue:43994", + "issue:43278|issue:43381", + "issue:43653|issue:44743", + "issue:44186|issue:45198", + "issue:43257|issue:43278", + "issue:44265|issue:45198", + "issue:44361|issue:45198", + "issue:43421|issue:45491", + "issue:43994|issue:45412", + "issue:42915|issue:42994", + "issue:44871|issue:44977", + "issue:44206|issue:44479", + "issue:43756|issue:45440", + "issue:43504|issue:44857", + "issue:29942|issue:30064", + "issue:44610|issue:45412", + "issue:33357|issue:39692", + "issue:45103|issue:45372", + "issue:32090|issue:33290", + "issue:29127|issue:34567", + "issue:45072|issue:45229" + ], + "prepared_review_unit_hash": "860a73e3913dbe1609cdafa958411863c33c7bbdabced4ba906e6a6ecc51f374", + "component_budget": { + "node_count": 367, + "item_count": 367, + "soft_pair_count": 1370, + "serialized_chars": 706986, + "estimated_input_tokens": 176747, + "estimated_eval_tokens": 353750 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 30, + "serialized_chars": 36127, + "estimated_input_tokens": 9032, + "estimated_eval_tokens": 18320 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These are mostly unrelated single-issue reports across different models and subsystems. The soft similarity links are all false positives; none of the pairs look like the same underlying bug or change.", + "confidence": 0.95, + "canonical_issue_reason": "No single canonical issue: the items are mostly unrelated bugs rather than one duplicate family.", + "canonical_pr_reason": "No PRs are present in this cluster.", + "best_issue_reason": "No strong global best issue either, since there is no coherent duplicate cluster to anchor on.", + "best_pr_reason": "No PRs are present in this cluster.", + "soft_edge_verdicts": [ + { + "left": "issue:41720", + "right": "issue:45229", + "accept": false, + "reason": "Both are inference/runtime problems, but one is Qwen3 auto device-mapping asserts and the other is Gemma4 multi-GPU OOM; different models and failure modes." + }, + { + "left": "issue:45245", + "right": "issue:45563", + "accept": false, + "reason": "Unrelated problems: category-count limit runtime error vs a stale warning about num_return_sequences in paged generate()." + }, + { + "left": "issue:32090", + "right": "issue:33357", + "accept": false, + "reason": "Trainer broadcast NoneType error and MacOS bus error in CLIP are distinct bugs with no shared code path." + }, + { + "left": "issue:43976", + "right": "issue:45440", + "accept": false, + "reason": "Python version compatibility vs DeepSeekV3MoE divergence; these are not the same defect." + }, + { + "left": "issue:45372", + "right": "issue:45459", + "accept": false, + "reason": "One is a missing ReasoningEffort import breaking Gemma 4 loading; the other is tokenizer error masking when protobuf is absent." + }, + { + "left": "issue:44451", + "right": "issue:45520", + "accept": false, + "reason": "Model-loading failure for ScandiBERT and a Python 3.13 flash_attn import KeyError are unrelated." + }, + { + "left": "issue:44448", + "right": "issue:44466", + "accept": false, + "reason": "Both mention version-dependent model behavior, but one is Pegasus output regression and the other is lm_head serialization/tied-weights inconsistency." + }, + { + "left": "issue:25251", + "right": "issue:32090", + "accept": false, + "reason": "Pipeline top_k nesting bug has nothing to do with Trainer GPU broadcast of NoneType." + }, + { + "left": "issue:43065", + "right": "issue:45412", + "accept": false, + "reason": "Dummy Conv2d in Sam3PixelDecoder and RT-DETR memory not released are different lifecycle/code-path issues." + }, + { + "left": "issue:43065", + "right": "issue:43994", + "accept": false, + "reason": "Sam3PixelDecoder architecture placeholder vs SigLIP2 wrong outputs; different models and symptoms." + }, + { + "left": "issue:43278", + "right": "issue:43381", + "accept": false, + "reason": "Embedding dtype drift during eval and gradient checkpointing in eval mode are separate evaluation-time issues." + }, + { + "left": "issue:43653", + "right": "issue:44743", + "accept": false, + "reason": "Tokenizer special-token registration bug is unrelated to Qwen3.5 recurrent state reset with cache." + }, + { + "left": "issue:44186", + "right": "issue:45198", + "accept": false, + "reason": "LayoutLMv2Tokenizer NER/padding crash and Wav2Vec2 save/tokenization failure affect different tokenizers and paths." + }, + { + "left": "issue:43257", + "right": "issue:43278", + "accept": false, + "reason": "Qwen3 MoE weight conversion with accelerate/deepspeed is unrelated to embedding dtype changes in eval." + }, + { + "left": "issue:44265", + "right": "issue:45198", + "accept": false, + "reason": "torch.export failure due to torch_compilable_check is not the same as Wav2Vec2 save_pretrained/tokenization failure." + }, + { + "left": "issue:44361", + "right": "issue:45198", + "accept": false, + "reason": "MLukeTokenizer AttributeError and Wav2Vec2 save/tokenization failure are separate tokenizer bugs." + }, + { + "left": "issue:43421", + "right": "issue:45491", + "accept": false, + "reason": "Runtime post-processor update bug and Gemma3 NaN embeddings from sliding-window attention are unrelated." + }, + { + "left": "issue:43994", + "right": "issue:45412", + "accept": false, + "reason": "SigLIP2 nonsensical outputs and RT-DETR memory leak are completely different problems." + }, + { + "left": "issue:42915", + "right": "issue:42994", + "accept": false, + "reason": "FineGrainedFP8Config failure on Qwen3Moe and quantized model saving failure are both quantization-related but not the same bug." + }, + { + "left": "issue:44871", + "right": "issue:44977", + "accept": false, + "reason": "Gemma-3 eos_token_id mismatch and Qwen3.5 flash-attention generation failure do not share a concrete code path." + }, + { + "left": "issue:44206", + "right": "issue:44479", + "accept": false, + "reason": "Feature-extractor center-arg regression and Qwen VL video-input regression are different model/input pipelines." + }, + { + "left": "issue:43756", + "right": "issue:45440", + "accept": false, + "reason": "Smollm3 RoPE-layer mismatch and DeepSeekV3MoE divergence are unrelated model-architecture complaints." + }, + { + "left": "issue:43504", + "right": "issue:44857", + "accept": false, + "reason": "Beit pretrained-load legacy field bug and LwDetrImageLoss AMP CUDA crash are distinct failures." + }, + { + "left": "issue:29942", + "right": "issue:30064", + "accept": false, + "reason": "Flash Attention 2 test failures and void segmentation-map processing failure are unrelated." + }, + { + "left": "issue:44610", + "right": "issue:45412", + "accept": false, + "reason": "OmDet-Turbo processor image-size mismatch and RT-DETR memory leak are different bugs." + }, + { + "left": "issue:33357", + "right": "issue:39692", + "accept": false, + "reason": "MacOS CLIP bus error and SigLIP2 documentation example errors are not the same underlying issue." + }, + { + "left": "issue:45103", + "right": "issue:45372", + "accept": false, + "reason": "auto_docstring future-annotations crash and Gemma 4 processor import failure are unrelated import/runtime errors." + }, + { + "left": "issue:32090", + "right": "issue:33290", + "accept": false, + "reason": "Trainer broadcast NoneType error and DeepSpeed Adafactor OOM are different training-time problems." + }, + { + "left": "issue:29127", + "right": "issue:34567", + "accept": false, + "reason": "LayoutLMv3 error-message clarity and TrainerState token counter not updating are separate issues." + }, + { + "left": "issue:45072", + "right": "issue:45229", + "accept": false, + "reason": "dtype mismatch in inference wrappers and Gemma4 multi-GPU OOM are not the same defect." + } + ] + }, + "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": 13, + "review_unit_index": 23, + "review_unit_count": 42, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-367", + "nodes": [ + "issue:29127", + "issue:29942", + "issue:33357", + "issue:33666", + "issue:35141", + "issue:36010", + "issue:42915", + "issue:43232", + "issue:43262", + "issue:43316", + "issue:43334", + "issue:43381", + "issue:43388", + "issue:43525", + "issue:43526", + "issue:43531", + "issue:43575", + "issue:43646", + "issue:43653", + "issue:43701", + "issue:43723", + "issue:43761", + "issue:43824", + "issue:43931", + "issue:44077", + "issue:44079", + "issue:44186", + "issue:44188", + "issue:44206", + "issue:44279", + "issue:44442", + "issue:44464", + "issue:44479", + "issue:44610", + "issue:44743", + "issue:44811", + "issue:44857", + "issue:44964", + "issue:44977", + "issue:45003", + "issue:45072", + "issue:45216", + "issue:45245", + "issue:45290", + "issue:45440", + "issue:45479", + "issue:45491", + "issue:45584" + ], + "soft_pairs": [ + "issue:43824|issue:44279", + "issue:43381|issue:43701", + "issue:43653|issue:44442", + "issue:43334|issue:43531", + "issue:43526|issue:44186", + "issue:44464|issue:45491", + "issue:29942|issue:36010", + "issue:29127|issue:33357", + "issue:44977|issue:45245", + "issue:43262|issue:44206", + "issue:29127|issue:35141", + "issue:43761|issue:44079", + "issue:42915|issue:43931", + "issue:44442|issue:45072", + "issue:43232|issue:43388", + "issue:43526|issue:44857", + "issue:44479|issue:45290", + "issue:43526|issue:44610", + "issue:44743|issue:44811", + "issue:43723|issue:45479", + "issue:45003|issue:45440", + "issue:43575|issue:43701", + "issue:43931|issue:45216", + "issue:44977|issue:45584", + "issue:43646|issue:44077", + "issue:43761|issue:44188", + "issue:43316|issue:43525", + "issue:29942|issue:33666", + "issue:44964|issue:45440" + ], + "prepared_review_unit_hash": "3af25fde097b959b0962bc4954bd24324d853cf36594b0d6fda0bd593eec03e4", + "component_budget": { + "node_count": 367, + "item_count": 367, + "soft_pair_count": 1370, + "serialized_chars": 706986, + "estimated_input_tokens": 176747, + "estimated_eval_tokens": 353750 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36402, + "estimated_input_tokens": 9101, + "estimated_eval_tokens": 18458 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This looks like a loose thematic cluster of Transformers bugs and regressions, not a set of true duplicates. The soft pairs are mostly same-subsystem overlaps (Qwen, tokenizers, generation, configs) but each points to a different concrete failure mode or code path.", + "confidence": 0.66, + "canonical_issue_reason": "issue:43824 has the strongest signal as a representative user-facing regression: high discussion, multiple inbound references, and a concrete import failure with clear reproduction impact. It is the best single issue to anchor the cluster, even though the cluster itself does not collapse into one duplicate set.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43824 is the most externally validated and broadly relevant issue in the set, with the clearest evidence of impact and follow-up from other references. It is a good global representative for triage.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43824", + "right": "issue:44279", + "accept": false, + "reason": "Both are import/dependency-adjacent at a high level, but one is a missing model export and the other is a generic dependency issue; different underlying failures." + }, + { + "left": "issue:43381", + "right": "issue:43701", + "accept": false, + "reason": "One is about gradient checkpointing in eval mode; the other is a resume-from-checkpoint key mismatch. Same training area, different bug." + }, + { + "left": "issue:43653", + "right": "issue:44442", + "accept": false, + "reason": "Both involve tokenizers, but one is a special-token registration/decode bug and the other is an AutoTokenizer load failure for a different tokenizer." + }, + { + "left": "issue:43334", + "right": "issue:43531", + "accept": false, + "reason": "Both are Qwen-related, but one is a missing pad_token_id on Qwen3-VL config and the other is a sliding_window issue on Qwen3-MoE." + }, + { + "left": "issue:43526", + "right": "issue:44186", + "accept": false, + "reason": "Both touch preprocessing, but one is BeitImageProcessorFast label reduction and the other is a LayoutLMv2Tokenizer crash on NER/batching." + }, + { + "left": "issue:44464", + "right": "issue:45491", + "accept": false, + "reason": "Both are generation/attention regressions, but they involve different models and different failures: compiled-forward inconsistency vs NaNs from sliding-window attention." + }, + { + "left": "issue:29942", + "right": "issue:36010", + "accept": false, + "reason": "Flash Attention test failures and GenerationMixin import errors are unrelated code paths." + }, + { + "left": "issue:29127", + "right": "issue:33357", + "accept": false, + "reason": "Different models and failures: LayoutLMv3 error messaging vs a MacOS bus error in CLIP." + }, + { + "left": "issue:44977", + "right": "issue:45245", + "accept": false, + "reason": "Qwen3.5 flash-attention generation issues are unrelated to the category-count runtime limit error." + }, + { + "left": "issue:43262", + "right": "issue:44206", + "accept": false, + "reason": "Both are audio preprocessing regressions, but they are different unsupported-argument failures in different components." + }, + { + "left": "issue:29127", + "right": "issue:35141", + "accept": false, + "reason": "One is an error-message clarity issue; the other is a post_init reinitialization bug after resizing token embeddings." + }, + { + "left": "issue:43761", + "right": "issue:44079", + "accept": false, + "reason": "CLIPVisionModel hidden-states regression and ModelOutput key assignment are separate output-handling bugs." + }, + { + "left": "issue:42915", + "right": "issue:43931", + "accept": false, + "reason": "Both are Qwen model issues, but one is FP8 config failure and the other is a weight-shape mismatch during loading." + }, + { + "left": "issue:44442", + "right": "issue:45072", + "accept": false, + "reason": "Tokenizer loading and dtype mismatch in inference are unrelated." + }, + { + "left": "issue:43232", + "right": "issue:43388", + "accept": false, + "reason": "Generation kwargs synchronization and metric-label gathering are different distributed-training bugs." + }, + { + "left": "issue:43526", + "right": "issue:44857", + "accept": false, + "reason": "Label reduction in an image processor and a mixed-precision CUDA crash are not the same bug." + }, + { + "left": "issue:44479", + "right": "issue:45290", + "accept": false, + "reason": "Video-input regression for Qwen models and chat-template crashes on empty tool-call messages are different issues." + }, + { + "left": "issue:43526", + "right": "issue:44610", + "accept": false, + "reason": "Both are processor-related, but one is a label-reduction bug and the other is an input-size mismatch." + }, + { + "left": "issue:44743", + "right": "issue:44811", + "accept": false, + "reason": "Recurrent-state reset with cache and Whisper batch_decode skip_special_tokens are unrelated." + }, + { + "left": "issue:43723", + "right": "issue:45479", + "accept": false, + "reason": "Tokenizer loading in v5 and degenerate zero-loss sequence classification are different failures." + }, + { + "left": "issue:45003", + "right": "issue:45440", + "accept": false, + "reason": "Unsafe sys.modules access and DeepSeekV3MoE divergence are unrelated code paths." + }, + { + "left": "issue:43575", + "right": "issue:43701", + "accept": false, + "reason": "Tensor-parallel OOM on Qwen2-57B and checkpoint key mismatch are distinct problems." + }, + { + "left": "issue:43931", + "right": "issue:45216", + "accept": false, + "reason": "Both involve Qwen3.5/VL loading or saving, but one is a load-time shape mismatch and the other is an incorrect saved checkpoint regression." + }, + { + "left": "issue:44977", + "right": "issue:45584", + "accept": false, + "reason": "Different models and symptoms: Qwen3.5 flash-attn generation vs Whisper empty-transcription handling." + }, + { + "left": "issue:43646", + "right": "issue:44077", + "accept": false, + "reason": "Both concern initialization/post_init behavior, but the broad custom-model-init breakage and the PatchTSMixer-specific post_init policy are not the same concrete bug." + }, + { + "left": "issue:43761", + "right": "issue:44188", + "accept": false, + "reason": "Hidden_states regression in CLIPVisionModel and attention-kernel divergence under torch.compile are unrelated." + }, + { + "left": "issue:43316", + "right": "issue:43525", + "accept": false, + "reason": "Both are config-API inconsistencies, but one is a Gemma3TextConfig parity issue and the other is a Llama4Config pad_token_id attribute error." + }, + { + "left": "issue:29942", + "right": "issue:33666", + "accept": false, + "reason": "Failing Flash Attention tests and Qwen2-VL multi-GPU training are unrelated." + }, + { + "left": "issue:44964", + "right": "issue:45440", + "accept": false, + "reason": "Phi-4 multimodal loading failure and DeepSeekV3MoE divergence are different model-specific 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": 13, + "review_unit_index": 24, + "review_unit_count": 42, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-367", + "nodes": [ + "issue:33290", + "issue:34567", + "issue:34689", + "issue:35141", + "issue:36010", + "issue:36331", + "issue:39692", + "issue:41628", + "issue:42175", + "issue:43316", + "issue:43381", + "issue:43388", + "issue:43404", + "issue:43441", + "issue:43493", + "issue:43540", + "issue:43575", + "issue:43611", + "issue:43643", + "issue:43673", + "issue:43688", + "issue:43720", + "issue:43749", + "issue:43827", + "issue:43856", + "issue:43873", + "issue:43931", + "issue:43994", + "issue:44077", + "issue:44162", + "issue:44190", + "issue:44206", + "issue:44387", + "issue:44442", + "issue:44479", + "issue:44496", + "issue:44514", + "issue:44743", + "issue:44792", + "issue:44855", + "issue:44871", + "issue:44877", + "issue:44898", + "issue:44913", + "issue:45137", + "issue:45372", + "issue:45491" + ], + "soft_pairs": [ + "issue:43575|issue:44387", + "issue:43994|issue:44206", + "issue:39692|issue:45491", + "issue:33290|issue:34567", + "issue:44442|issue:44898", + "issue:43673|issue:44743", + "issue:34689|issue:43611", + "issue:43381|issue:43873", + "issue:43493|issue:43827", + "issue:33290|issue:35141", + "issue:33290|issue:34689", + "issue:41628|issue:45372", + "issue:44077|issue:44479", + "issue:43720|issue:44442", + "issue:43643|issue:44913", + "issue:43749|issue:43931", + "issue:43873|issue:44871", + "issue:44190|issue:44496", + "issue:43381|issue:43856", + "issue:33290|issue:36331", + "issue:44855|issue:45137", + "issue:34689|issue:36010", + "issue:43540|issue:44514", + "issue:43316|issue:44877", + "issue:44162|issue:44792", + "issue:43388|issue:43688", + "issue:42175|issue:43316", + "issue:43404|issue:43441" + ], + "prepared_review_unit_hash": "2888c4df9578ba1cb91aef0e6a2ff904c93690021c2fab7913c1d266999ba1eb", + "component_budget": { + "node_count": 367, + "item_count": 367, + "soft_pair_count": 1370, + "serialized_chars": 706986, + "estimated_input_tokens": 176747, + "estimated_eval_tokens": 353750 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 28, + "serialized_chars": 35699, + "estimated_input_tokens": 8925, + "estimated_eval_tokens": 18106 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is heterogeneous: most items are distinct bug reports about different models, configs, or regressions, with only superficial similarity around loading/memory/cache issues. I do not see any strong duplicate groups; the soft edges should be rejected.", + "confidence": 0.33, + "canonical_issue_reason": "Issue 44387 is the strongest representative only in the sense that it is a high-signal, broadly framed regression report with substantial discussion and impact; however, it is not a true duplicate anchor for the rest of the cluster.", + "canonical_pr_reason": null, + "best_issue_reason": "44387 is the best standalone issue because it is clearly scoped, has strong discussion activity, and describes a concrete regression rather than a narrow model-specific symptom.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43575", + "right": "issue:44387", + "accept": false, + "reason": "Both mention OOM, but they describe different causes and contexts: TP loading for a specific Qwen model versus increased CUDA reserved memory under int4 quantization." + }, + { + "left": "issue:43994", + "right": "issue:44206", + "accept": false, + "reason": "These are unrelated failures: one is SigLIP2 producing bad outputs with AutoModel/pipeline, the other is a feature extractor argument regression." + }, + { + "left": "issue:39692", + "right": "issue:45491", + "accept": false, + "reason": "Different models and symptoms: SigLIP2 docs example errors versus Gemma-3 NaN embeddings during mixed-length batching." + }, + { + "left": "issue:33290", + "right": "issue:34567", + "accept": false, + "reason": "OOM with Adafactor in DeepSpeed is unrelated to TrainerState's token counter not updating." + }, + { + "left": "issue:44442", + "right": "issue:44898", + "accept": false, + "reason": "One is a tokenizer load failure for FastSpeech2ConformerTokenizer; the other is a Perceiver image-classification failure on non-default resolution." + }, + { + "left": "issue:43673", + "right": "issue:44743", + "accept": false, + "reason": "Both involve cache semantics, but they affect different code paths and models: v5 chunked prefill cache lookup versus Qwen3.5 recurrent state reset." + }, + { + "left": "issue:34689", + "right": "issue:43611", + "accept": false, + "reason": "Both are loading-related, but one is a model-specific Llama 3.2 Vision issue and the other is a general base_model_prefix loading regression." + }, + { + "left": "issue:43381", + "right": "issue:43873", + "accept": false, + "reason": "Gradient checkpointing in eval mode and quantization offloading are different mechanisms with different failure modes." + }, + { + "left": "issue:43493", + "right": "issue:43827", + "accept": false, + "reason": "SigLIP2 implementation drift and outdated pipeline docs are unrelated." + }, + { + "left": "issue:33290", + "right": "issue:35141", + "accept": false, + "reason": "Adafactor/DeepSpeed OOM is not the same bug as output embeddings being reinitialized after resizing token embeddings." + }, + { + "left": "issue:33290", + "right": "issue:34689", + "accept": false, + "reason": "These are separate issues: optimizer/deepspeed OOM versus a specific model-loading breakage." + }, + { + "left": "issue:41628", + "right": "issue:45372", + "accept": false, + "reason": "Both are import errors, but they involve different missing symbols and different affected components." + }, + { + "left": "issue:44077", + "right": "issue:44479", + "accept": false, + "reason": "PatchTSMixer post_init policy and Qwen video-input regression are unrelated." + }, + { + "left": "issue:43720", + "right": "issue:44442", + "accept": false, + "reason": "Packed-weight unpacking during accelerate loading is unrelated to tokenizer loading." + }, + { + "left": "issue:43643", + "right": "issue:44913", + "accept": false, + "reason": "Missing fields with trust_remote_code and GPTNeoX rotary_pct reset are both config-related but not the same underlying defect." + }, + { + "left": "issue:43749", + "right": "issue:43931", + "accept": false, + "reason": "FSDP CPU RAM efficient loading and Qwen3-VL weight-shape mismatch are different loading problems." + }, + { + "left": "issue:43873", + "right": "issue:44871", + "accept": false, + "reason": "Quantization offloading and Gemma-3 eos_token_id mismatch are unrelated." + }, + { + "left": "issue:44190", + "right": "issue:44496", + "accept": false, + "reason": "A local dataset loading failure in an example script is not the same as an unrecognized model/config loading error." + }, + { + "left": "issue:43381", + "right": "issue:43856", + "accept": false, + "reason": "One is a gradient-checkpointing eval-mode restriction; the other is a Qwen3 MoE memory-efficiency complaint." + }, + { + "left": "issue:33290", + "right": "issue:36331", + "accept": false, + "reason": "OOM in DeepSpeed is unrelated to a custom trainer signature change." + }, + { + "left": "issue:44855", + "right": "issue:45137", + "accept": false, + "reason": "Python 3.13 parsing/import error in DebertaV2Model is unrelated to a ZeRO3 deque underflow." + }, + { + "left": "issue:34689", + "right": "issue:36010", + "accept": false, + "reason": "A model-loading regression for Llama 3.2 Vision is not the same as the GenerationMixin import-path break." + }, + { + "left": "issue:43540", + "right": "issue:44514", + "accept": false, + "reason": "Both are Qwen-related processor failures, but one is video input processing in Qwen3OmniMoe and the other is batched chat-template padding behavior." + }, + { + "left": "issue:43316", + "right": "issue:44877", + "accept": false, + "reason": "Gemma3TextConfig API inconsistency and strict config loading for granite_speech are different config issues." + }, + { + "left": "issue:44162", + "right": "issue:44792", + "accept": false, + "reason": "ESM2 breakage and a janus model test failure do not describe the same underlying bug." + }, + { + "left": "issue:43388", + "right": "issue:43688", + "accept": false, + "reason": "Metric gathering dropping tuple labels and auxiliary-loss normalization are unrelated training bugs." + }, + { + "left": "issue:42175", + "right": "issue:43316", + "accept": false, + "reason": "Backend packaging for TensorFlow and a Gemma3TextConfig API discrepancy are unrelated." + }, + { + "left": "issue:43404", + "right": "issue:43441", + "accept": false, + "reason": "Weight tying in Mistral3ForConditionalGeneration and a FlashAttention failure in Ministral-3 are different issues, even if both concern related 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": 2, + "component_count": 13, + "review_unit_index": 25, + "review_unit_count": 42, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-367", + "nodes": [ + "issue:16998", + "issue:22355", + "issue:25251", + "issue:28282", + "issue:32090", + "issue:35141", + "issue:41950", + "issue:42617", + "issue:43065", + "issue:43295", + "issue:43316", + "issue:43452", + "issue:43475", + "issue:43550", + "issue:43575", + "issue:43606", + "issue:43646", + "issue:43701", + "issue:43716", + "issue:43742", + "issue:43761", + "issue:43825", + "issue:43873", + "issue:43906", + "issue:43931", + "issue:44060", + "issue:44077", + "issue:44242", + "issue:44464", + "issue:44479", + "issue:44496", + "issue:44521", + "issue:44534", + "issue:44617", + "issue:44734", + "issue:44792", + "issue:44871", + "issue:44913", + "issue:44933", + "issue:44977", + "issue:45127", + "issue:45200", + "issue:45290", + "issue:45362", + "issue:45381", + "issue:45406", + "issue:45563", + "issue:45584" + ], + "soft_pairs": [ + "issue:16998|issue:22355", + "issue:25251|issue:28282", + "issue:43316|issue:43716", + "issue:43065|issue:44617", + "issue:41950|issue:43475", + "issue:45127|issue:45200", + "issue:43295|issue:43825", + "issue:44871|issue:45584", + "issue:43761|issue:44479", + "issue:44077|issue:44534", + "issue:43550|issue:43701", + "issue:43295|issue:44871", + "issue:43606|issue:43701", + "issue:44734|issue:45406", + "issue:43646|issue:44913", + "issue:43906|issue:45290", + "issue:43452|issue:44496", + "issue:43761|issue:44242", + "issue:43575|issue:43742", + "issue:42617|issue:43575", + "issue:44617|issue:45584", + "issue:45563|issue:45584", + "issue:43931|issue:44060", + "issue:44521|issue:45381", + "issue:43575|issue:43873", + "issue:32090|issue:35141", + "issue:44479|issue:44734", + "issue:44933|issue:45362", + "issue:44464|issue:44871", + "issue:44464|issue:45563", + "issue:44792|issue:44977" + ], + "prepared_review_unit_hash": "083a739e41d2bae7b0e9b3f54f2cfd28b209db1a779577fc4ab29f799a17ccb6", + "component_budget": { + "node_count": 367, + "item_count": 367, + "soft_pair_count": 1370, + "serialized_chars": 706986, + "estimated_input_tokens": 176747, + "estimated_eval_tokens": 353750 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 36789, + "estimated_input_tokens": 9198, + "estimated_eval_tokens": 18652 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "All suggested soft edges look like false positives. The items span unrelated bugs across loading, generation, multimodal processing, training, and model-specific regressions, with no pair clearly describing the same underlying fix. No pull requests are present.", + "confidence": 0.97, + "canonical_issue_reason": "No single canonical issue stands out: the candidates are about different subsystems and failure modes, so none is a safe representative for deduplication.", + "canonical_pr_reason": null, + "best_issue_reason": "No issue is a strong global representative because the cluster is not a true duplicate set; choosing one would conflate unrelated reports.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:16998", + "right": "issue:22355", + "accept": false, + "reason": "Different problems: DeBERTa max length handling vs missing `transformers.onnx` module import." + }, + { + "left": "issue:25251", + "right": "issue:28282", + "accept": false, + "reason": "Pipeline `top_k` output nesting bug is unrelated to the PyTorch import error for `AutoModel`." + }, + { + "left": "issue:43316", + "right": "issue:43716", + "accept": false, + "reason": "Gemma3TextConfig API inconsistency is unrelated to the Mistral-3 image preprocessor dtype mismatch." + }, + { + "left": "issue:43065", + "right": "issue:44617", + "accept": false, + "reason": "`Sam3PixelDecoder` dummy conv issue is a different bug from Sam3Video CUDA OOM." + }, + { + "left": "issue:41950", + "right": "issue:43475", + "accept": false, + "reason": "Video-classification processor lookup and missing `fpn_position_embeddings` are not the same code-path." + }, + { + "left": "issue:45127", + "right": "issue:45200", + "accept": false, + "reason": "LoRA merge/tied embeddings collapse is unrelated to Gemma 4 `mm_token_type_ids` defaults." + }, + { + "left": "issue:43295", + "right": "issue:43825", + "accept": false, + "reason": "Processor/tokenizer regression in v4.57.5 is different from the pipeline translation-task error message." + }, + { + "left": "issue:44871", + "right": "issue:45584", + "accept": false, + "reason": "Gemma-3 EOS token config mismatch is unrelated to Whisper empty-transcription handling." + }, + { + "left": "issue:43761", + "right": "issue:44479", + "accept": false, + "reason": "CLIP hidden_states regression and Qwen video-input regression affect different models and failures." + }, + { + "left": "issue:44077", + "right": "issue:44534", + "accept": false, + "reason": "`patchtsmixer` post_init restriction is unrelated to non-persistent buffer junk in v5." + }, + { + "left": "issue:43550", + "right": "issue:43701", + "accept": false, + "reason": "Bamba torch.compile/SDPA failure and resume_from_checkpoint key mismatch are different bugs." + }, + { + "left": "issue:43295", + "right": "issue:44871", + "accept": false, + "reason": "Custom processor/tokenizer regression is not the same as Gemma-3 EOS token inconsistency." + }, + { + "left": "issue:43606", + "right": "issue:43701", + "accept": false, + "reason": "CPU offload device mismatch for Bark is unrelated to checkpoint key mismatch on resume." + }, + { + "left": "issue:44734", + "right": "issue:45406", + "accept": false, + "reason": "KV-cache continuation indexing in serve is unrelated to Gemma4Processor missing `_tokenizer`." + }, + { + "left": "issue:43646", + "right": "issue:44913", + "accept": false, + "reason": "Custom model initialization breakage and GPTNeoX rotary_pct reload loss are different configuration bugs." + }, + { + "left": "issue:43906", + "right": "issue:45290", + "accept": false, + "reason": "An isolated reproduction of a prior issue is not the same as the tool-call assistant message crash." + }, + { + "left": "issue:43452", + "right": "issue:44496", + "accept": false, + "reason": "`gguf_file` loading regression is distinct from the unrecognized-model/config.json error." + }, + { + "left": "issue:43761", + "right": "issue:44242", + "accept": false, + "reason": "CLIP hidden_states regression is unrelated to MoE load-balancing loss omission." + }, + { + "left": "issue:43575", + "right": "issue:43742", + "accept": false, + "reason": "Qwen2-57B tensor-parallel OOM and MobileLLM key error are unrelated model-loading failures." + }, + { + "left": "issue:42617", + "right": "issue:43575", + "accept": false, + "reason": "Unable to run `3d_parallel.py` is not the same as Qwen2-57B TP OOM." + }, + { + "left": "issue:44617", + "right": "issue:45584", + "accept": false, + "reason": "Sam3Video OOM is unrelated to Whisper empty transcription after token alignment." + }, + { + "left": "issue:45563", + "right": "issue:45584", + "accept": false, + "reason": "Paged generate warning for `num_return_sequences` is unrelated to Whisper empty transcription." + }, + { + "left": "issue:43931", + "right": "issue:44060", + "accept": false, + "reason": "Qwen3-VL shape mismatch and Qwen3-Next tied-weights warning are different model/config bugs." + }, + { + "left": "issue:44521", + "right": "issue:45381", + "accept": false, + "reason": "Assistant-mask failure for multimodal chat is not the same as qwen2.5-vl video position-id regression." + }, + { + "left": "issue:43575", + "right": "issue:43873", + "accept": false, + "reason": "Qwen2-57B OOM is a different symptom from offloading/quantization behavior." + }, + { + "left": "issue:32090", + "right": "issue:35141", + "accept": false, + "reason": "Trainer `_gpu_broadcast_one` NoneType error and embedding reinitialization after resize are unrelated." + }, + { + "left": "issue:44479", + "right": "issue:44734", + "accept": false, + "reason": "Qwen video-input regression and serve KV-cache indexing crash affect different paths." + }, + { + "left": "issue:44933", + "right": "issue:45362", + "accept": false, + "reason": "Nonexistent `image_utils` import and Qwen3.5 chat crash are unrelated." + }, + { + "left": "issue:44464", + "right": "issue:44871", + "accept": false, + "reason": "Chunked generation inconsistency with compiled forward is unrelated to Gemma-3 EOS config mismatch." + }, + { + "left": "issue:44464", + "right": "issue:45563", + "accept": false, + "reason": "Chunked generation/compiled forward differs from the stale paged-generate warning." + }, + { + "left": "issue:44792", + "right": "issue:44977", + "accept": false, + "reason": "Janus test failure and Qwen3.5 flash-attention generation bug are different model-specific 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": 13, + "review_unit_index": 26, + "review_unit_count": 42, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-367", + "nodes": [ + "issue:36246", + "issue:38617", + "issue:39692", + "issue:41628", + "issue:41720", + "issue:41950", + "issue:43065", + "issue:43116", + "issue:43295", + "issue:43335", + "issue:43388", + "issue:43404", + "issue:43479", + "issue:43540", + "issue:43550", + "issue:43582", + "issue:43644", + "issue:43650", + "issue:43761", + "issue:43827", + "issue:43976", + "issue:43994", + "issue:44016", + "issue:44077", + "issue:44079", + "issue:44186", + "issue:44190", + "issue:44220", + "issue:44355", + "issue:44466", + "issue:44496", + "issue:44534", + "issue:44610", + "issue:44734", + "issue:44743", + "issue:44855", + "issue:44913", + "issue:44938", + "issue:44964", + "issue:45127", + "issue:45216", + "issue:45245", + "issue:45276", + "issue:45357", + "issue:45372", + "issue:45375", + "issue:45542", + "issue:45584" + ], + "soft_pairs": [ + "issue:43827|issue:43994", + "issue:43065|issue:44610", + "issue:44534|issue:44913", + "issue:41720|issue:45245", + "issue:44938|issue:45542", + "issue:44190|issue:44964", + "issue:43650|issue:44220", + "issue:43388|issue:43582", + "issue:45127|issue:45357", + "issue:43761|issue:44743", + "issue:45276|issue:45584", + "issue:43644|issue:44913", + "issue:44077|issue:44079", + "issue:44355|issue:44855", + "issue:43540|issue:44734", + "issue:43976|issue:44855", + "issue:44496|issue:45372", + "issue:43335|issue:44855", + "issue:43550|issue:44190", + "issue:43404|issue:45216", + "issue:43295|issue:45276", + "issue:43116|issue:44190", + "issue:45245|issue:45584", + "issue:39692|issue:44190", + "issue:41950|issue:44016", + "issue:43479|issue:45375", + "issue:36246|issue:41628", + "issue:44186|issue:44734", + "issue:36246|issue:38617", + "issue:44466|issue:44913" + ], + "prepared_review_unit_hash": "a60c003ffb334595e4457c551ee93cf0ce67a0ec85a04ca83cde65a9dddebff4", + "component_budget": { + "node_count": 367, + "item_count": 367, + "soft_pair_count": 1370, + "serialized_chars": 706986, + "estimated_input_tokens": 176747, + "estimated_eval_tokens": 353750 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 37070, + "estimated_input_tokens": 9268, + "estimated_eval_tokens": 18792 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This looks like a grab-bag of unrelated Transformers bugs. None of the proposed soft edges appear to be the same underlying issue, so I would not merge any of these items into a duplicate cluster.", + "confidence": 0.96, + "canonical_issue_reason": "No clear canonical issue: the items span distinct failures (imports, config loading, model serialization, training scripts, and serving/runtime bugs) rather than one shared bug.", + "canonical_pr_reason": null, + "best_issue_reason": "If forced to pick a representative, issue 43295 is the broadest user-facing regression, but it still does not represent the other issues well enough to serve as a true canonical duplicate.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43827", + "right": "issue:43994", + "accept": false, + "reason": "Docs still referencing pipeline() removals is unrelated to SigLIP2 producing bad outputs; different symptom and code path." + }, + { + "left": "issue:43065", + "right": "issue:44610", + "accept": false, + "reason": "Sam3PixelDecoder dummy conv2d and OmDet-Turbo processor size mismatch are separate vision-model bugs." + }, + { + "left": "issue:44534", + "right": "issue:44913", + "accept": false, + "reason": "Non-persistent buffer junk serialization and GPTNeoX rotary_pct reload are different configuration/serialization problems." + }, + { + "left": "issue:41720", + "right": "issue:45245", + "accept": false, + "reason": "Qwen3 auto device mapping CUDA assert is unrelated to the category-count limit RuntimeError." + }, + { + "left": "issue:44938", + "right": "issue:45542", + "accept": false, + "reason": "Python 3.14 import/load failure and missing TensorFlow backend error are distinct environment/package issues." + }, + { + "left": "issue:44190", + "right": "issue:44964", + "accept": false, + "reason": "Local dataset loading in an image-classification script is not the same bug as loading Phi-4 multimodal models." + }, + { + "left": "issue:43650", + "right": "issue:44220", + "accept": false, + "reason": "The placeholder \"ADD THE DATA\" issue and the fbank feature extraction bug do not match." + }, + { + "left": "issue:43388", + "right": "issue:43582", + "accept": false, + "reason": "Metric label truncation and Apple Silicon allocator warmup TypeError are unrelated." + }, + { + "left": "issue:45127", + "right": "issue:45357", + "accept": false, + "reason": "Both involve Qwen models, but one is LoRA/extended-vocab merge collapse and the other is incorrect visual-encoder save keys; different concrete failures." + }, + { + "left": "issue:43761", + "right": "issue:44743", + "accept": false, + "reason": "CLIPVisionModel hidden_states regression and Qwen3 modular recurrent-state reset are separate model behaviors." + }, + { + "left": "issue:45276", + "right": "issue:45584", + "accept": false, + "reason": "Gemma4 embedding resize behavior and Whisper empty transcription after align_special_tokens are unrelated." + }, + { + "left": "issue:43644", + "right": "issue:44913", + "accept": false, + "reason": "Non-persistent buffer junk in v5 and GPTNeoX rotary_pct reload are different bugs." + }, + { + "left": "issue:44077", + "right": "issue:44079", + "accept": false, + "reason": "Optional post_init handling in patchtsmixer is not the same as ModelOutput key assignment when values were previously None." + }, + { + "left": "issue:44355", + "right": "issue:44855", + "accept": false, + "reason": "Compiled Python file errors and DebertaV2Model import indentation parsing are different failure modes." + }, + { + "left": "issue:43540", + "right": "issue:44734", + "accept": false, + "reason": "Qwen3OmniMoe video-input validation and serve KV-cache continuation indexing affect different code paths." + }, + { + "left": "issue:43976", + "right": "issue:44855", + "accept": false, + "reason": "Python version compatibility for Transformers 5.1.0 is unrelated to the DebertaV2 @torch.jit.script parsing error." + }, + { + "left": "issue:44496", + "right": "issue:45372", + "accept": false, + "reason": "Unrecognized model/config loading and mistral_common ReasoningEffort import breakage are different loader issues." + }, + { + "left": "issue:43335", + "right": "issue:44855", + "accept": false, + "reason": "Sparse-layer creation in SwitchTransformersConfig is unrelated to the DebertaV2 import parsing problem." + }, + { + "left": "issue:43550", + "right": "issue:44190", + "accept": false, + "reason": "Bamba torch.compile+SDPA failure and local dataset loading in an image-classification script are unrelated." + }, + { + "left": "issue:43404", + "right": "issue:45216", + "accept": false, + "reason": "lm_head weight tying in Mistral3ForConditionalGeneration and Qwen3.5 save_pretrained regression are both weight/save issues but not the same concrete bug." + }, + { + "left": "issue:43295", + "right": "issue:45276", + "accept": false, + "reason": "Processor.tokenizer/image-to-tokenizer regression and Gemma4 embedding resize not propagating are different API regressions." + }, + { + "left": "issue:43116", + "right": "issue:44190", + "accept": false, + "reason": "Multilabel classification returning empty results is not the same as failing to load a local dataset." + }, + { + "left": "issue:45245", + "right": "issue:45584", + "accept": false, + "reason": "Category-cardinality limit and Whisper empty transcription are unrelated errors." + }, + { + "left": "issue:39692", + "right": "issue:44190", + "accept": false, + "reason": "SigLIP2 docs example errors and the local image-classification dataset loading bug are different issues." + }, + { + "left": "issue:41950", + "right": "issue:44016", + "accept": false, + "reason": "Video-classification pipeline processor lookup and a notebook syntax error are unrelated." + }, + { + "left": "issue:43479", + "right": "issue:45375", + "accept": false, + "reason": "Phi4MultimodalConfig default-init behavior and Qwen3_5MoeVisionConfig missing field handling are separate config bugs." + }, + { + "left": "issue:36246", + "right": "issue:41628", + "accept": false, + "reason": "Both are import errors, but they refer to different missing symbols and different import surfaces; not the same bug." + }, + { + "left": "issue:44186", + "right": "issue:44734", + "accept": false, + "reason": "LayoutLMv2Tokenizer NER/padding crashes are unrelated to serve KV-cache continuation indexing." + }, + { + "left": "issue:36246", + "right": "issue:38617", + "accept": false, + "reason": "Different missing imports in different modules; no shared underlying fix is evident." + }, + { + "left": "issue:44466", + "right": "issue:44913", + "accept": false, + "reason": "Device-dependent lm_head serialization inconsistency and rotary_pct reload persistence are different configuration 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": 13, + "review_unit_index": 27, + "review_unit_count": 42, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-367", + "nodes": [ + "issue:38175", + "issue:42175", + "issue:42907", + "issue:43065", + "issue:43295", + "issue:43299", + "issue:43441", + "issue:43479", + "issue:43525", + "issue:43526", + "issue:43577", + "issue:43643", + "issue:43673", + "issue:43824", + "issue:43854", + "issue:43901", + "issue:43976", + "issue:44016", + "issue:44062", + "issue:44075", + "issue:44112", + "issue:44220", + "issue:44242", + "issue:44246", + "issue:44279", + "issue:44336", + "issue:44351", + "issue:44368", + "issue:44496", + "issue:44521", + "issue:44683", + "issue:44792", + "issue:44933", + "issue:44964", + "issue:44987", + "issue:45003", + "issue:45042", + "issue:45200", + "issue:45216", + "issue:45278", + "issue:45325", + "issue:45335", + "issue:45357", + "issue:45405", + "issue:45446", + "issue:45542", + "issue:45584" + ], + "soft_pairs": [ + "issue:38175|issue:43065", + "issue:44683|issue:45446", + "issue:44016|issue:44220", + "issue:44246|issue:45003", + "issue:42175|issue:44279", + "issue:44112|issue:44792", + "issue:43643|issue:44496", + "issue:43299|issue:45216", + "issue:43295|issue:45335", + "issue:42907|issue:43441", + "issue:44016|issue:44062", + "issue:44112|issue:44964", + "issue:45335|issue:45584", + "issue:43673|issue:44336", + "issue:43976|issue:44987", + "issue:43854|issue:44964", + "issue:44496|issue:45042", + "issue:43824|issue:44933", + "issue:43526|issue:43901", + "issue:43976|issue:44279", + "issue:43299|issue:45357", + "issue:43577|issue:45216", + "issue:44075|issue:44368", + "issue:44521|issue:45325", + "issue:43901|issue:44496", + "issue:44242|issue:45200", + "issue:44683|issue:45542", + "issue:43976|issue:45405", + "issue:44351|issue:45278", + "issue:43479|issue:43525" + ], + "prepared_review_unit_hash": "b373b4767fd548755848602c879d709fd65c603b800bacc49d79b57ed17283e3", + "component_budget": { + "node_count": 367, + "item_count": 367, + "soft_pair_count": 1370, + "serialized_chars": 706986, + "estimated_input_tokens": 176747, + "estimated_eval_tokens": 353750 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 30, + "serialized_chars": 35976, + "estimated_input_tokens": 8994, + "estimated_eval_tokens": 18244 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a grab bag of Transformers regression/import/model-specific issues rather than true duplicates. All soft-edge pairs look like false positives; several share a subsystem or symptom class, but not the same concrete bug.", + "confidence": 0.84, + "canonical_issue_reason": "issue:45278 is the broadest representative of the cluster\u2019s recurring theme: post-upgrade import/runtime breakage. It\u2019s still not a true duplicate of the others, but it best captures the generalized failure mode compared with the many narrower model- or API-specific reports.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45278 is the best single issue to stand in for this cluster because it is the most umbrella-like and most likely to subsume related import failures after upgrading Transformers.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:38175", + "right": "issue:43065", + "accept": false, + "reason": "Different models and different failures: zero probabilities in SigLIP2 versus a dummy Conv2d in Sam3PixelDecoder." + }, + { + "left": "issue:44683", + "right": "issue:45446", + "accept": false, + "reason": "Both involve flex_attention, but they describe different concrete problems: a torch>=2.9 compiled flex_attention failure versus a PyTorch version check for AuxRequest import." + }, + { + "left": "issue:44016", + "right": "issue:44220", + "accept": false, + "reason": "A notebook syntax error and an audio feature-extraction bug are unrelated." + }, + { + "left": "issue:44246", + "right": "issue:45003", + "accept": false, + "reason": "Import slowdowns and unsafe sys.modules access are not the same bug; one is performance, the other is unsafe import handling." + }, + { + "left": "issue:42175", + "right": "issue:44279", + "accept": false, + "reason": "The first is a specific missing TensorFlow backend in a torch install; the second is a vague dependency issue with no evidence of the same root cause." + }, + { + "left": "issue:44112", + "right": "issue:44792", + "accept": false, + "reason": "Different failing tests in different models; same general CI symptom, not the same underlying defect." + }, + { + "left": "issue:43643", + "right": "issue:44496", + "accept": false, + "reason": "One is a trust_remote_code/config field propagation issue; the other is an unrecognized model config problem." + }, + { + "left": "issue:43299", + "right": "issue:45216", + "accept": false, + "reason": "Both are Qwen-related regressions, but one is a model loading failure and the other is a save_pretrained checkpoint serialization bug." + }, + { + "left": "issue:43295", + "right": "issue:45335", + "accept": false, + "reason": "Different APIs and code paths: processor/tokenizer access regression versus resize_token_embeddings not updating decoder embeddings." + }, + { + "left": "issue:42907", + "right": "issue:43441", + "accept": false, + "reason": "Both mention Ministral, but one is about saving dequantized models and the other about FlashAttention failures." + }, + { + "left": "issue:44016", + "right": "issue:44062", + "accept": false, + "reason": "Notebook syntax error versus AddedToken special-argument TypeError; unrelated." + }, + { + "left": "issue:44112", + "right": "issue:44964", + "accept": false, + "reason": "Different models and failures; a GraniteSpeech CI test issue is not the same as Phi-4 multimodal loading failure." + }, + { + "left": "issue:45335", + "right": "issue:45584", + "accept": false, + "reason": "Tokenizer resize behavior and Whisper empty-transcription generation are different bugs." + }, + { + "left": "issue:43673", + "right": "issue:44336", + "accept": false, + "reason": "Generation cache missing during chunked prefill and ANSI code leakage in loading_report are unrelated." + }, + { + "left": "issue:43976", + "right": "issue:44987", + "accept": false, + "reason": "A Python-version compatibility issue is not the same as a model-loading failure for physical-intelligence/fast." + }, + { + "left": "issue:43854", + "right": "issue:44964", + "accept": false, + "reason": "Different model families and different failures: GLM-4.7-Flash unit-test loading versus Phi-4 multimodal loading." + }, + { + "left": "issue:44496", + "right": "issue:45042", + "accept": false, + "reason": "An unrecognized model config and a torchvision dependency regression in PIL image processors are unrelated." + }, + { + "left": "issue:43824", + "right": "issue:44933", + "accept": false, + "reason": "Both are import errors, but for different missing symbols and different modules; too generic to treat as duplicates." + }, + { + "left": "issue:43526", + "right": "issue:43901", + "accept": false, + "reason": "A label-reduction bug in BeitImageProcessorFast is unrelated to stale pipeline documentation." + }, + { + "left": "issue:43976", + "right": "issue:44279", + "accept": false, + "reason": "A Python compatibility problem is not the same as a vague dependency issue." + }, + { + "left": "issue:43299", + "right": "issue:45357", + "accept": false, + "reason": "Both are Qwen3.5-related, but one is a loading regression and the other is incorrect visual-encoder key saving." + }, + { + "left": "issue:43577", + "right": "issue:45216", + "accept": false, + "reason": "Blip2 dtype propagation and Qwen3.5 checkpoint saving are different code paths and different bugs." + }, + { + "left": "issue:44075", + "right": "issue:44368", + "accept": false, + "reason": "Optimizer argument handling and a tie_word_embeddings warning are unrelated." + }, + { + "left": "issue:44521", + "right": "issue:45325", + "accept": false, + "reason": "Both are multimodal, but assistant mask generation and rope/position-id scaling are different failures." + }, + { + "left": "issue:43901", + "right": "issue:44496", + "accept": false, + "reason": "Documentation mismatch versus model config recognition failure; unrelated." + }, + { + "left": "issue:44242", + "right": "issue:45200", + "accept": false, + "reason": "Load-balancing loss behavior in MoE models is not the same as mm_token_type_ids defaulting for text-only fine-tuning." + }, + { + "left": "issue:44683", + "right": "issue:45542", + "accept": false, + "reason": "Compiled flex_attention on torch>=2.9 and missing TensorFlow backend installation are unrelated." + }, + { + "left": "issue:43976", + "right": "issue:45405", + "accept": false, + "reason": "Python version support and a PEFT minimum-version bump are separate dependency issues." + }, + { + "left": "issue:44351", + "right": "issue:45278", + "accept": false, + "reason": "45278 is a broad umbrella for many import errors after upgrade, but 44351 is a specific HybridCache import failure; not enough evidence they share one exact root cause." + }, + { + "left": "issue:43479", + "right": "issue:43525", + "accept": false, + "reason": "Different config bugs: default multimodal sub-config initialization versus missing pad_token_id on Llama4Config." + } + ] + }, + "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": 13, + "review_unit_index": 28, + "review_unit_count": 42, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-367", + "nodes": [ + "issue:38175", + "issue:39692", + "issue:41950", + "issue:42994", + "issue:43065", + "issue:43122", + "issue:43262", + "issue:43295", + "issue:43299", + "issue:43317", + "issue:43441", + "issue:43450", + "issue:43550", + "issue:43650", + "issue:43673", + "issue:43723", + "issue:43761", + "issue:43784", + "issue:43854", + "issue:43901", + "issue:44016", + "issue:44162", + "issue:44261", + "issue:44263", + "issue:44279", + "issue:44291", + "issue:44297", + "issue:44462", + "issue:44466", + "issue:44496", + "issue:44521", + "issue:44617", + "issue:44623", + "issue:44671", + "issue:44683", + "issue:44749", + "issue:44779", + "issue:44877", + "issue:44898", + "issue:44964", + "issue:44987", + "issue:45020", + "issue:45081", + "issue:45276", + "issue:45290", + "issue:45397", + "issue:45405", + "issue:45542" + ], + "soft_pairs": [ + "issue:41950|issue:43650", + "issue:44496|issue:44877", + "issue:44462|issue:45081", + "issue:41950|issue:45020", + "issue:44297|issue:44462", + "issue:43550|issue:44964", + "issue:44279|issue:45405", + "issue:43299|issue:43854", + "issue:43295|issue:44779", + "issue:44462|issue:44779", + "issue:39692|issue:43901", + "issue:38175|issue:44617", + "issue:43650|issue:45290", + "issue:44016|issue:44291", + "issue:43723|issue:44162", + "issue:43450|issue:44279", + "issue:43441|issue:44683", + "issue:44466|issue:45276", + "issue:43262|issue:44521", + "issue:42994|issue:43317", + "issue:44261|issue:44263", + "issue:43065|issue:44898", + "issue:44623|issue:45397", + "issue:43122|issue:43650", + "issue:44623|issue:44749", + "issue:43784|issue:45542", + "issue:44671|issue:44779", + "issue:43673|issue:43761", + "issue:44987|issue:45542" + ], + "prepared_review_unit_hash": "2c0faa0868625c96c14ea576dc70a798fbb0720820a4f0bd530d6338b3b9a9f1", + "component_budget": { + "node_count": 367, + "item_count": 367, + "soft_pair_count": 1370, + "serialized_chars": 706986, + "estimated_input_tokens": 176747, + "estimated_eval_tokens": 353750 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36343, + "estimated_input_tokens": 9086, + "estimated_eval_tokens": 18428 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Mostly unrelated issues with superficial overlap around Transformers v5 regressions, tokenizers, processors, and model loading. None of the soft pairs look like the same underlying bug/change.", + "confidence": 0.91, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:41950", + "right": "issue:43650", + "accept": false, + "reason": "Video pipeline processor lookup is unrelated to the vague \"ADD THE DATA\" report." + }, + { + "left": "issue:44496", + "right": "issue:44877", + "accept": false, + "reason": "Both are loading/config issues, but they fail for different reasons and different models." + }, + { + "left": "issue:44462", + "right": "issue:45081", + "accept": false, + "reason": "AutoTokenizer ignoring repo tokenizer.json is a different bug from Mistral regex patch crashing on backend_tokenizer." + }, + { + "left": "issue:41950", + "right": "issue:45020", + "accept": false, + "reason": "A video-classification processor bug is not the same as a broad remote_code compatibility regression." + }, + { + "left": "issue:44297", + "right": "issue:44462", + "accept": false, + "reason": "Tokenizer save metadata mismatch and tokenizer.json precedence are distinct serialization vs loading issues." + }, + { + "left": "issue:43550", + "right": "issue:44964", + "accept": false, + "reason": "Different models and failure modes: torch.compile/SDPA vs model loading for Phi-4 multimodal." + }, + { + "left": "issue:44279", + "right": "issue:45405", + "accept": false, + "reason": "Unrelated dependency problems: transformers dependency issue vs an unreleased PEFT version bump." + }, + { + "left": "issue:43299", + "right": "issue:43854", + "accept": false, + "reason": "Both are model-loading problems, but they affect different models and code paths." + }, + { + "left": "issue:43295", + "right": "issue:44779", + "accept": false, + "reason": "Processor.tokenizer regression is not the same as Deepseek tokenizer output drift." + }, + { + "left": "issue:44462", + "right": "issue:44779", + "accept": false, + "reason": "Repo tokenizer.json selection and Deepseek tokenization changes are different tokenizer bugs." + }, + { + "left": "issue:39692", + "right": "issue:43901", + "accept": false, + "reason": "Both mention docs, but one is a broken SigLIP2 example and the other is stale pipeline documentation." + }, + { + "left": "issue:38175", + "right": "issue:44617", + "accept": false, + "reason": "Unexpected zero probabilities in SigLIP2 is unrelated to Sam3Video CUDA OOM." + }, + { + "left": "issue:43650", + "right": "issue:45290", + "accept": false, + "reason": "\"ADD THE DATA\" is not a substantive duplicate target for a chat-template tool-call crash." + }, + { + "left": "issue:44016", + "right": "issue:44291", + "accept": false, + "reason": "Notebook syntax error and init_empty_weights TypeError are unrelated failures." + }, + { + "left": "issue:43723", + "right": "issue:44162", + "accept": false, + "reason": "Tokenizer loading regression and ESM2 breakage are different model/component issues." + }, + { + "left": "issue:43450", + "right": "issue:44279", + "accept": false, + "reason": "Batched video processor shape bugs are unrelated to a generic dependency issue." + }, + { + "left": "issue:43441", + "right": "issue:44683", + "accept": false, + "reason": "FlashAttention failure and compiled flex_attention failure are different kernels and triggers." + }, + { + "left": "issue:44466", + "right": "issue:45276", + "accept": false, + "reason": "lm_head serialization inconsistency and resize_token_embeddings propagation are different weight-handling bugs." + }, + { + "left": "issue:43262", + "right": "issue:44521", + "accept": false, + "reason": "Audio chat-template sampling-rate default and multimodal assistant_mask generation are different failures in the same API." + }, + { + "left": "issue:42994", + "right": "issue:43317", + "accept": false, + "reason": "Saving quantized models and loading dequantized offloaded models are separate quantization workflows." + }, + { + "left": "issue:44261", + "right": "issue:44263", + "accept": false, + "reason": "Different low-level bugs: missing config.rms_norm_eps vs torch.split return-value handling." + }, + { + "left": "issue:43065", + "right": "issue:44898", + "accept": false, + "reason": "Dummy Conv2d in Sam3PixelDecoder and Perceiver resize/interpolation failure are unrelated vision-model issues." + }, + { + "left": "issue:44623", + "right": "issue:45397", + "accept": false, + "reason": "Processor.save_pretrained missing files is not the same as a zero3 from_pretrained loading failure." + }, + { + "left": "issue:43122", + "right": "issue:43650", + "accept": false, + "reason": "Tokenizer version drift is unrelated to the non-specific \"ADD THE DATA\" issue." + }, + { + "left": "issue:44623", + "right": "issue:44749", + "accept": false, + "reason": "Missing save_pretrained files and a performance regression after upgrade are different classes of bugs." + }, + { + "left": "issue:43784", + "right": "issue:45542", + "accept": false, + "reason": "Importing sentence-transformers breaks for nn, while the other issue is about tensorboard/TensorFlow backend detection." + }, + { + "left": "issue:44671", + "right": "issue:44779", + "accept": false, + "reason": "CamemBERT masked-LM prediction regression is a different bug from Deepseek tokenizer output changes." + }, + { + "left": "issue:43673", + "right": "issue:43761", + "accept": false, + "reason": "Chunked-prefill cache regression and CLIPVisionModel hidden_states regression are separate v5 breakages." + }, + { + "left": "issue:44987", + "right": "issue:45542", + "accept": false, + "reason": "Model-loading failure for physical-intelligence/fast is unrelated to the tensorboard/TensorFlow backend issue." + } + ] + }, + "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": 13, + "review_unit_index": 29, + "review_unit_count": 42, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-367", + "nodes": [ + "issue:36246", + "issue:41628", + "issue:41950", + "issue:42175", + "issue:42491", + "issue:42757", + "issue:43295", + "issue:43299", + "issue:43352", + "issue:43475", + "issue:43525", + "issue:43526", + "issue:43531", + "issue:43673", + "issue:43824", + "issue:43881", + "issue:43901", + "issue:43976", + "issue:43994", + "issue:44016", + "issue:44162", + "issue:44188", + "issue:44220", + "issue:44246", + "issue:44295", + "issue:44297", + "issue:44336", + "issue:44351", + "issue:44373", + "issue:44462", + "issue:44485", + "issue:44496", + "issue:44561", + "issue:44589", + "issue:44617", + "issue:44661", + "issue:45003", + "issue:45200", + "issue:45250", + "issue:45292", + "issue:45356", + "issue:45362", + "issue:45397", + "issue:45542", + "issue:45561", + "issue:45563", + "issue:45584" + ], + "soft_pairs": [ + "issue:43901|issue:43994", + "issue:42175|issue:43976", + "issue:44297|issue:45397", + "issue:45561|issue:45563", + "issue:43475|issue:44016", + "issue:43824|issue:43976", + "issue:43352|issue:45250", + "issue:43525|issue:43881", + "issue:44589|issue:45362", + "issue:43295|issue:44617", + "issue:44188|issue:44336", + "issue:43901|issue:44246", + "issue:41628|issue:43976", + "issue:44295|issue:45397", + "issue:42491|issue:43299", + "issue:41950|issue:44220", + "issue:44336|issue:44373", + "issue:44462|issue:45356", + "issue:44246|issue:44496", + "issue:44661|issue:45542", + "issue:45003|issue:45561", + "issue:43531|issue:45362", + "issue:43526|issue:45003", + "issue:36246|issue:42757", + "issue:36246|issue:44351", + "issue:44336|issue:44485", + "issue:43673|issue:44561", + "issue:44162|issue:45200", + "issue:45292|issue:45584" + ], + "prepared_review_unit_hash": "35c0d608082a1ee9f0f70fb8748080b4f90df9ed54ac335cc012f9d412ca0ce5", + "component_budget": { + "node_count": 367, + "item_count": 367, + "soft_pair_count": 1370, + "serialized_chars": 706986, + "estimated_input_tokens": 176747, + "estimated_eval_tokens": 353750 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 29, + "serialized_chars": 35485, + "estimated_input_tokens": 8872, + "estimated_eval_tokens": 18000 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is a loose mix of unrelated Transformers issues. The soft-similarity pairs are mostly same-area but different bugs, regressions, or docs/runtime symptoms, so none should be merged as duplicates.", + "confidence": 0.94, + "canonical_issue_reason": "If forced to pick one anchor, issue 43299 is the clearest concrete regression report and the most representative of the Qwen3-MoE loading-breakage subgroup; there is no single true canonical issue for the whole set.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 43299 is the best overall representative: it is specific, reproducible, and describes a real model-loading regression rather than a generic symptom or unrelated API/docs problem.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43901", + "right": "issue:43994", + "accept": false, + "reason": "One is a docs/behavior mismatch for text classification; the other is a model quality/loading problem for SigLIP2. Different failure modes." + }, + { + "left": "issue:42175", + "right": "issue:43976", + "accept": false, + "reason": "TensorFlow backend omission from a pip extra is unrelated to the Python-version support regression." + }, + { + "left": "issue:44297", + "right": "issue:45397", + "accept": false, + "reason": "Tokenizer class metadata mismatch vs gemma-4 ZeRO3 from_pretrained failure; different code paths and symptoms." + }, + { + "left": "issue:45561", + "right": "issue:45563", + "accept": false, + "reason": "A pytest-xdist file race is unrelated to a stale generation warning in paged generate()." + }, + { + "left": "issue:43475", + "right": "issue:44016", + "accept": false, + "reason": "SAM3 video attribute bug and a notebook syntax error are entirely separate issues." + }, + { + "left": "issue:43824", + "right": "issue:43976", + "accept": false, + "reason": "Import export failure for a Qwen2.5-VL symbol is not the same as a Python compatibility issue." + }, + { + "left": "issue:43352", + "right": "issue:45250", + "accept": false, + "reason": "A specific model lacking Flash Attention 2.0 support is not the same as the general Flash Attention 2.0 topic." + }, + { + "left": "issue:43525", + "right": "issue:43881", + "accept": false, + "reason": "Missing pad_token_id on Llama4Config and glm-4v-9b loading failure are different bugs." + }, + { + "left": "issue:44589", + "right": "issue:45362", + "accept": false, + "reason": "Float8 storage deserialization failure and Qwen3.5 chat crash are unrelated runtime errors." + }, + { + "left": "issue:43295", + "right": "issue:44617", + "accept": false, + "reason": "Processor/tokenizer regression and SAM3Video OOM are different problems." + }, + { + "left": "issue:44188", + "right": "issue:44336", + "accept": false, + "reason": "Torch.compile attention-kernel divergence has nothing to do with ANSI codes in loading_report." + }, + { + "left": "issue:43901", + "right": "issue:44246", + "accept": false, + "reason": "Docs mismatch for return_all_scores is unrelated to intermittent slow imports." + }, + { + "left": "issue:41628", + "right": "issue:43976", + "accept": false, + "reason": "AutoImageProcessor import failure and Python version support are separate import/runtime issues." + }, + { + "left": "issue:44295", + "right": "issue:45397", + "accept": false, + "reason": "position_ids buffer access error is unrelated to gemma-4 zero3 from_pretrained loading." + }, + { + "left": "issue:42491", + "right": "issue:43299", + "accept": false, + "reason": "Both involve Qwen3-MoE family breakage, but one is a LoRA/hf4.x-to-hf5.x compatibility report and the other is a specific Qwen3VL-MoE loading regression; not the same concrete bug." + }, + { + "left": "issue:41950", + "right": "issue:44220", + "accept": false, + "reason": "Video-classification pipeline processor lookup and torch fbank feature extraction are different audio/video preprocessing bugs." + }, + { + "left": "issue:44336", + "right": "issue:44373", + "accept": false, + "reason": "Terminal ANSI output leakage and a wrong docstring for position_ids are unrelated." + }, + { + "left": "issue:44462", + "right": "issue:45356", + "accept": false, + "reason": "AutoTokenizer ignoring tokenizer.json and Kimi tokenizer codec regression are both tokenizer issues, but not the same underlying bug." + }, + { + "left": "issue:44246", + "right": "issue:44496", + "accept": false, + "reason": "Slow import behavior and unrecognized model config are unrelated." + }, + { + "left": "issue:44661", + "right": "issue:45542", + "accept": false, + "reason": "Tokenizer mapping validation and missing TensorFlow backend handling are separate packaging/runtime issues." + }, + { + "left": "issue:45003", + "right": "issue:45561", + "accept": false, + "reason": "Unsafe sys.modules access in modeling_utils is unrelated to a captured_info.txt xdist race." + }, + { + "left": "issue:43531", + "right": "issue:45362", + "accept": false, + "reason": "Qwen3-MoE sliding_window behavior and Qwen3.5 chat crashes are not the same code-path problem." + }, + { + "left": "issue:43526", + "right": "issue:45003", + "accept": false, + "reason": "BEiT image processor label reduction bug and modeling_utils sys.modules access are unrelated." + }, + { + "left": "issue:36246", + "right": "issue:42757", + "accept": false, + "reason": "Both are import errors, but they involve different packages and different missing symbols." + }, + { + "left": "issue:36246", + "right": "issue:44351", + "accept": false, + "reason": "Qwen2.5-VL image processor import failure is not the same as HybridCache export failure." + }, + { + "left": "issue:44336", + "right": "issue:44485", + "accept": false, + "reason": "ANSI output leakage and GLM-5 RoPE implementation discussion are unrelated." + }, + { + "left": "issue:43673", + "right": "issue:44561", + "accept": false, + "reason": "Chunked-prefill cache handling and removal of is_torch_fx_available affect different subsystems and code paths." + }, + { + "left": "issue:44162", + "right": "issue:45200", + "accept": false, + "reason": "ESM2 breakage and Gemma 4 mm_token_type_ids defaults are separate model-specific bugs." + }, + { + "left": "issue:45292", + "right": "issue:45584", + "accept": false, + "reason": "resize_token_embeddings not updating output embeddings is unrelated to Whisper empty-transcription generation 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": 13, + "review_unit_index": 30, + "review_unit_count": 42, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-367", + "nodes": [ + "issue:36246", + "issue:41628", + "issue:41950", + "issue:42886", + "issue:43065", + "issue:43066", + "issue:43232", + "issue:43257", + "issue:43262", + "issue:43295", + "issue:43408", + "issue:43441", + "issue:43475", + "issue:43540", + "issue:43596", + "issue:43606", + "issue:43653", + "issue:43673", + "issue:43901", + "issue:44060", + "issue:44077", + "issue:44162", + "issue:44242", + "issue:44246", + "issue:44295", + "issue:44336", + "issue:44368", + "issue:44466", + "issue:44496", + "issue:44734", + "issue:44749", + "issue:44779", + "issue:44857", + "issue:44871", + "issue:44877", + "issue:44933", + "issue:44977", + "issue:44987", + "issue:44995", + "issue:45003", + "issue:45137", + "issue:45200", + "issue:45229", + "issue:45292", + "issue:45335", + "issue:45375", + "issue:45459", + "issue:45563" + ], + "soft_pairs": [ + "issue:44242|issue:44857", + "issue:44246|issue:44871", + "issue:44295|issue:44749", + "issue:44779|issue:45292", + "issue:44877|issue:45375", + "issue:43441|issue:43540", + "issue:36246|issue:43596", + "issue:43066|issue:44779", + "issue:44987|issue:45459", + "issue:42886|issue:45229", + "issue:41628|issue:41950", + "issue:44995|issue:45200", + "issue:43901|issue:45003", + "issue:43257|issue:45137", + "issue:43065|issue:43408", + "issue:43232|issue:44336", + "issue:43295|issue:45292", + "issue:44496|issue:45003", + "issue:44060|issue:44077", + "issue:43262|issue:44246", + "issue:43475|issue:43540", + "issue:43653|issue:44162", + "issue:44466|issue:45335", + "issue:44368|issue:44977", + "issue:44368|issue:45563", + "issue:44933|issue:45459", + "issue:43441|issue:43606", + "issue:43673|issue:44734" + ], + "prepared_review_unit_hash": "0d8d9c917531b1966a671a19eec80590d91b0b0db6208db571e901442a629241", + "component_budget": { + "node_count": 367, + "item_count": 367, + "soft_pair_count": 1370, + "serialized_chars": 706986, + "estimated_input_tokens": 176747, + "estimated_eval_tokens": 353750 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 28, + "serialized_chars": 36206, + "estimated_input_tokens": 9052, + "estimated_eval_tokens": 18360 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is mostly heterogeneous; the soft pairs look like superficially similar Transformers issues but not the same underlying bug or fix. I would anchor on the broad import-performance report as the best issue representative, but I would not merge any of the suggested soft edges.", + "confidence": 0.57, + "canonical_issue_reason": "issue:44246 is the broadest, most general user-facing report in the set and best matches the cluster\u2019s core import/runtime theme, even though it is not a duplicate of the other items.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:44246 has the widest scope and strongest general relevance among the issues here; it is the most plausible anchor for a mixed cluster of Transformers import/runtime complaints.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44242", + "right": "issue:44857", + "accept": false, + "reason": "Different bugs: one is router/load-balancing loss behavior, the other is AMP/CUDA crash in LwDetrImageLoss." + }, + { + "left": "issue:44246", + "right": "issue:44871", + "accept": false, + "reason": "Import slowdown vs Gemma-3 EOS config inconsistency; unrelated failure modes." + }, + { + "left": "issue:44295", + "right": "issue:44749", + "accept": false, + "reason": "Position_ids buffer handling is not the same as the reported data filtering slowdown." + }, + { + "left": "issue:44779", + "right": "issue:45292", + "accept": false, + "reason": "Tokenizer output regression vs resize_token_embeddings not updating output embeddings; different code paths." + }, + { + "left": "issue:44877", + "right": "issue:45375", + "accept": false, + "reason": "Both involve config handling, but one is strict config loading and the other is a missing model field; not the same bug." + }, + { + "left": "issue:43441", + "right": "issue:43540", + "accept": false, + "reason": "FlashAttention failure for one model family vs video-input ValueError in another; unrelated." + }, + { + "left": "issue:36246", + "right": "issue:43596", + "accept": false, + "reason": "ImportError for Qwen2.5-VL image processor vs DeepSpeed ZeRO3/BertModel index error; no shared code-path." + }, + { + "left": "issue:43066", + "right": "issue:44779", + "accept": false, + "reason": "Tokenizer decoder-type warning/regression vs incorrect DeepSeek tokenizer outputs; same area, but not the same underlying defect." + }, + { + "left": "issue:44987", + "right": "issue:45459", + "accept": false, + "reason": "Model loading failure for physical-intelligence/fast is unrelated to protobuf exception masking in tokenizer decode errors." + }, + { + "left": "issue:42886", + "right": "issue:45229", + "accept": false, + "reason": "Offline cache tokenizer loading vs Gemma4 multi-GPU OOM; completely different problems." + }, + { + "left": "issue:41628", + "right": "issue:41950", + "accept": false, + "reason": "Importing AutoImageProcessor vs video-classification pipeline choosing image processors; related subsystem, but not the same bug." + }, + { + "left": "issue:44995", + "right": "issue:45200", + "accept": false, + "reason": "Stale indexer cache crash vs missing mm_token_type_ids defaults; different model behaviors." + }, + { + "left": "issue:43901", + "right": "issue:45003", + "accept": false, + "reason": "Pipeline docs mismatch vs unsafe sys.modules access in modeling_utils; unrelated." + }, + { + "left": "issue:43257", + "right": "issue:45137", + "accept": false, + "reason": "Qwen3 MoE weight conversion issue vs ZeRO3 deque IndexError; different failure mechanisms." + }, + { + "left": "issue:43065", + "right": "issue:43408", + "accept": false, + "reason": "Dummy Conv2d implementation issue vs model-type warning between sam3_video and sam3_tracker; not the same underlying defect." + }, + { + "left": "issue:43232", + "right": "issue:44336", + "accept": false, + "reason": "Generation kwargs/cache issue vs ANSI codes in loading_report; unrelated." + }, + { + "left": "issue:43295", + "right": "issue:45292", + "accept": false, + "reason": "Processor/tokenizer regression with images vs resize_token_embeddings not updating output embeddings; different API surfaces." + }, + { + "left": "issue:44496", + "right": "issue:45003", + "accept": false, + "reason": "Unrecognized model/config loading error vs unsafe sys.modules access; not the same bug." + }, + { + "left": "issue:44060", + "right": "issue:44077", + "accept": false, + "reason": "Tied-weights warning bug vs optional post_init restriction for patchtsmixer; unrelated." + }, + { + "left": "issue:43262", + "right": "issue:44246", + "accept": false, + "reason": "Audio chat-template sample-rate default vs import time regression; different subsystems." + }, + { + "left": "issue:43475", + "right": "issue:43540", + "accept": false, + "reason": "Missing fpn_position_embeddings attribute vs Qwen3OmniMoe video-input processing error; not the same code path." + }, + { + "left": "issue:43653", + "right": "issue:44162", + "accept": false, + "reason": "BigBirdTokenizer special-token registration vs ESM2 model breakage; not a duplicate." + }, + { + "left": "issue:44466", + "right": "issue:45335", + "accept": false, + "reason": "Serialization inconsistency for tied weights vs resize_token_embeddings not affecting decoder.embed_tokens; related theme, but distinct bugs." + }, + { + "left": "issue:44368", + "right": "issue:44977", + "accept": false, + "reason": "Tie_word_embeddings warning during LoRA fine-tuning vs flash-attention generation failure; different problems." + }, + { + "left": "issue:44368", + "right": "issue:45563", + "accept": false, + "reason": "Tie_word_embeddings warning vs stale num_return_sequences warning in paged generate; unrelated." + }, + { + "left": "issue:44933", + "right": "issue:45459", + "accept": false, + "reason": "Missing image_utils import vs protobuf decode-error masking in tokenizers; no overlap in root cause." + }, + { + "left": "issue:43441", + "right": "issue:43606", + "accept": false, + "reason": "FlashAttention model failure vs CPU offload device mismatch; different runtime issues." + }, + { + "left": "issue:43673", + "right": "issue:44734", + "accept": false, + "reason": "Chunked-prefill cache missing in GenerationMixin vs server KV-cache continuation tensor indexing; both cache-related but not the same fixable defect." + } + ] + }, + "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": 13, + "review_unit_index": 31, + "review_unit_count": 42, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-367", + "nodes": [ + "issue:28282", + "issue:30990", + "issue:41628", + "issue:42222", + "issue:42371", + "issue:42673", + "issue:42831", + "issue:42994", + "issue:43232", + "issue:43475", + "issue:43482", + "issue:43504", + "issue:43525", + "issue:43643", + "issue:43646", + "issue:43673", + "issue:43701", + "issue:43704", + "issue:43873", + "issue:43931", + "issue:44038", + "issue:44162", + "issue:44188", + "issue:44222", + "issue:44297", + "issue:44351", + "issue:44361", + "issue:44368", + "issue:44623", + "issue:44661", + "issue:44749", + "issue:44869", + "issue:44933", + "issue:44945", + "issue:45081", + "issue:45092", + "issue:45103", + "issue:45200", + "issue:45230", + "issue:45278", + "issue:45310", + "issue:45356", + "issue:45362", + "issue:45375", + "issue:45381", + "issue:45397", + "issue:45478", + "issue:45561" + ], + "soft_pairs": [ + "issue:44869|issue:45356", + "issue:43931|issue:44368", + "issue:43232|issue:43673", + "issue:44297|issue:44623", + "issue:43673|issue:45278", + "issue:43646|issue:44162", + "issue:44361|issue:44749", + "issue:28282|issue:30990", + "issue:42673|issue:43704", + "issue:44749|issue:45310", + "issue:45230|issue:45397", + "issue:44038|issue:44623", + "issue:42994|issue:43482", + "issue:43475|issue:43643", + "issue:43701|issue:44749", + "issue:43475|issue:45381", + "issue:42371|issue:45092", + "issue:45081|issue:45103", + "issue:45375|issue:45478", + "issue:41628|issue:44933", + "issue:45310|issue:45375", + "issue:42994|issue:43525", + "issue:42222|issue:45200", + "issue:43475|issue:44661", + "issue:43873|issue:44945", + "issue:45362|issue:45375", + "issue:42831|issue:44222", + "issue:45310|issue:45397", + "issue:43232|issue:44188", + "issue:44222|issue:45310", + "issue:41628|issue:44351", + "issue:43504|issue:45561" + ], + "prepared_review_unit_hash": "e16a39bf9904bc734a02553e028b64a3998b12d9ccfc47dc3e3e57e560c9f3a4", + "component_budget": { + "node_count": 367, + "item_count": 367, + "soft_pair_count": 1370, + "serialized_chars": 706986, + "estimated_input_tokens": 176747, + "estimated_eval_tokens": 353750 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 32, + "serialized_chars": 36998, + "estimated_input_tokens": 9250, + "estimated_eval_tokens": 18756 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is mostly a mix of Transformers v5 import/loading/config regressions plus a few model-specific bugs. There are two clear duplicate pairs in the packet: issue:42673/issue:43704 and issue:45310/issue:45478; the rest of the soft links are topical at best, not true duplicates.", + "confidence": 0.81, + "canonical_issue_reason": "issue:45310 is a clean exact duplicate pair with issue:45478 and is a representative concrete v5 from_pretrained regression within the cluster.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45278 is the broadest umbrella for the cluster\u2019s v5-era import/loading regression reports, even though it is vague.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44869", + "right": "issue:45356", + "accept": false, + "reason": "Different bugs: Whisper timestamp decode crash vs Kimi-K2.5 tokenizer codec/regex regression." + }, + { + "left": "issue:43931", + "right": "issue:44368", + "accept": false, + "reason": "Different failures: Qwen3-VL weight-shape loading error vs Qwen3.5 LoRA warning about tie_word_embeddings." + }, + { + "left": "issue:43232", + "right": "issue:43673", + "accept": false, + "reason": "Both involve generation, but they affect different code paths: sync_gpus kwarg handling vs chunked_prefill cache regression." + }, + { + "left": "issue:44297", + "right": "issue:44623", + "accept": false, + "reason": "Both are save_pretrained-related, but tokenizer metadata mismatch and missing processor files are distinct bugs." + }, + { + "left": "issue:43673", + "right": "issue:45278", + "accept": false, + "reason": "The latter is a broad import-error umbrella after upgrade, not the same concrete cache-missing bug." + }, + { + "left": "issue:43646", + "right": "issue:44162", + "accept": false, + "reason": "Transformers 5 custom-model init breakage is broader than the specific ESM2 failure; not enough evidence they are the same defect." + }, + { + "left": "issue:44361", + "right": "issue:44749", + "accept": false, + "reason": "Tokenizer AttributeError and a severe filtering slowdown are unrelated." + }, + { + "left": "issue:28282", + "right": "issue:30990", + "accept": false, + "reason": "PyTorch missing ImportError and a sentence-transformers loading hang are different issues." + }, + { + "left": "issue:42673", + "right": "issue:43704", + "accept": true, + "reason": "Same title and same concrete bug: Qwen3ForCausalLM VRAM leak when used across multiple dataloader threads." + }, + { + "left": "issue:44749", + "right": "issue:45310", + "accept": false, + "reason": "Data filtering slowdown vs Qwen3.5-Moe from_pretrained error are unrelated." + }, + { + "left": "issue:45230", + "right": "issue:45397", + "accept": false, + "reason": "Vague bug report vs Gemma-4 ZeRO3 from_pretrained failure; no shared concrete bug." + }, + { + "left": "issue:44038", + "right": "issue:44623", + "accept": false, + "reason": "Qwen3-VL-Moe transformers v5 bug and processor.save_pretrained missing files are different problems." + }, + { + "left": "issue:42994", + "right": "issue:43482", + "accept": false, + "reason": "Quantized model saving failure and Qwen2.5-GGUF loading failure are distinct save/load bugs." + }, + { + "left": "issue:43475", + "right": "issue:43643", + "accept": false, + "reason": "SAM 3 Video missing attribute and trust_remote_code missing fields are different model-loading issues." + }, + { + "left": "issue:43701", + "right": "issue:44749", + "accept": false, + "reason": "Checkpoint resume key mismatch and data filtering slowdown are unrelated." + }, + { + "left": "issue:43475", + "right": "issue:45381", + "accept": false, + "reason": "Both are vision/video model issues, but the specific missing attribute and wrong vision_position_ids bugs are not the same." + }, + { + "left": "issue:42371", + "right": "issue:45092", + "accept": false, + "reason": "TF32 API settings warning and InternVL2 meta-init incompatibility are unrelated." + }, + { + "left": "issue:45081", + "right": "issue:45103", + "accept": false, + "reason": "Tokenizer regex patch crash and auto_docstring annotation crash are separate code paths." + }, + { + "left": "issue:45375", + "right": "issue:45478", + "accept": false, + "reason": "Missing vision config field and Qwen3.5-Moe from_pretrained error are not the same concrete bug." + }, + { + "left": "issue:41628", + "right": "issue:44933", + "accept": false, + "reason": "Different missing-import symptoms: AutoImageProcessor vs an image_utils import." + }, + { + "left": "issue:45310", + "right": "issue:45375", + "accept": false, + "reason": "Qwen3.5-Moe load error and missing vision config field are not the same issue." + }, + { + "left": "issue:42994", + "right": "issue:43525", + "accept": false, + "reason": "Quantized saving failure and Llama4Config missing pad_token_id are different bugs." + }, + { + "left": "issue:42222", + "right": "issue:45200", + "accept": false, + "reason": "Vitpose model breakage and Gemma 4 mm_token_type_ids handling are separate model-specific problems." + }, + { + "left": "issue:43475", + "right": "issue:44661", + "accept": false, + "reason": "SAM 3 Video attribute error and add-new-model-like/tokenizer-mapping failure are unrelated." + }, + { + "left": "issue:43873", + "right": "issue:44945", + "accept": false, + "reason": "Quantization offloading bug and pipeline-parallel output corruption are different concerns." + }, + { + "left": "issue:45362", + "right": "issue:45375", + "accept": false, + "reason": "Qwen3.5-35B chat crash and missing Qwen3_5MoeVisionConfig field are different failures." + }, + { + "left": "issue:42831", + "right": "issue:44222", + "accept": false, + "reason": "FineGrainedFP8 accuracy issue and FP8 save_pretrained bug are not the same code-path problem." + }, + { + "left": "issue:45310", + "right": "issue:45397", + "accept": false, + "reason": "Same from_pretrained wording, but different models and different failure modes." + }, + { + "left": "issue:43232", + "right": "issue:44188", + "accept": false, + "reason": "Generation kwarg update bug and torch.compile attention-kernel divergence are distinct." + }, + { + "left": "issue:44222", + "right": "issue:45310", + "accept": false, + "reason": "FP8 save_pretrained bug and Qwen3.5-Moe load error are unrelated." + }, + { + "left": "issue:41628", + "right": "issue:44351", + "accept": false, + "reason": "Both are import errors, but they are for different symbols and likely different fixes." + }, + { + "left": "issue:43504", + "right": "issue:45561", + "accept": false, + "reason": "Beit pretrained-load legacy-field issue and pytest-xdist captured_info.txt race are unrelated." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "The summary is not fully grounded: the packet contains two clear duplicate pairs, not just one. In addition to issue:45310/issue:45478, issue:42673/issue:43704 is the same Qwen3ForCausalLM VRAM leak report with identical titles. The broader characterization of the cluster as mostly v5 import/loading/config regressions is reasonable, but the duplicate-count claim should be corrected." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 13, + "review_unit_index": 32, + "review_unit_count": 42, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-367", + "nodes": [ + "issue:30990", + "issue:35141", + "issue:36010", + "issue:41628", + "issue:42175", + "issue:42898", + "issue:42994", + "issue:43232", + "issue:43278", + "issue:43295", + "issue:43377", + "issue:43479", + "issue:43526", + "issue:43673", + "issue:43723", + "issue:43784", + "issue:43824", + "issue:43874", + "issue:43976", + "issue:44038", + "issue:44060", + "issue:44077", + "issue:44079", + "issue:44164", + "issue:44188", + "issue:44222", + "issue:44242", + "issue:44297", + "issue:44373", + "issue:44448", + "issue:44496", + "issue:44568", + "issue:44623", + "issue:44734", + "issue:44779", + "issue:44792", + "issue:44869", + "issue:44933", + "issue:44964", + "issue:45092", + "issue:45103", + "issue:45125", + "issue:45137", + "issue:45276", + "issue:45310", + "issue:45356", + "issue:45459", + "issue:45561" + ], + "soft_pairs": [ + "issue:42994|issue:44164", + "issue:44496|issue:45561", + "issue:43526|issue:45561", + "issue:44038|issue:45310", + "issue:43232|issue:44792", + "issue:43377|issue:44242", + "issue:30990|issue:35141", + "issue:43673|issue:44188", + "issue:36010|issue:41628", + "issue:42898|issue:44568", + "issue:43479|issue:44079", + "issue:43479|issue:44869", + "issue:44964|issue:45137", + "issue:43976|issue:44933", + "issue:42175|issue:44933", + "issue:43784|issue:43824", + "issue:44623|issue:45310", + "issue:43278|issue:45137", + "issue:44188|issue:44373", + "issue:44060|issue:45125", + "issue:44734|issue:45103", + "issue:42898|issue:45356", + "issue:43723|issue:44568", + "issue:42898|issue:44779", + "issue:45137|issue:45276", + "issue:43874|issue:45092", + "issue:44077|issue:44242", + "issue:42898|issue:44448", + "issue:41628|issue:45459", + "issue:44222|issue:44297", + "issue:42898|issue:43295" + ], + "prepared_review_unit_hash": "c8b871d9dfba5f683f9beb8e264bbe434410e43e33d226e303de8b76ae5d25c4", + "component_budget": { + "node_count": 367, + "item_count": 367, + "soft_pair_count": 1370, + "serialized_chars": 706986, + "estimated_input_tokens": 176747, + "estimated_eval_tokens": 353750 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 37073, + "estimated_input_tokens": 9269, + "estimated_eval_tokens": 18794 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is very heterogeneous: most items are unrelated Transformers v5 regression reports, import/load failures, tokenizer quirks, and a few model-specific bugs. None of the proposed soft pairs look like the same underlying bug/change closely enough to merge, and there are no PRs in the set.", + "confidence": 0.33, + "canonical_issue_reason": "issue:42898 is the closest umbrella regression report because it captures a broad Transformers v5 behavior change, but the cluster overall mixes many different failure modes and subsystems.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:42898 is the best single issue candidate as the broadest v5-regression style report in the set, though it is not a true duplicate representative for most of the other items.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:42994", + "right": "issue:44164", + "accept": false, + "reason": "Both involve save/pretrained, but one is quantized model saving and the other is extra_state handling; different bugs and code paths." + }, + { + "left": "issue:44496", + "right": "issue:45561", + "accept": false, + "reason": "Unrelated: model config loading failure vs pytest-xdist race in testing utils." + }, + { + "left": "issue:43526", + "right": "issue:45561", + "accept": false, + "reason": "Different subsystems: BeitImageProcessorFast label reduction vs a test capture-file race." + }, + { + "left": "issue:44038", + "right": "issue:45310", + "accept": false, + "reason": "Both mention Qwen/MoE and v5, but the specific models and failure modes differ; not clearly the same bug." + }, + { + "left": "issue:43232", + "right": "issue:44792", + "accept": false, + "reason": "Generation kwargs/sync_gpus bug vs janus image-generation test failure; no shared concrete code path." + }, + { + "left": "issue:43377", + "right": "issue:44242", + "accept": false, + "reason": "MIMI padding-mask behavior and load-balancing-loss gating are distinct issues." + }, + { + "left": "issue:30990", + "right": "issue:35141", + "accept": false, + "reason": "Sentence-Transformers loading hang and embedding reinit on resize_token_embeddings are unrelated." + }, + { + "left": "issue:43673", + "right": "issue:44188", + "accept": false, + "reason": "Cache missing in chunked prefill and attention-kernel divergence under torch.compile are separate generation bugs." + }, + { + "left": "issue:36010", + "right": "issue:41628", + "accept": false, + "reason": "Both are import errors, but for different symbols and likely different export regressions." + }, + { + "left": "issue:42898", + "right": "issue:44568", + "accept": false, + "reason": "Tokenizer cleanup semantics and BOS/EOS insertion behavior are different tokenizer regressions." + }, + { + "left": "issue:43479", + "right": "issue:44079", + "accept": false, + "reason": "Config default initialization and ModelOutput key assignment are unrelated." + }, + { + "left": "issue:43479", + "right": "issue:44869", + "accept": false, + "reason": "Phi4 multimodal config init and Whisper timestamp decoding crash are different bugs." + }, + { + "left": "issue:44964", + "right": "issue:45137", + "accept": false, + "reason": "Model loading incompatibility vs DeepSpeed ZeRO3 deque underflow; no overlap in cause." + }, + { + "left": "issue:43976", + "right": "issue:44933", + "accept": false, + "reason": "Python version support issue and nonexistent image_utils import are unrelated." + }, + { + "left": "issue:42175", + "right": "issue:44933", + "accept": false, + "reason": "Missing TensorFlow extra and missing image_utils import are different problems." + }, + { + "left": "issue:43784", + "right": "issue:43824", + "accept": false, + "reason": "Both are import errors, but for different packages/symbols and different root causes." + }, + { + "left": "issue:44623", + "right": "issue:45310", + "accept": false, + "reason": "processor.save_pretrained missing files and Qwen3.5 MoE from_pretrained error are different save/load failures." + }, + { + "left": "issue:43278", + "right": "issue:45137", + "accept": false, + "reason": "Embedding dtype drift in eval and a DeepSpeed deque error are unrelated." + }, + { + "left": "issue:44188", + "right": "issue:44373", + "accept": false, + "reason": "Attention-kernel branching and a wrong docstring are not the same issue." + }, + { + "left": "issue:44060", + "right": "issue:45125", + "accept": false, + "reason": "Incorrect tied-weights warning and missing _tp_plan for tensor parallelism are different model-architecture issues." + }, + { + "left": "issue:44734", + "right": "issue:45103", + "accept": false, + "reason": "Serve KV-cache indexing crash and auto_docstring annotations crash affect different code paths." + }, + { + "left": "issue:42898", + "right": "issue:45356", + "accept": false, + "reason": "Tokenizer cleanup behavior change and Kimi-K2.5 codec handling regression are different tokenizer bugs." + }, + { + "left": "issue:43723", + "right": "issue:44568", + "accept": false, + "reason": "Tokenizer loading failure and add_special_tokens BOS/EOS behavior are not the same bug." + }, + { + "left": "issue:42898", + "right": "issue:44779", + "accept": false, + "reason": "Both are tokenizer regressions, but the observed behavior and likely code paths differ substantially." + }, + { + "left": "issue:45137", + "right": "issue:45276", + "accept": false, + "reason": "DeepSpeed deque error and Gemma4 resize_token_embeddings propagation bug are unrelated." + }, + { + "left": "issue:43874", + "right": "issue:45092", + "accept": false, + "reason": "Missing image-patch helper and old InternVL2 meta-init incompatibility are distinct multimodal issues." + }, + { + "left": "issue:44077", + "right": "issue:44242", + "accept": false, + "reason": "Optional post_init policy and missing load-balancing loss are different model implementation bugs." + }, + { + "left": "issue:42898", + "right": "issue:44448", + "accept": false, + "reason": "Both are v5 regressions, but Pegasus output mismatch is broader generation behavior, not the specific clean_up_tokenization_spaces change." + }, + { + "left": "issue:41628", + "right": "issue:45459", + "accept": false, + "reason": "AutoImageProcessor import failure and hidden tokenizer errors without protobuf are unrelated." + }, + { + "left": "issue:44222", + "right": "issue:44297", + "accept": false, + "reason": "Both touch save_pretrained, but one is FP8 MoE model saving and the other is tokenizer metadata mismatch." + }, + { + "left": "issue:42898", + "right": "issue:43295", + "accept": false, + "reason": "Tokenizer cleanup change and processor/tokenizer image-passing regression are different v5 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": 13, + "review_unit_index": 33, + "review_unit_count": 42, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-367", + "nodes": [ + "issue:22355", + "issue:28282", + "issue:30990", + "issue:33357", + "issue:38175", + "issue:41628", + "issue:42757", + "issue:42831", + "issue:42886", + "issue:42994", + "issue:43066", + "issue:43122", + "issue:43479", + "issue:43576", + "issue:43644", + "issue:43723", + "issue:43784", + "issue:43824", + "issue:44038", + "issue:44117", + "issue:44206", + "issue:44222", + "issue:44261", + "issue:44263", + "issue:44295", + "issue:44297", + "issue:44360", + "issue:44373", + "issue:44442", + "issue:44483", + "issue:44484", + "issue:44485", + "issue:44514", + "issue:44521", + "issue:44623", + "issue:44704", + "issue:44871", + "issue:44913", + "issue:44977", + "issue:44991", + "issue:45092", + "issue:45103", + "issue:45278", + "issue:45290", + "issue:45310", + "issue:45357", + "issue:45406", + "issue:45459" + ], + "soft_pairs": [ + "issue:43479|issue:43644", + "issue:44442|issue:45459", + "issue:44222|issue:44623", + "issue:42831|issue:44623", + "issue:44206|issue:45103", + "issue:22355|issue:28282", + "issue:44297|issue:45310", + "issue:41628|issue:42757", + "issue:44483|issue:45406", + "issue:44222|issue:44484", + "issue:44263|issue:44360", + "issue:43723|issue:44977", + "issue:44263|issue:45092", + "issue:44295|issue:44623", + "issue:44514|issue:45103", + "issue:43479|issue:44913", + "issue:44373|issue:44485", + "issue:43723|issue:44871", + "issue:44261|issue:45092", + "issue:43723|issue:44991", + "issue:43723|issue:44442", + "issue:45092|issue:45357", + "issue:30990|issue:38175", + "issue:42886|issue:42994", + "issue:42831|issue:45310", + "issue:41628|issue:43824", + "issue:42831|issue:44038", + "issue:43576|issue:43784", + "issue:44263|issue:44373", + "issue:42757|issue:45278", + "issue:43122|issue:44521", + "issue:42886|issue:43066", + "issue:44117|issue:44704", + "issue:43122|issue:44514", + "issue:30990|issue:33357", + "issue:44704|issue:45290" + ], + "prepared_review_unit_hash": "f3a1ffd23eae00a1c85c5e5a9f214c8145564cfa8d797e2411e87c15fcde902a", + "component_budget": { + "node_count": 367, + "item_count": 367, + "soft_pair_count": 1370, + "serialized_chars": 706986, + "estimated_input_tokens": 176747, + "estimated_eval_tokens": 353750 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 36, + "serialized_chars": 38110, + "estimated_input_tokens": 9528, + "estimated_eval_tokens": 19312 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is heterogeneous: a few issues are about v5 import/tokenizer regressions, but many others are unrelated (FP8, multimodal processors, chat templates, config serialization, etc.). I don\u2019t see any soft pair that clearly describes the same underlying bug/change.", + "confidence": 0.91, + "canonical_issue_reason": "Issue 45278 is the broadest umbrella in the set, covering the general v5 import-regression theme better than the narrower single-symbol reports.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 45278 is the most representative cluster anchor because it is the most general report of the recurring v5 breakage pattern.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43479", + "right": "issue:43644", + "accept": false, + "reason": "Different bugs: default config initialization for Phi4MultimodalConfig vs junk-filled non-persistent buffers in v5.0.0." + }, + { + "left": "issue:44442", + "right": "issue:45459", + "accept": false, + "reason": "Both involve tokenizer loading, but one is a missing tokenizer class while the other hides protobuf-related decode errors; not the same bug." + }, + { + "left": "issue:44222", + "right": "issue:44623", + "accept": false, + "reason": "FP8 moe save_pretrained failure and processor.save_pretrained missing files are separate save paths with different failure modes." + }, + { + "left": "issue:42831", + "right": "issue:44623", + "accept": false, + "reason": "Accuracy regression in FineGrainedFP8 is unrelated to processor save_pretrained file omissions." + }, + { + "left": "issue:44206", + "right": "issue:45103", + "accept": false, + "reason": "Unsupported center arg in LasrFeatureExtractor and auto_docstring AttributeError are unrelated code paths." + }, + { + "left": "issue:22355", + "right": "issue:28282", + "accept": false, + "reason": "Both are import errors, but one is missing transformers.onnx and the other is missing PyTorch; different root causes." + }, + { + "left": "issue:44297", + "right": "issue:45310", + "accept": false, + "reason": "Tokenizer config mismatch on save is not the same as Qwen3.5 MoE from_pretrained failure." + }, + { + "left": "issue:41628", + "right": "issue:42757", + "accept": false, + "reason": "Different missing imports from different packages (AutoImageProcessor vs is_offline_mode)." + }, + { + "left": "issue:44483", + "right": "issue:45406", + "accept": false, + "reason": "Chat/completions request rejection and Gemma4Processor missing _tokenizer are not the same failure." + }, + { + "left": "issue:44222", + "right": "issue:44484", + "accept": false, + "reason": "Both mention save_pretrained, but one is FP8/moe serialization and the other is shard-size default behavior." + }, + { + "left": "issue:44263", + "right": "issue:44360", + "accept": false, + "reason": "Same subsystem name, but one is about torch.split outputs and the other is about a missing ReLU in the DSA indexer." + }, + { + "left": "issue:43723", + "right": "issue:44977", + "accept": false, + "reason": "Tokenizer-loading regression vs flash-attention generation bug; different code paths." + }, + { + "left": "issue:44263", + "right": "issue:45092", + "accept": false, + "reason": "GLM DSA indexing bug is unrelated to remote-code checkpoint meta-initialization incompatibility." + }, + { + "left": "issue:44295", + "right": "issue:44623", + "accept": false, + "reason": "Position_ids buffer read error is unrelated to missing processor save files." + }, + { + "left": "issue:44514", + "right": "issue:45103", + "accept": false, + "reason": "Qwen2_5_VLProcessor batched apply_chat_template crash is unrelated to auto_docstring annotation handling." + }, + { + "left": "issue:43479", + "right": "issue:44913", + "accept": false, + "reason": "Both are config-related, but they affect different models and different config fields." + }, + { + "left": "issue:44373", + "right": "issue:44485", + "accept": false, + "reason": "Docstring correction for position_ids is not the same as a RoPE implementation bug." + }, + { + "left": "issue:43723", + "right": "issue:44871", + "accept": false, + "reason": "Both concern tokenizer/model loading, but they describe different regressions with different symptoms and targets." + }, + { + "left": "issue:44261", + "right": "issue:45092", + "accept": false, + "reason": "MLA rms_norm_eps precision issue is unrelated to remote-code meta initialization compatibility." + }, + { + "left": "issue:43723", + "right": "issue:44991", + "accept": false, + "reason": "Generic AutoTokenizer load issue and a specific EMBEDDIA/est-roberta tokenizer failure are not clearly the same bug." + }, + { + "left": "issue:43723", + "right": "issue:44442", + "accept": false, + "reason": "Both are tokenizer loading problems, but they point to different tokenizer classes and likely different causes." + }, + { + "left": "issue:45092", + "right": "issue:45357", + "accept": false, + "reason": "Loading incompatibility for old InternVL2 checkpoints is unrelated to Qwen3.5 save_pretrained key serialization." + }, + { + "left": "issue:30990", + "right": "issue:38175", + "accept": false, + "reason": "Sentence Transformers loading hang and zero-probability SigLIP output are different issues." + }, + { + "left": "issue:42886", + "right": "issue:42994", + "accept": false, + "reason": "Offline tokenizer cache loading and quantized model saving failures are unrelated." + }, + { + "left": "issue:42831", + "right": "issue:45310", + "accept": false, + "reason": "FP8 accuracy regression and Qwen3.5 MoE from_pretrained error are different bugs." + }, + { + "left": "issue:41628", + "right": "issue:43824", + "accept": false, + "reason": "Different missing imports from different symbols and model classes." + }, + { + "left": "issue:42831", + "right": "issue:44038", + "accept": false, + "reason": "FineGrainedFP8 accuracy problems are unrelated to Qwen3-VL-Moe loading on v5." + }, + { + "left": "issue:43576", + "right": "issue:43784", + "accept": false, + "reason": "Broken transformers env command and sentence-transformers import NameError are unrelated." + }, + { + "left": "issue:44263", + "right": "issue:44373", + "accept": false, + "reason": "Indexing behavior and docstring text are not the same underlying bug." + }, + { + "left": "issue:42757", + "right": "issue:45278", + "accept": false, + "reason": "A specific missing hub symbol is not enough to call it the same bug as a broad 'many import errors' report." + }, + { + "left": "issue:43122", + "right": "issue:44521", + "accept": false, + "reason": "Different tokenization problems: reload changes tokenization vs assistant mask handling in multimodal chat templates." + }, + { + "left": "issue:42886", + "right": "issue:43066", + "accept": false, + "reason": "Offline cache loading failure and tokenizer decoder type mismatch are unrelated." + }, + { + "left": "issue:44117", + "right": "issue:44704", + "accept": false, + "reason": "TOKENIZER_MAPPING_NAMES returning None and AutoProcessor kwargs forwarding are different failure points." + }, + { + "left": "issue:43122", + "right": "issue:44514", + "accept": false, + "reason": "Tokenization changes across versions and batched chat-template crashes are different bugs." + }, + { + "left": "issue:30990", + "right": "issue:33357", + "accept": false, + "reason": "Loading stuck vs MacOS bus error on a different model/version; no same code-path evidence." + }, + { + "left": "issue:44704", + "right": "issue:45290", + "accept": false, + "reason": "AutoProcessor cached_file kwargs issue is unrelated to apply_chat_template crashing on tool-call messages." + } + ] + }, + "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": 13, + "review_unit_index": 34, + "review_unit_count": 42, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-367", + "nodes": [ + "issue:29127", + "issue:30990", + "issue:36296", + "issue:39401", + "issue:41628", + "issue:42222", + "issue:42371", + "issue:42757", + "issue:42831", + "issue:42907", + "issue:43116", + "issue:43122", + "issue:43232", + "issue:43278", + "issue:43408", + "issue:43441", + "issue:43452", + "issue:43475", + "issue:43824", + "issue:43937", + "issue:43976", + "issue:43992", + "issue:44038", + "issue:44222", + "issue:44261", + "issue:44263", + "issue:44295", + "issue:44297", + "issue:44351", + "issue:44360", + "issue:44373", + "issue:44485", + "issue:44488", + "issue:44521", + "issue:44625", + "issue:44704", + "issue:44743", + "issue:44779", + "issue:44829", + "issue:44945", + "issue:44964", + "issue:44977", + "issue:45200", + "issue:45216", + "issue:45245", + "issue:45278", + "issue:45478" + ], + "soft_pairs": [ + "issue:42371|issue:44263", + "issue:44295|issue:44297", + "issue:42907|issue:44945", + "issue:43976|issue:44945", + "issue:43232|issue:45245", + "issue:43232|issue:44743", + "issue:39401|issue:44488", + "issue:42907|issue:43937", + "issue:43408|issue:43475", + "issue:44625|issue:44829", + "issue:41628|issue:45278", + "issue:44360|issue:44485", + "issue:43116|issue:45200", + "issue:29127|issue:30990", + "issue:39401|issue:44521", + "issue:44261|issue:44373", + "issue:43122|issue:44779", + "issue:44038|issue:44297", + "issue:42371|issue:42831", + "issue:44038|issue:44295", + "issue:42907|issue:44964", + "issue:36296|issue:43116", + "issue:42907|issue:43452", + "issue:44038|issue:44222", + "issue:43992|issue:44704", + "issue:45216|issue:45478", + "issue:43122|issue:44977", + "issue:43824|issue:44351", + "issue:43278|issue:44945", + "issue:42907|issue:44222", + "issue:42757|issue:44351", + "issue:42222|issue:43441" + ], + "prepared_review_unit_hash": "cd100f53f5dddd64782cb1f70e28d97e95d7d381ca0536995a4b498dc79e54ef", + "component_budget": { + "node_count": 367, + "item_count": 367, + "soft_pair_count": 1370, + "serialized_chars": 706986, + "estimated_input_tokens": 176747, + "estimated_eval_tokens": 353750 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 32, + "serialized_chars": 36173, + "estimated_input_tokens": 9044, + "estimated_eval_tokens": 18344 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly heterogeneous: import regressions, tokenizer issues, generation bugs, multimodal/model loading problems, and FP8/parallelism issues are all mixed together. I did not find any high-confidence duplicate pair; the closest look-alike is a tokenizer regression pair, but the titles are still too broad to merge confidently.", + "confidence": 0.43, + "canonical_issue_reason": "issue:43824 is the most central concrete regression in the set, with multiple inbound references and a clear import-failure symptom; the rest of the items are too diverse to form one true duplicate cluster.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43824 is the best representative issue here because it is specific, user-facing, and has the strongest linkage among the listed issues.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:42371", + "right": "issue:44263", + "accept": false, + "reason": "TF32 behavior guidance vs a torch.split indexing bug; unrelated symptoms and code paths." + }, + { + "left": "issue:44295", + "right": "issue:44297", + "accept": false, + "reason": "Buffer access error for position_ids vs tokenizer_class mismatch on save; different subsystems." + }, + { + "left": "issue:42907", + "right": "issue:44945", + "accept": false, + "reason": "Dequantized model save failure vs incorrect output under pipeline parallelism; not the same bug." + }, + { + "left": "issue:43976", + "right": "issue:44945", + "accept": false, + "reason": "Python version compatibility issue vs pipeline-parallel output bug; different failure modes." + }, + { + "left": "issue:43232", + "right": "issue:45245", + "accept": false, + "reason": "Generation kwargs/cache bug vs category-cardinality runtime error; unrelated." + }, + { + "left": "issue:43232", + "right": "issue:44743", + "accept": false, + "reason": "Both touch generation state, but one is sync_gpus/update_model_kwargs and the other is recurrent-state reset in modular_qwen3_5; too distinct." + }, + { + "left": "issue:39401", + "right": "issue:44488", + "accept": false, + "reason": "Qwen3 tokenizer offset_mapping regression vs model load failure for a different model; not the same issue." + }, + { + "left": "issue:42907", + "right": "issue:43937", + "accept": false, + "reason": "Dequantized save failure vs invalid GenerationConfig for GLM-5; different models and code paths." + }, + { + "left": "issue:43408", + "right": "issue:43475", + "accept": false, + "reason": "Model-type warning during instantiation vs missing attribute on Sam3VisionEncoderOutput; related family, but different concrete bugs." + }, + { + "left": "issue:44625", + "right": "issue:44829", + "accept": false, + "reason": "num_labels propagation bug vs flash_attention_3 degenerate training; unrelated." + }, + { + "left": "issue:41628", + "right": "issue:45278", + "accept": false, + "reason": "Specific AutoImageProcessor import error vs broad post-upgrade import errors; too generic to treat as the same bug." + }, + { + "left": "issue:44360", + "right": "issue:44485", + "accept": false, + "reason": "DSA indexer ReLU discussion vs GLM-5 RoPE implementation; unrelated." + }, + { + "left": "issue:43116", + "right": "issue:45200", + "accept": false, + "reason": "Example-script multi-label empty results vs mm_token_type_ids defaulting; different model/task paths." + }, + { + "left": "issue:29127", + "right": "issue:30990", + "accept": false, + "reason": "LayoutLMv3 error-message clarity vs Sentence Transformers loading hang; unrelated." + }, + { + "left": "issue:39401", + "right": "issue:44521", + "accept": false, + "reason": "Tokenizer offset_mapping regression vs multimodal assistant_mask bug; different functionality." + }, + { + "left": "issue:44261", + "right": "issue:44373", + "accept": false, + "reason": "LayerNorm precision/config issue vs docstring mismatch; not duplicates." + }, + { + "left": "issue:43122", + "right": "issue:44779", + "accept": false, + "reason": "Both mention tokenizer regressions in v5, but one is broad and the other is Deepseek-specific; not enough evidence they are the same underlying bug." + }, + { + "left": "issue:44038", + "right": "issue:44297", + "accept": false, + "reason": "Qwen3-VL-Moe bug vs tokenizer save metadata mismatch; different symptoms." + }, + { + "left": "issue:42371", + "right": "issue:42831", + "accept": false, + "reason": "TF32 configuration warning vs FineGrainedFP8 accuracy issue; unrelated." + }, + { + "left": "issue:44038", + "right": "issue:44295", + "accept": false, + "reason": "Qwen3-VL-Moe bug vs position_ids buffer read error; different code paths." + }, + { + "left": "issue:42907", + "right": "issue:44964", + "accept": false, + "reason": "Dequantized save failure vs Phi-4 multimodal load failure; not the same bug." + }, + { + "left": "issue:36296", + "right": "issue:43116", + "accept": false, + "reason": "Tensor parallel training bug vs classification example returning empty results; unrelated." + }, + { + "left": "issue:42907", + "right": "issue:43452", + "accept": false, + "reason": "Saving dequantized Ministral/Devstral models vs gguf_file handling for from_pretrained; different failure modes." + }, + { + "left": "issue:44038", + "right": "issue:44222", + "accept": false, + "reason": "Qwen3-VL-Moe bug vs FP8 save_pretrained issue; too broad and not concretely the same bug." + }, + { + "left": "issue:43992", + "right": "issue:44704", + "accept": false, + "reason": "Missing embed_tokens.weight in UMT5Encoder loading vs AutoProcessor kwargs not passed to cached_file; unrelated." + }, + { + "left": "issue:45216", + "right": "issue:45478", + "accept": false, + "reason": "Qwen3.5 save_pretrained regression vs Qwen3.5 Moe from_pretrained error; same family but different operations and likely different fixes." + }, + { + "left": "issue:43122", + "right": "issue:44977", + "accept": false, + "reason": "Tokenizer regression vs flash-attention generation bug; not the same code path." + }, + { + "left": "issue:43824", + "right": "issue:44351", + "accept": false, + "reason": "Missing Qwen2_5_VLForConditionalGeneration import vs missing HybridCache import; both import errors, but distinct symbols and likely distinct causes." + }, + { + "left": "issue:43278", + "right": "issue:44945", + "accept": false, + "reason": "Embedding dtype mismatch between train/eval vs incorrect output under pipeline parallelism; different problems." + }, + { + "left": "issue:42907", + "right": "issue:44222", + "accept": false, + "reason": "Dequantized save failure vs FP8 save_pretrained moe; both involve saving, but not the same concrete bug." + }, + { + "left": "issue:42757", + "right": "issue:44351", + "accept": false, + "reason": "Missing is_offline_mode from huggingface_hub vs missing HybridCache from transformers; both imports, but different packages and errors." + }, + { + "left": "issue:42222", + "right": "issue:43441", + "accept": false, + "reason": "Broken vitpose models vs Ministral-3 FlashAttention failure; unrelated models and 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": 13, + "review_unit_index": 35, + "review_unit_count": 42, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-367", + "nodes": [ + "issue:22355", + "issue:29127", + "issue:33453", + "issue:38617", + "issue:39401", + "issue:41628", + "issue:42371", + "issue:42673", + "issue:43064", + "issue:43116", + "issue:43421", + "issue:43450", + "issue:43493", + "issue:43550", + "issue:43673", + "issue:43784", + "issue:43825", + "issue:43827", + "issue:43874", + "issue:43906", + "issue:43931", + "issue:43937", + "issue:43976", + "issue:43986", + "issue:44038", + "issue:44075", + "issue:44222", + "issue:44291", + "issue:44484", + "issue:44485", + "issue:44779", + "issue:44829", + "issue:44936", + "issue:44945", + "issue:44977", + "issue:45071", + "issue:45072", + "issue:45083", + "issue:45216", + "issue:45278", + "issue:45310", + "issue:45341", + "issue:45405", + "issue:45446", + "issue:45478", + "issue:45561", + "issue:45563" + ], + "soft_pairs": [ + "issue:43784|issue:45446", + "issue:42673|issue:45072", + "issue:43064|issue:43673", + "issue:38617|issue:45278", + "issue:43550|issue:43784", + "issue:22355|issue:29127", + "issue:43937|issue:44485", + "issue:43116|issue:43493", + "issue:42371|issue:43874", + "issue:44075|issue:44936", + "issue:44222|issue:45478", + "issue:44038|issue:45478", + "issue:43116|issue:43825", + "issue:43976|issue:45446", + "issue:43937|issue:45563", + "issue:39401|issue:44779", + "issue:33453|issue:39401", + "issue:43931|issue:44945", + "issue:45083|issue:45446", + "issue:45341|issue:45561", + "issue:43937|issue:44977", + "issue:43827|issue:45446", + "issue:42673|issue:43906", + "issue:44291|issue:45071", + "issue:45405|issue:45446", + "issue:43450|issue:43986", + "issue:44829|issue:45216", + "issue:44484|issue:45310", + "issue:39401|issue:43421", + "issue:38617|issue:41628" + ], + "prepared_review_unit_hash": "168ecd36525d07682a988987201d2484fd63639928bfab1326aa9931f811aab2", + "component_budget": { + "node_count": 367, + "item_count": 367, + "soft_pair_count": 1370, + "serialized_chars": 706986, + "estimated_input_tokens": 176747, + "estimated_eval_tokens": 353750 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 30, + "serialized_chars": 35833, + "estimated_input_tokens": 8959, + "estimated_eval_tokens": 18174 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These items are mostly unrelated regression reports across different areas (imports, tokenizers, generation, training, multimodal, testing). I do not see any soft pair that clearly describes the same underlying bug/change, so I would not merge any of the suggested edges. If one issue must serve as an umbrella, 45278 is the broadest open import-regression report.", + "confidence": 0.77, + "canonical_issue_reason": "issue:45278 is the broadest open report in the set and the only plausible umbrella-style tracker ('many import errors after update from 4.57.0 to 5.5.0').", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45278 is the most general/open issue and the closest thing to a cluster sink, even though the rest of the set is too heterogeneous for true duplication.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43784", + "right": "issue:45446", + "accept": false, + "reason": "Different failures: import-time NameError in sentence-transformers integration vs a PyTorch version-check bug in flex_attention." + }, + { + "left": "issue:42673", + "right": "issue:45072", + "accept": false, + "reason": "Unrelated problems: VRAM leak in Qwen3ForCausalLM thread usage vs dtype mismatches in inference tests." + }, + { + "left": "issue:43064", + "right": "issue:43673", + "accept": false, + "reason": "Different code paths: FSDP2/PEFT optimizer-state issue vs GenerationMixin cache handling during chunked prefill." + }, + { + "left": "issue:38617", + "right": "issue:45278", + "accept": false, + "reason": "45278 is a broad import-error report, but 38617 is one specific missing symbol; not the same concrete bug." + }, + { + "left": "issue:43550", + "right": "issue:43784", + "accept": false, + "reason": "Different subsystems and symptoms: torch.compile/SDPA failure for Bamba vs sentence-transformers import NameError." + }, + { + "left": "issue:22355", + "right": "issue:29127", + "accept": false, + "reason": "No shared underlying bug: ONNX module import failure vs LayoutLMv3 error-message clarity issue." + }, + { + "left": "issue:43937", + "right": "issue:44485", + "accept": false, + "reason": "Both mention GLM-5, but one is invalid GenerationConfig handling and the other is a RoPE implementation discussion." + }, + { + "left": "issue:43116", + "right": "issue:43493", + "accept": false, + "reason": "Different artifacts: example script multi-label classification bug vs SigLIP2 model discrepancy with JAX." + }, + { + "left": "issue:42371", + "right": "issue:43874", + "accept": false, + "reason": "TF32-behavior API guidance and a missing image-patch helper in GLM46V are unrelated." + }, + { + "left": "issue:44075", + "right": "issue:44936", + "accept": false, + "reason": "Optimizer argument handling is unrelated to trainer.evaluate() failing after training." + }, + { + "left": "issue:44222", + "right": "issue:45478", + "accept": false, + "reason": "Both touch MoE/Qwen3.5, but one is FP8 save_pretrained behavior and the other is a from_pretrained load error." + }, + { + "left": "issue:44038", + "right": "issue:45478", + "accept": false, + "reason": "Different Qwen3-VL/MoE reports with different failure modes; not the same concrete bug." + }, + { + "left": "issue:43116", + "right": "issue:43825", + "accept": false, + "reason": "Example-script classification bug is unrelated to the pipeline() translation-task warning." + }, + { + "left": "issue:43976", + "right": "issue:45446", + "accept": false, + "reason": "Python-version compatibility regression and a flex_attention import/version check are not the same issue." + }, + { + "left": "issue:43937", + "right": "issue:45563", + "accept": false, + "reason": "GenerationConfig invalidity is unrelated to a stale warning in paged generate()." + }, + { + "left": "issue:39401", + "right": "issue:44779", + "accept": false, + "reason": "Both are tokenizer regressions, but they affect different models and different observable failures." + }, + { + "left": "issue:33453", + "right": "issue:39401", + "accept": false, + "reason": "Generic tokenizer-loading regression and Qwen3 offset_mapping bug are too different to treat as one fix." + }, + { + "left": "issue:43931", + "right": "issue:44945", + "accept": false, + "reason": "Weight-shape mismatch during model loading is unrelated to incorrect output under pipeline parallelism." + }, + { + "left": "issue:45083", + "right": "issue:45446", + "accept": false, + "reason": "Helper-function behavior in qwen3_omni_moe and a PyTorch version-check bug in flex_attention are different code paths." + }, + { + "left": "issue:45341", + "right": "issue:45561", + "accept": false, + "reason": "A generic testing_utils bug and a pytest-xdist race on captured_info.txt are not the same underlying defect." + }, + { + "left": "issue:43937", + "right": "issue:44977", + "accept": false, + "reason": "Different generation bugs: invalid GenerationConfig vs flash-attention-related generation failure in Qwen3.5." + }, + { + "left": "issue:43827", + "right": "issue:45446", + "accept": false, + "reason": "Docs still referencing pipeline() after v5 removals is unrelated to the flex_attention version-check issue." + }, + { + "left": "issue:42673", + "right": "issue:43906", + "accept": false, + "reason": "The second item is only an isolated reproduction of another issue; there is no evidence it matches the VRAM leak." + }, + { + "left": "issue:44291", + "right": "issue:45071", + "accept": false, + "reason": "Unexpected _is_hf_initialized argument in init_empty_weights and PreTrainedConfig type-checking breakage are distinct loading/type-system regressions." + }, + { + "left": "issue:45405", + "right": "issue:45446", + "accept": false, + "reason": "PEFT version pinning problem is unrelated to a PyTorch version check for flex_attention." + }, + { + "left": "issue:43450", + "right": "issue:43986", + "accept": false, + "reason": "Batched video-processor shape bug and AutoProcessor crash without torchvision are different failures." + }, + { + "left": "issue:44829", + "right": "issue:45216", + "accept": false, + "reason": "Flash-attention-3 training degeneration and save_pretrained checkpoint correctness are not the same bug." + }, + { + "left": "issue:44484", + "right": "issue:45310", + "accept": false, + "reason": "A question about max_shard_size and a Qwen3.5 MoE from_pretrained error are unrelated." + }, + { + "left": "issue:38617", + "right": "issue:41628", + "accept": false, + "reason": "Both are import errors, but one is a missing internal symbol and the other is a missing top-level export; no same fix is evident." + } + ] + }, + "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": 13, + "review_unit_index": 36, + "review_unit_count": 42, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-367", + "nodes": [ + "issue:38617", + "issue:41628", + "issue:42175", + "issue:42222", + "issue:42371", + "issue:42617", + "issue:42673", + "issue:43012", + "issue:43066", + "issue:43097", + "issue:43122", + "issue:43335", + "issue:43408", + "issue:43441", + "issue:43452", + "issue:43577", + "issue:43582", + "issue:43673", + "issue:43742", + "issue:43874", + "issue:43931", + "issue:43937", + "issue:43986", + "issue:44038", + "issue:44117", + "issue:44220", + "issue:44261", + "issue:44263", + "issue:44368", + "issue:44410", + "issue:44483", + "issue:44484", + "issue:44485", + "issue:44493", + "issue:44496", + "issue:44829", + "issue:44933", + "issue:44987", + "issue:45003", + "issue:45071", + "issue:45081", + "issue:45278", + "issue:45310", + "issue:45341", + "issue:45459", + "issue:45464", + "issue:45478", + "issue:45542" + ], + "soft_pairs": [ + "issue:44117|issue:45081", + "issue:44410|issue:45310", + "issue:42673|issue:43582", + "issue:43408|issue:43742", + "issue:42673|issue:43577", + "issue:42222|issue:45542", + "issue:42371|issue:44261", + "issue:43937|issue:44496", + "issue:38617|issue:44220", + "issue:42175|issue:42222", + "issue:44483|issue:45464", + "issue:42617|issue:45341", + "issue:44933|issue:45278", + "issue:44484|issue:45478", + "issue:43066|issue:43122", + "issue:43673|issue:45341", + "issue:43097|issue:44493", + "issue:44410|issue:45478", + "issue:42222|issue:45459", + "issue:43874|issue:44261", + "issue:43441|issue:44829", + "issue:44263|issue:44485", + "issue:43012|issue:43408", + "issue:43986|issue:44484", + "issue:41628|issue:45071", + "issue:45071|issue:45542", + "issue:45003|issue:45341", + "issue:43931|issue:44410", + "issue:44117|issue:44987", + "issue:43097|issue:43452", + "issue:43986|issue:45341", + "issue:43931|issue:44038", + "issue:43335|issue:45071", + "issue:43408|issue:44368", + "issue:43874|issue:44263" + ], + "prepared_review_unit_hash": "c861c0621562764fca53427cb4152bc0b96a8fcb801cd1d34d0077ffe49bfed8", + "component_budget": { + "node_count": 367, + "item_count": 367, + "soft_pair_count": 1370, + "serialized_chars": 706986, + "estimated_input_tokens": 176747, + "estimated_eval_tokens": 353750 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 35, + "serialized_chars": 38012, + "estimated_input_tokens": 9503, + "estimated_eval_tokens": 19262 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These items are mostly unrelated single-issue reports; none of the soft pairs look like the same underlying bug/change closely enough to merge. The closest thematic overlap is around import/regression reports, but even there the concrete failure modes differ.", + "confidence": 0.88, + "canonical_issue_reason": "If one issue must anchor the cluster, issue 45278 is the broadest import-regression report ('Many import errors after update from 4.57.0 to 5.5.0'), but it is still not a true duplicate umbrella for the others.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 45278 is the best global representative because it is the most general and user-facing import regression in the set, making it the most plausible triage hub among the issue-only items.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44117", + "right": "issue:45081", + "accept": false, + "reason": "Both are tokenizer-loading failures, but one is a mapping/assumption bug and the other is a Mistral regex patch crash; different code paths." + }, + { + "left": "issue:44410", + "right": "issue:45310", + "accept": false, + "reason": "Both involve Qwen-family models, but the reported failures are different model/config loading problems, not the same concrete bug." + }, + { + "left": "issue:42673", + "right": "issue:43582", + "accept": false, + "reason": "VRAM leakage in Qwen3ForCausalLM is unrelated to an Apple Silicon TypeError in caching_allocator_warmup." + }, + { + "left": "issue:43408", + "right": "issue:43742", + "accept": false, + "reason": "A SAM tracker/model-type warning is not the same as a MobileLLM key error during loading." + }, + { + "left": "issue:42673", + "right": "issue:43577", + "accept": false, + "reason": "Threaded VRAM leakage and Blip2 dtype propagation are different runtime issues." + }, + { + "left": "issue:42222", + "right": "issue:45542", + "accept": false, + "reason": "VitPose model breakage and missing TensorFlow backend handling are unrelated failures." + }, + { + "left": "issue:42371", + "right": "issue:44261", + "accept": false, + "reason": "TF32 API warnings and MLA rms_norm_eps precision drift are different configuration/precision issues." + }, + { + "left": "issue:43937", + "right": "issue:44496", + "accept": false, + "reason": "One is invalid GenerationConfig handling; the other is missing model_type during model resolution." + }, + { + "left": "issue:38617", + "right": "issue:44220", + "accept": false, + "reason": "ImportError for a config utility and a speech feature extraction bug do not share a code path." + }, + { + "left": "issue:42175", + "right": "issue:42222", + "accept": false, + "reason": "Backend dependency installation and VitPose model breakage are not the same bug." + }, + { + "left": "issue:44483", + "right": "issue:45464", + "accept": false, + "reason": "Both mention chat/completions, but one is a general API acceptance problem and the other is a streaming inference failure on a specific model." + }, + { + "left": "issue:42617", + "right": "issue:45341", + "accept": false, + "reason": "A script execution failure and a testing_utils bug are unrelated." + }, + { + "left": "issue:44933", + "right": "issue:45278", + "accept": false, + "reason": "The first is a specific missing import from image_utils; the second is a broad post-upgrade import-regression report, so they are not clearly the same bug." + }, + { + "left": "issue:44484", + "right": "issue:45478", + "accept": false, + "reason": "Sharding-size behavior in save_pretrained and a Qwen3.5-Moe from_pretrained error are unrelated." + }, + { + "left": "issue:43066", + "right": "issue:43122", + "accept": false, + "reason": "Tokenizer decoder type mismatch and changed tokenization output are related in theme but not the same concrete failure." + }, + { + "left": "issue:43673", + "right": "issue:45341", + "accept": false, + "reason": "Generation cache handling in v5.0.0 and a testing utility bug are unrelated." + }, + { + "left": "issue:43097", + "right": "issue:44493", + "accept": false, + "reason": "Removing tie_embeddings_and_encoder_decoder is a config/API change, while unexpected position-id keys are a separate model loading warning." + }, + { + "left": "issue:44410", + "right": "issue:45478", + "accept": false, + "reason": "Both are Qwen-related, but they concern different model families and different load-time failures." + }, + { + "left": "issue:42222", + "right": "issue:45459", + "accept": false, + "reason": "VitPose model breakage and tokenizer protobuf error masking are unrelated." + }, + { + "left": "issue:43874", + "right": "issue:44261", + "accept": false, + "reason": "An image processor method omission is not the same as an MLA normalization-precision bug." + }, + { + "left": "issue:43441", + "right": "issue:44829", + "accept": false, + "reason": "FlashAttention failure on Ministral-3 and degenerate training with flash_attention_3 are distinct code-path problems." + }, + { + "left": "issue:44263", + "right": "issue:44485", + "accept": false, + "reason": "A torch.split return-value issue is not the same as a GLM-5 RoPE implementation discussion." + }, + { + "left": "issue:43012", + "right": "issue:43408", + "accept": false, + "reason": "A bfloat16 compilation warning and a SAM model-type warning are unrelated." + }, + { + "left": "issue:43986", + "right": "issue:44484", + "accept": false, + "reason": "AutoProcessor without torchvision and save_pretrained shard sizing are different areas entirely." + }, + { + "left": "issue:41628", + "right": "issue:45071", + "accept": false, + "reason": "An AutoImageProcessor import error and a PreTrainedConfig type-checking regression do not share the same underlying bug." + }, + { + "left": "issue:45071", + "right": "issue:45542", + "accept": false, + "reason": "PreTrainedConfig type checking and missing TensorFlow backend detection are unrelated." + }, + { + "left": "issue:45003", + "right": "issue:45341", + "accept": false, + "reason": "Unsafe sys.modules access in modeling_utils and a testing_utils bug are not duplicates." + }, + { + "left": "issue:43931", + "right": "issue:44410", + "accept": false, + "reason": "Qwen3-VL weight-shape mismatch and missing projections in qwen3next are different model-loading failures." + }, + { + "left": "issue:44117", + "right": "issue:44987", + "accept": false, + "reason": "A tokenizer mapping None/assumption bug is not the same as a package load failure for physical-intelligence/fast." + }, + { + "left": "issue:43097", + "right": "issue:43452", + "accept": false, + "reason": "Removing a config attribute and gguf_file handling breakage are unrelated." + }, + { + "left": "issue:43986", + "right": "issue:45341", + "accept": false, + "reason": "Video-model loading without torchvision and a testing utility bug are unrelated." + }, + { + "left": "issue:43931", + "right": "issue:44038", + "accept": false, + "reason": "A Qwen3-VL weight-shape mismatch and a Qwen3-VL-Moe bug are different failures." + }, + { + "left": "issue:43335", + "right": "issue:45071", + "accept": false, + "reason": "A SwitchTransformers sparse-layer construction bug is unrelated to PreTrainedConfig type checking." + }, + { + "left": "issue:43408", + "right": "issue:44368", + "accept": false, + "reason": "A sam3 tracker/model-type warning and a tie_word_embeddings warning in Qwen3.5 are not the same issue." + }, + { + "left": "issue:43874", + "right": "issue:44263", + "accept": false, + "reason": "An image processor helper method missing and a torch.split return-value 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": 13, + "review_unit_index": 37, + "review_unit_count": 42, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-367", + "nodes": [ + "issue:38617", + "issue:42757", + "issue:42886", + "issue:42907", + "issue:43012", + "issue:43064", + "issue:43066", + "issue:43097", + "issue:43208", + "issue:43352", + "issue:43381", + "issue:43502", + "issue:43525", + "issue:43526", + "issue:43531", + "issue:43550", + "issue:43618", + "issue:43824", + "issue:43825", + "issue:43873", + "issue:43901", + "issue:43906", + "issue:43976", + "issue:43986", + "issue:43992", + "issue:44038", + "issue:44062", + "issue:44117", + "issue:44261", + "issue:44315", + "issue:44351", + "issue:44368", + "issue:44380", + "issue:44410", + "issue:44483", + "issue:44484", + "issue:44485", + "issue:44514", + "issue:44655", + "issue:44704", + "issue:44829", + "issue:45071", + "issue:45103", + "issue:45245", + "issue:45341", + "issue:45372", + "issue:45375", + "issue:45479" + ], + "soft_pairs": [ + "issue:44829|issue:45245", + "issue:43976|issue:45071", + "issue:43986|issue:45372", + "issue:43531|issue:45341", + "issue:43064|issue:44315", + "issue:38617|issue:42757", + "issue:44261|issue:44485", + "issue:44410|issue:45375", + "issue:38617|issue:44351", + "issue:44038|issue:44410", + "issue:44062|issue:44117", + "issue:44704|issue:45103", + "issue:43986|issue:44117", + "issue:43502|issue:43992", + "issue:43097|issue:43873", + "issue:43992|issue:44484", + "issue:43824|issue:43986", + "issue:43066|issue:43906", + "issue:43825|issue:44655", + "issue:43352|issue:44380", + "issue:38617|issue:43097", + "issue:43097|issue:43901", + "issue:43012|issue:43550", + "issue:43012|issue:44368", + "issue:42886|issue:43381", + "issue:43208|issue:43550", + "issue:42907|issue:44655", + "issue:43525|issue:44117", + "issue:44380|issue:45479", + "issue:43352|issue:43976", + "issue:43064|issue:44380", + "issue:43208|issue:43526", + "issue:44483|issue:44514", + "issue:43618|issue:44380" + ], + "prepared_review_unit_hash": "ee2bd8d3a5159d2ec7e9ddbc220dfa194d25d77ef5cc67fe0eecb8438538b695", + "component_budget": { + "node_count": 367, + "item_count": 367, + "soft_pair_count": 1370, + "serialized_chars": 706986, + "estimated_input_tokens": 176747, + "estimated_eval_tokens": 353750 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 34, + "serialized_chars": 37815, + "estimated_input_tokens": 9454, + "estimated_eval_tokens": 19164 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is mostly a grab bag of distinct Transformers bugs/regressions (imports, model configs, loaders, training, and docs). The soft edges are all superficial similarities rather than true duplicates.", + "confidence": 0.96, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": "issue:43097 is the closest cluster anchor because it is a broad v5 API regression with some inbound references, but it still does not subsume the other reports.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44829", + "right": "issue:45245", + "accept": false, + "reason": "Different failures: flash-attention training degeneration vs category-cardinality runtime error." + }, + { + "left": "issue:43976", + "right": "issue:45071", + "accept": false, + "reason": "Python version compatibility problem vs PretrainedConfig type-checking regression." + }, + { + "left": "issue:43986", + "right": "issue:45372", + "accept": false, + "reason": "Different loading paths and missing dependencies: torchvision vs mistral_common import." + }, + { + "left": "issue:43531", + "right": "issue:45341", + "accept": false, + "reason": "Unrelated bugs in Qwen3-MoE sliding windows and testing utilities." + }, + { + "left": "issue:43064", + "right": "issue:44315", + "accept": false, + "reason": "Different training issues: FSDP2 optimizer state corruption vs Liger kernel not applied." + }, + { + "left": "issue:38617", + "right": "issue:42757", + "accept": false, + "reason": "Both are import errors, but for different symbols and different modules." + }, + { + "left": "issue:44261", + "right": "issue:44485", + "accept": false, + "reason": "Different model-specific bugs: RMS norm precision vs GLM-5 RoPE implementation." + }, + { + "left": "issue:44410", + "right": "issue:45375", + "accept": false, + "reason": "Both involve Qwen-related configs, but the missing projections and missing config field are separate bugs." + }, + { + "left": "issue:38617", + "right": "issue:44351", + "accept": false, + "reason": "Separate missing-export import errors with different affected APIs." + }, + { + "left": "issue:44038", + "right": "issue:44410", + "accept": false, + "reason": "Different Qwen model families and different broken code paths." + }, + { + "left": "issue:44062", + "right": "issue:44117", + "accept": false, + "reason": "Tokenizer construction error vs tokenizer-mapping None handling; not the same defect." + }, + { + "left": "issue:44704", + "right": "issue:45103", + "accept": false, + "reason": "AutoProcessor forwarding bug vs auto_docstring crash; unrelated subsystems." + }, + { + "left": "issue:43986", + "right": "issue:44117", + "accept": false, + "reason": "Video-model loading dependency issue vs tokenizer mapping assumptions." + }, + { + "left": "issue:43502", + "right": "issue:43992", + "accept": false, + "reason": "Offline/local-file loading leak vs missing encoder weights on load." + }, + { + "left": "issue:43097", + "right": "issue:43873", + "accept": false, + "reason": "Broad v5 API removal vs quantization/offloading behavior; different problems." + }, + { + "left": "issue:43992", + "right": "issue:44484", + "accept": false, + "reason": "Load-state mismatch vs save_pretrained shard-size default question." + }, + { + "left": "issue:43824", + "right": "issue:43986", + "accept": false, + "reason": "Different import/load failures affecting different processor/model paths." + }, + { + "left": "issue:43066", + "right": "issue:43906", + "accept": false, + "reason": "Tokenizer decoder-type regression vs isolated reproduction of another issue." + }, + { + "left": "issue:43825", + "right": "issue:44655", + "accept": false, + "reason": "Pipeline error-message bug vs pipeline save_pretrained failure." + }, + { + "left": "issue:43352", + "right": "issue:44380", + "accept": false, + "reason": "Different attention-backend issues on different models and with different symptoms." + }, + { + "left": "issue:38617", + "right": "issue:43097", + "accept": false, + "reason": "ImportError for a symbol vs v5 embedding-link API removal; unrelated." + }, + { + "left": "issue:43097", + "right": "issue:43901", + "accept": false, + "reason": "Core API removal vs documentation mismatch for return_all_scores." + }, + { + "left": "issue:43012", + "right": "issue:43550", + "accept": false, + "reason": "Compilation warning on bfloat16 vs torch.compile SDPA failure; distinct bugs." + }, + { + "left": "issue:43012", + "right": "issue:44368", + "accept": false, + "reason": "Both are warnings, but about different APIs and causes." + }, + { + "left": "issue:42886", + "right": "issue:43381", + "accept": false, + "reason": "Offline cache loading bug vs gradient-checkpointing eval-mode limitation." + }, + { + "left": "issue:43208", + "right": "issue:43550", + "accept": false, + "reason": "xLSTM training blockers vs Bamba compile/SDPA failure." + }, + { + "left": "issue:42907", + "right": "issue:44655", + "accept": false, + "reason": "Both involve saving, but dequantized model save and pipeline save are different code paths." + }, + { + "left": "issue:43525", + "right": "issue:44117", + "accept": false, + "reason": "Missing config attribute vs tokenizer-mapping None handling." + }, + { + "left": "issue:44380", + "right": "issue:45479", + "accept": false, + "reason": "Attention scaling backend regression vs degenerate classification loss; not the same bug." + }, + { + "left": "issue:43352", + "right": "issue:43976", + "accept": false, + "reason": "Model/backend compatibility issue vs Python version packaging issue." + }, + { + "left": "issue:43064", + "right": "issue:44380", + "accept": false, + "reason": "FSDP optimizer-state corruption vs GPT2 attention scaling ignored in backends." + }, + { + "left": "issue:43208", + "right": "issue:43526", + "accept": false, + "reason": "Different model families and unrelated training vs image-processor bugs." + }, + { + "left": "issue:44483", + "right": "issue:44514", + "accept": false, + "reason": "Chat-completions request rejection vs batched processor crash; different endpoints and causes." + }, + { + "left": "issue:43618", + "right": "issue:44380", + "accept": false, + "reason": "CLIPOutput attention field regression vs GPT2 backend scaling 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": 13, + "review_unit_index": 38, + "review_unit_count": 42, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-367", + "nodes": [ + "issue:36010", + "issue:36683", + "issue:38617", + "issue:41669", + "issue:42757", + "issue:42994", + "issue:43010", + "issue:43012", + "issue:43064", + "issue:43066", + "issue:43299", + "issue:43352", + "issue:43475", + "issue:43502", + "issue:43525", + "issue:43582", + "issue:43618", + "issue:43756", + "issue:43761", + "issue:43824", + "issue:43867", + "issue:43901", + "issue:43931", + "issue:43937", + "issue:43976", + "issue:43992", + "issue:44038", + "issue:44077", + "issue:44188", + "issue:44230", + "issue:44242", + "issue:44246", + "issue:44263", + "issue:44380", + "issue:44483", + "issue:44485", + "issue:44492", + "issue:44655", + "issue:44683", + "issue:44938", + "issue:45216", + "issue:45278", + "issue:45290", + "issue:45310", + "issue:45375", + "issue:45381", + "issue:45440", + "issue:45478" + ], + "soft_pairs": [ + "issue:43618|issue:43761", + "issue:36683|issue:43582", + "issue:44483|issue:45290", + "issue:43012|issue:44683", + "issue:44038|issue:44230", + "issue:42757|issue:43824", + "issue:44483|issue:45381", + "issue:43299|issue:43867", + "issue:38617|issue:43502", + "issue:43756|issue:44485", + "issue:43901|issue:44655", + "issue:42994|issue:44655", + "issue:36683|issue:43475", + "issue:43867|issue:45216", + "issue:43352|issue:43937", + "issue:43066|issue:45381", + "issue:44655|issue:45478", + "issue:44077|issue:44380", + "issue:36683|issue:44077", + "issue:43064|issue:44492", + "issue:44263|issue:45440", + "issue:36683|issue:43525", + "issue:43976|issue:44938", + "issue:36010|issue:42757", + "issue:44188|issue:44380", + "issue:43618|issue:45216", + "issue:43992|issue:45310", + "issue:36683|issue:44242", + "issue:43010|issue:45278", + "issue:36683|issue:43352", + "issue:43352|issue:43931", + "issue:41669|issue:44246", + "issue:43867|issue:45375", + "issue:36683|issue:43064" + ], + "prepared_review_unit_hash": "df347db9ef5a6ccc91ac01a77eca053f9b3401fc984ca19156423449a139f8d0", + "component_budget": { + "node_count": 367, + "item_count": 367, + "soft_pair_count": 1370, + "serialized_chars": 706986, + "estimated_input_tokens": 176747, + "estimated_eval_tokens": 353750 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 34, + "serialized_chars": 37423, + "estimated_input_tokens": 9356, + "estimated_eval_tokens": 18968 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Mostly a mixed Transformers regression set rather than one tight duplicate cluster. issue:45278 is a reasonable umbrella anchor for the broad import-error reports, the CLIPVisionModel v5 output regressions (43618/43761) look like the clearest duplicate-like pair, and there is also a related Qwen3.5 MoE save/load mini-cluster (45216/45310/45478) that seems thematically close but not provably identical from the snippets alone.", + "confidence": 0.41, + "canonical_issue_reason": "issue:45278 is the broadest open umbrella in the packet and best fits the general post-upgrade import-regression theme without overcommitting to a narrower model-specific bug.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45278 is the most useful triage anchor because it is open, broad, and central to the import-error family represented in the packet.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43618", + "right": "issue:43761", + "accept": true, + "reason": "Both point to the same CLIPVisionModel v5 forward-path regression where optional outputs are dropped from the returned object." + }, + { + "left": "issue:36683", + "right": "issue:43582", + "accept": false, + "reason": "Different failures and subsystems: Gemma3Config vocab_size vs an Apple Silicon allocator TypeError." + }, + { + "left": "issue:44483", + "right": "issue:45290", + "accept": false, + "reason": "One is chat/completions request validation; the other is apply_chat_template handling of tool-call assistant messages." + }, + { + "left": "issue:43012", + "right": "issue:44683", + "accept": false, + "reason": "Both are PyTorch/attention-related, but the reported bugs are different: a bfloat16 warning versus a flex_attention failure on newer torch." + }, + { + "left": "issue:44038", + "right": "issue:44230", + "accept": false, + "reason": "Same model family, but one is a loading regression and the other is fp8 inference support; not the same concrete bug." + }, + { + "left": "issue:42757", + "right": "issue:43824", + "accept": false, + "reason": "Both are import errors, but for different symbols and likely different breakpoints." + }, + { + "left": "issue:44483", + "right": "issue:45381", + "accept": false, + "reason": "Different API surfaces and symptoms: request acceptance vs Qwen2.5-VL video position-id handling." + }, + { + "left": "issue:43299", + "right": "issue:43867", + "accept": false, + "reason": "Both are model-loading issues, but one is a Qwen3-VL-Moe regression and the other is a state_dict sorting/load-order problem." + }, + { + "left": "issue:38617", + "right": "issue:43502", + "accept": false, + "reason": "Unrelated problems: missing configuration_utils symbol versus local_files_only still triggering network access." + }, + { + "left": "issue:43756", + "right": "issue:44485", + "accept": false, + "reason": "Different models and different concerns: Smollm3 RoPE-layer dropping versus GLM-5 RoPE discussion/regression." + }, + { + "left": "issue:43901", + "right": "issue:44655", + "accept": false, + "reason": "Docs/behavior mismatch for return_all_scores is separate from Pipeline.save_pretrained serialization failure." + }, + { + "left": "issue:42994", + "right": "issue:44655", + "accept": false, + "reason": "Both involve saving, but quantized-model saving and pipeline serialization are different code paths." + }, + { + "left": "issue:36683", + "right": "issue:43475", + "accept": false, + "reason": "Distinct attribute errors on different model configs and different forward paths." + }, + { + "left": "issue:43867", + "right": "issue:45216", + "accept": false, + "reason": "A state_dict-order load error is not the same as a Qwen3.5 save_pretrained regression." + }, + { + "left": "issue:43352", + "right": "issue:43937", + "accept": false, + "reason": "Nemotron FlashAttention support gating and GLM-5 GenerationConfig validation are unrelated." + }, + { + "left": "issue:43066", + "right": "issue:45381", + "accept": false, + "reason": "Wrong tokenizer decoder type and wrong Qwen2.5-VL video position ids are separate bugs." + }, + { + "left": "issue:44655", + "right": "issue:45478", + "accept": false, + "reason": "Pipeline save_pretrained and Qwen3.5 MoE from_pretrained failures do not look like the same underlying issue." + }, + { + "left": "issue:44077", + "right": "issue:44380", + "accept": false, + "reason": "PatchTSMixer post_init policy and GPT2 attention scaling under SDPA/FlashAttention are different changes." + }, + { + "left": "issue:36683", + "right": "issue:44077", + "accept": false, + "reason": "Gemma3Config vocab_size and patchtsmixer post_init policy are unrelated." + }, + { + "left": "issue:43064", + "right": "issue:44492", + "accept": false, + "reason": "A distributed optimizer-state bug and a cache-strategy typo are not the same issue." + }, + { + "left": "issue:44263", + "right": "issue:45440", + "accept": false, + "reason": "Different model implementation problems: GlmMoeDsaIndexer split behavior versus DeepseekV3MoE divergence." + }, + { + "left": "issue:36683", + "right": "issue:43525", + "accept": false, + "reason": "Different config attribute regressions on different models." + }, + { + "left": "issue:43976", + "right": "issue:44938", + "accept": false, + "reason": "Both concern Python compatibility, but they describe different version constraints and failures." + }, + { + "left": "issue:36010", + "right": "issue:42757", + "accept": false, + "reason": "Both are import errors, but they involve different missing symbols and likely unrelated fixes." + }, + { + "left": "issue:44188", + "right": "issue:44380", + "accept": false, + "reason": "Attention/compile behavior is the common area, but the concrete bugs are different." + }, + { + "left": "issue:43618", + "right": "issue:45216", + "accept": false, + "reason": "CLIPVisionModel output regression and Qwen3.5 save_pretrained regression are unrelated." + }, + { + "left": "issue:43992", + "right": "issue:45310", + "accept": false, + "reason": "UMT5Encoder missing embed_tokens.weight is not the same bug as Qwen3.5 Moe from_pretrained failure." + }, + { + "left": "issue:36683", + "right": "issue:44242", + "accept": false, + "reason": "Gemma3Config vocab_size and missing load-balancing loss are separate model issues." + }, + { + "left": "issue:43010", + "right": "issue:45278", + "accept": false, + "reason": "Cache update no_grad decoration is a targeted API behavior change, not the broad import-regression cluster." + }, + { + "left": "issue:36683", + "right": "issue:43352", + "accept": false, + "reason": "Unrelated config attribute error versus FlashAttention support gating." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "The broad anchor choice (issue:45278) is grounded, and the CLIPVisionModel pair looks plausibly duplicate. However, the summary is a bit too strong: the packet also contains a closely related Qwen3.5 MoE save/load cluster (e.g. 45216, 45310, 45478) that makes \"the only plausible duplicate pair\" overconfident. Several soft-edge rejections also read as more definitive than the packet supports, given that we only have titles/body snippets and no diff evidence. Please keep the cluster characterization more conservative." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 13, + "review_unit_index": 39, + "review_unit_count": 42, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-367", + "nodes": [ + "issue:28282", + "issue:30990", + "issue:31795", + "issue:33453", + "issue:36296", + "issue:42548", + "issue:42890", + "issue:43010", + "issue:43065", + "issue:43232", + "issue:43299", + "issue:43502", + "issue:43575", + "issue:43618", + "issue:43723", + "issue:43867", + "issue:43976", + "issue:43992", + "issue:44062", + "issue:44075", + "issue:44077", + "issue:44112", + "issue:44222", + "issue:44315", + "issue:44351", + "issue:44355", + "issue:44360", + "issue:44368", + "issue:44373", + "issue:44387", + "issue:44484", + "issue:44485", + "issue:44556", + "issue:44749", + "issue:44857", + "issue:44871", + "issue:44938", + "issue:44964", + "issue:44977", + "issue:45071", + "issue:45125", + "issue:45341", + "issue:45357", + "issue:45375", + "issue:45381", + "issue:45440", + "issue:45478", + "issue:45561" + ], + "soft_pairs": [ + "issue:42890|issue:45561", + "issue:43992|issue:45478", + "issue:44938|issue:45071", + "issue:43867|issue:45357", + "issue:43010|issue:43976", + "issue:44938|issue:45381", + "issue:44360|issue:45440", + "issue:43618|issue:44077", + "issue:36296|issue:43575", + "issue:43867|issue:44484", + "issue:43618|issue:44857", + "issue:42890|issue:44964", + "issue:43618|issue:44977", + "issue:28282|issue:33453", + "issue:42890|issue:44112", + "issue:43010|issue:44075", + "issue:43723|issue:44556", + "issue:42890|issue:45341", + "issue:43502|issue:44062", + "issue:44222|issue:45125", + "issue:44387|issue:44556", + "issue:43010|issue:43065", + "issue:43010|issue:43232", + "issue:45125|issue:45478", + "issue:42548|issue:44351", + "issue:45125|issue:45375", + "issue:44355|issue:44749", + "issue:43502|issue:44485", + "issue:30990|issue:31795", + "issue:44075|issue:44871", + "issue:44315|issue:44360", + "issue:43010|issue:44368", + "issue:42890|issue:44373", + "issue:43502|issue:44222", + "issue:43299|issue:44556" + ], + "prepared_review_unit_hash": "d8e2723ad30220130109499c7e2ef6790a42457d573c6bcd29b0f0e53e069fb1", + "component_budget": { + "node_count": 367, + "item_count": 367, + "soft_pair_count": 1370, + "serialized_chars": 706986, + "estimated_input_tokens": 176747, + "estimated_eval_tokens": 353750 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 35, + "serialized_chars": 37644, + "estimated_input_tokens": 9411, + "estimated_eval_tokens": 19078 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is broad and mostly not duplicate-worthy: the soft pairs are only superficially similar (same subsystem or same release) but do not look like the same underlying bug or change. No PRs are present.", + "confidence": 0.46, + "canonical_issue_reason": "Issue 43723 is the clearest, most generic loading-regression report in the set, with an explicit API target (`AutoTokenizer.from_pretrained`) and a theme that matches several nearby loading-related reports.", + "canonical_pr_reason": null, + "best_issue_reason": "43723 is the best standalone representative of the cluster\u2019s loading/import regression subset: concrete repro, explicit failing path, and broader applicability than the model-specific reports.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:42890", + "right": "issue:45561", + "accept": false, + "reason": "Different test failures: missing `set_seed()` flakiness vs xdist race on `captured_info.txt`." + }, + { + "left": "issue:43992", + "right": "issue:45478", + "accept": false, + "reason": "Different model-loading bugs with different causes; shared `from_pretrained` wording is not enough." + }, + { + "left": "issue:44938", + "right": "issue:45071", + "accept": false, + "reason": "Both are v5 regressions, but one is a Python 3.14 import/load failure and the other is `PretrainedConfig` type checking." + }, + { + "left": "issue:43867", + "right": "issue:45357", + "accept": false, + "reason": "One is a state-dict load ordering issue; the other is incorrect keys saved for Qwen3.5 visual encoder." + }, + { + "left": "issue:43010", + "right": "issue:43976", + "accept": false, + "reason": "`@torch.no_grad` on cache updates is unrelated to Python-version compatibility." + }, + { + "left": "issue:44938", + "right": "issue:45381", + "accept": false, + "reason": "Python 3.14 loading failure vs Qwen2.5-VL video `vision_position_ids` bug; different code paths." + }, + { + "left": "issue:44360", + "right": "issue:45440", + "accept": false, + "reason": "DSA indexer activation bug and DeepSeekV3 implementation divergence are unrelated." + }, + { + "left": "issue:43618", + "right": "issue:44077", + "accept": false, + "reason": "Different issues: missing CLIP attentions assignment vs PatchTSMixer `post_init` API constraint." + }, + { + "left": "issue:36296", + "right": "issue:43575", + "accept": false, + "reason": "Both mention tensor parallelism, but one is a training bug and the other is an OOM while loading a specific model." + }, + { + "left": "issue:43867", + "right": "issue:44484", + "accept": false, + "reason": "State-dict loading error vs `save_pretrained` shard-size default; unrelated." + }, + { + "left": "issue:43618", + "right": "issue:44857", + "accept": false, + "reason": "CLIP output regression vs AMP float16 crash in a different loss module." + }, + { + "left": "issue:42890", + "right": "issue:44964", + "accept": false, + "reason": "Test-seed flakiness is unrelated to loading `Phi-4-multimodal-instruct`." + }, + { + "left": "issue:43618", + "right": "issue:44977", + "accept": false, + "reason": "Different bugs: CLIP attentions regression vs Qwen3.5 flash-attention generation problem." + }, + { + "left": "issue:28282", + "right": "issue:33453", + "accept": false, + "reason": "PyTorch import requirement error vs tokenizer-loading regression; no shared root cause." + }, + { + "left": "issue:42890", + "right": "issue:44112", + "accept": false, + "reason": "Unrelated CI/test flake and stale device-override test failure." + }, + { + "left": "issue:43010", + "right": "issue:44075", + "accept": false, + "reason": "Cache update decoration issue vs SGD optimizer args not being used." + }, + { + "left": "issue:43723", + "right": "issue:44556", + "accept": false, + "reason": "Both involve loading, but one is tokenizer loading in v5 and the other is checkpoint reload compatibility after upgrading versions." + }, + { + "left": "issue:42890", + "right": "issue:45341", + "accept": false, + "reason": "Different test utilities issues with no common underlying bug." + }, + { + "left": "issue:43502", + "right": "issue:44062", + "accept": false, + "reason": "Offline-loading network leak vs `AddedToken(..., special=...)` constructor error are unrelated." + }, + { + "left": "issue:44222", + "right": "issue:45125", + "accept": false, + "reason": "FP8 `save_pretrained` bug vs missing tensor-parallel plan on a Qwen MoE model." + }, + { + "left": "issue:44387", + "right": "issue:44556", + "accept": false, + "reason": "Int4 CUDA memory/OOM regression vs inability to reload checkpoints after version upgrade." + }, + { + "left": "issue:43010", + "right": "issue:43065", + "accept": false, + "reason": "`no_grad` on cache updates is unrelated to the dummy Conv2d in Sam3PixelDecoder." + }, + { + "left": "issue:43010", + "right": "issue:43232", + "accept": false, + "reason": "Different generation-path bugs: cache update decoration vs `_update_model_kwargs_for_generation` after `sync_gpus`." + }, + { + "left": "issue:45125", + "right": "issue:45478", + "accept": false, + "reason": "Related model family, but different failures: missing `_tp_plan` vs `from_pretrained` loading error in v5.4+." + }, + { + "left": "issue:42548", + "right": "issue:44351", + "accept": false, + "reason": "Missing `PreTrainedModel` import and missing `HybridCache` import are distinct import regressions." + }, + { + "left": "issue:45125", + "right": "issue:45375", + "accept": false, + "reason": "Tensor-parallel plan issue vs missing `deepstack_visual_indexes` config field; different root causes." + }, + { + "left": "issue:44355", + "right": "issue:44749", + "accept": false, + "reason": "Compiled Python file errors vs a performance regression after upgrading from 4.57.3 to 5.3.0." + }, + { + "left": "issue:43502", + "right": "issue:44485", + "accept": false, + "reason": "Offline-loading bug vs GLM-5 RoPE implementation discussion." + }, + { + "left": "issue:30990", + "right": "issue:31795", + "accept": false, + "reason": "Sentence Transformers loading hang and forward-doc confusion are unrelated." + }, + { + "left": "issue:44075", + "right": "issue:44871", + "accept": false, + "reason": "Unused SGD args vs Gemma-3 `eos_token_id` configuration mismatch." + }, + { + "left": "issue:44315", + "right": "issue:44360", + "accept": false, + "reason": "Model-init Liger-kernel application bug vs DSA indexer activation issue." + }, + { + "left": "issue:43010", + "right": "issue:44368", + "accept": false, + "reason": "`no_grad` cache-update bug vs warning about `tie_word_embeddings` in Qwen3.5 LoRA fine-tuning." + }, + { + "left": "issue:42890", + "right": "issue:44373", + "accept": false, + "reason": "Test flake vs wrong docstring for `position_ids`; different severity and code path." + }, + { + "left": "issue:43502", + "right": "issue:44222", + "accept": false, + "reason": "Local-files-only network bug vs FP8 MoE save/load issue; not the same change." + }, + { + "left": "issue:43299", + "right": "issue:44556", + "accept": false, + "reason": "Qwen3VL MoE loading breakage and post-upgrade checkpoint reload failure are different 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": 13, + "review_unit_index": 40, + "review_unit_count": 42, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-367", + "nodes": [ + "issue:31795", + "issue:33357", + "issue:33453", + "issue:34689", + "issue:36296", + "issue:38617", + "issue:41628", + "issue:41669", + "issue:42371", + "issue:42548", + "issue:42757", + "issue:42831", + "issue:42890", + "issue:43012", + "issue:43502", + "issue:43526", + "issue:43576", + "issue:43673", + "issue:43824", + "issue:43827", + "issue:43874", + "issue:43901", + "issue:44038", + "issue:44222", + "issue:44230", + "issue:44246", + "issue:44261", + "issue:44273", + "issue:44351", + "issue:44355", + "issue:44360", + "issue:44393", + "issue:44485", + "issue:44492", + "issue:44556", + "issue:44623", + "issue:44655", + "issue:44704", + "issue:44908", + "issue:44961", + "issue:45003", + "issue:45071", + "issue:45092", + "issue:45125", + "issue:45230", + "issue:45468", + "issue:45542" + ], + "soft_pairs": [ + "issue:43502|issue:44961", + "issue:44222|issue:44230", + "issue:44355|issue:44655", + "issue:44355|issue:45071", + "issue:42831|issue:43874", + "issue:33453|issue:34689", + "issue:42757|issue:43576", + "issue:43827|issue:44961", + "issue:33453|issue:36296", + "issue:43576|issue:45071", + "issue:43576|issue:43824", + "issue:43673|issue:44492", + "issue:38617|issue:43576", + "issue:33357|issue:33453", + "issue:44961|issue:45468", + "issue:31795|issue:38617", + "issue:43576|issue:44351", + "issue:44908|issue:44961", + "issue:41628|issue:42548", + "issue:44485|issue:44961", + "issue:36296|issue:43901", + "issue:36296|issue:43526", + "issue:41628|issue:44355", + "issue:36296|issue:43012", + "issue:44273|issue:45003", + "issue:44261|issue:44360", + "issue:44273|issue:45542", + "issue:44623|issue:45230", + "issue:43576|issue:44556", + "issue:41669|issue:43673", + "issue:44393|issue:45071", + "issue:44038|issue:45125", + "issue:44556|issue:45092", + "issue:42371|issue:44492", + "issue:42890|issue:44908", + "issue:44704|issue:44908", + "issue:44246|issue:44273" + ], + "prepared_review_unit_hash": "3a47f5ff87530614cf0d6f95de2ca3042027c662bc4b1e16b1fd8cf5adc3497c", + "component_budget": { + "node_count": 367, + "item_count": 367, + "soft_pair_count": 1370, + "serialized_chars": 706986, + "estimated_input_tokens": 176747, + "estimated_eval_tokens": 353750 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 37, + "serialized_chars": 37411, + "estimated_input_tokens": 9353, + "estimated_eval_tokens": 18962 + }, + "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-era regressions and import/loading complaints. Most pairs are only loosely related by subsystem or release, not by the same concrete bug. The one clear duplicate-like edge is the lazy-loading failure versus unsafe `sys.modules` access in `modeling_utils`.", + "confidence": 0.78, + "canonical_issue_reason": "issue:45003 is the most code-path-specific report in the set: it points to an unsafe `sys.modules[]` access in `modeling_utils`, which plausibly explains the broader lazy-loading/import failures around it.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45003 is the best cluster anchor because it names a concrete mechanism rather than a symptom-only report.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43502", + "right": "issue:44961", + "accept": false, + "reason": "No shared bug; local_files_only networking behavior is unrelated to the placeholder/nonsense racoon issue." + }, + { + "left": "issue:44222", + "right": "issue:44230", + "accept": false, + "reason": "Both mention FP8/MoE, but one is save_pretrained and the other is inference support; different code paths." + }, + { + "left": "issue:44355", + "right": "issue:44655", + "accept": false, + "reason": "Compiled Python file errors are unrelated to Pipeline save_pretrained missing files." + }, + { + "left": "issue:44355", + "right": "issue:45071", + "accept": false, + "reason": "No overlap beyond general v5 breakage; compiled-file errors and PretrainedConfig type-checking are different bugs." + }, + { + "left": "issue:42831", + "right": "issue:43874", + "accept": false, + "reason": "Different models and failure modes: FP8 accuracy vs a missing image-patch helper AttributeError." + }, + { + "left": "issue:33453", + "right": "issue:34689", + "accept": false, + "reason": "Both are loading regressions, but they affect different loaders/models and do not look like the same bug." + }, + { + "left": "issue:42757", + "right": "issue:43576", + "accept": false, + "reason": "Missing `is_offline_mode` import from huggingface_hub is separate from the v5 env command breakage." + }, + { + "left": "issue:43827", + "right": "issue:44961", + "accept": false, + "reason": "Docs referencing removed pipeline APIs is unrelated to the racoon issue." + }, + { + "left": "issue:33453", + "right": "issue:36296", + "accept": false, + "reason": "Tokenizer loading regression and tensor-parallel training bug are unrelated." + }, + { + "left": "issue:43576", + "right": "issue:45071", + "accept": false, + "reason": "Environment command breakage and `PretrainedConfig` type checking are different regressions." + }, + { + "left": "issue:43576", + "right": "issue:43824", + "accept": false, + "reason": "v5 env command issue is unrelated to missing Qwen2.5-VL import/export." + }, + { + "left": "issue:43673", + "right": "issue:44492", + "accept": false, + "reason": "Cache-missing during chunked prefill is not the same as a typo in cache strategies." + }, + { + "left": "issue:38617", + "right": "issue:43576", + "accept": false, + "reason": "ImportError for `layer_type_validation` is a specific missing symbol issue, not the env command regression." + }, + { + "left": "issue:33357", + "right": "issue:33453", + "accept": false, + "reason": "MacOS CLIP bus error and tokenizer loading regression are not the same underlying bug." + }, + { + "left": "issue:44961", + "right": "issue:45468", + "accept": false, + "reason": "Racoon is not meaningfully connected to Gemma-4 audio positional encoding." + }, + { + "left": "issue:31795", + "right": "issue:38617", + "accept": false, + "reason": "Documentation confusion about forward args is unrelated to the missing import error." + }, + { + "left": "issue:43576", + "right": "issue:44351", + "accept": false, + "reason": "v5 env command breakage and missing `HybridCache` import are distinct issues." + }, + { + "left": "issue:44908", + "right": "issue:44961", + "accept": false, + "reason": "Scheduler kwargs bug is unrelated to the racoon issue." + }, + { + "left": "issue:41628", + "right": "issue:42548", + "accept": false, + "reason": "Both are import errors, but they concern different top-level symbols and no concrete shared code path is evident." + }, + { + "left": "issue:44485", + "right": "issue:44961", + "accept": false, + "reason": "GLM-5 RoPE implementation discussion is unrelated to the racoon issue." + }, + { + "left": "issue:36296", + "right": "issue:43901", + "accept": false, + "reason": "Tensor-parallel training bug and TextClassificationPipeline docs mismatch are unrelated." + }, + { + "left": "issue:36296", + "right": "issue:43526", + "accept": false, + "reason": "Tensor-parallel training bug and BeitImageProcessorFast label reduction are unrelated." + }, + { + "left": "issue:41628", + "right": "issue:44355", + "accept": false, + "reason": "AutoImageProcessor import failure and compiled Python file errors do not share the same bug." + }, + { + "left": "issue:36296", + "right": "issue:43012", + "accept": false, + "reason": "Tensor parallel training and bf16 compilation warnings are different problems." + }, + { + "left": "issue:44273", + "right": "issue:45003", + "accept": true, + "reason": "Both point at lazy-loading/import machinery; `sys.modules[]` access in modeling utils is a plausible root cause for lazy loading not working properly." + }, + { + "left": "issue:44261", + "right": "issue:44360", + "accept": false, + "reason": "Missing `rms_norm_eps` precision drift and a DSA indexer ReLU comment are unrelated." + }, + { + "left": "issue:44273", + "right": "issue:45542", + "accept": false, + "reason": "Lazy loading problems and the TensorFlow/tensorboard backend error are different failures." + }, + { + "left": "issue:44623", + "right": "issue:45230", + "accept": false, + "reason": "Missing processor save files is not the same as a generic bug report." + }, + { + "left": "issue:43576", + "right": "issue:44556", + "accept": false, + "reason": "Broken env command and checkpoint reload incompatibility after upgrade are different migration issues." + }, + { + "left": "issue:41669", + "right": "issue:43673", + "accept": false, + "reason": "Import-star slow import issue and missing cache in chunked prefill are unrelated." + }, + { + "left": "issue:44393", + "right": "issue:45071", + "accept": false, + "reason": "Qwen3-VL bbox output error and PretrainedConfig type checking are unrelated." + }, + { + "left": "issue:44038", + "right": "issue:45125", + "accept": false, + "reason": "Both involve Qwen3/VL/MoE, but one is broad v5 support and the other is missing `_tp_plan` for tensor parallelism." + }, + { + "left": "issue:44556", + "right": "issue:45092", + "accept": false, + "reason": "Both are upgrade/loading compatibility reports, but they affect different checkpoint/model scenarios and do not look like one bug." + }, + { + "left": "issue:42371", + "right": "issue:44492", + "accept": false, + "reason": "TF32 settings guidance and a cache-strategy typo are unrelated." + }, + { + "left": "issue:42890", + "right": "issue:44908", + "accept": false, + "reason": "Test flakiness from missing `set_seed()` is unrelated to the inverse_sqrt scheduler kwargs bug." + }, + { + "left": "issue:44704", + "right": "issue:44908", + "accept": false, + "reason": "AutoProcessor kwargs forwarding and scheduler kwargs handling are different code paths." + }, + { + "left": "issue:44246", + "right": "issue:44273", + "accept": false, + "reason": "Slow import and lazy loading malfunction are related at a high level, but not the same concrete bug." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "Grounded overall: the summary stays close to the packet, and the only accepted soft edge (44273 \u2194 45003) is a conservative, plausible lazy-loading/sys.modules match. The other pair judgments are generally cautious and do not overclaim shared bugs." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 2, + "component_count": 13, + "review_unit_index": 41, + "review_unit_count": 42, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-367", + "nodes": [ + "issue:31795", + "issue:33453", + "issue:34689", + "issue:37428", + "issue:41669", + "issue:42371", + "issue:42890", + "issue:42913", + "issue:43066", + "issue:43352", + "issue:43408", + "issue:43502", + "issue:43519", + "issue:43576", + "issue:43606", + "issue:43618", + "issue:43704", + "issue:43723", + "issue:43827", + "issue:43976", + "issue:44077", + "issue:44261", + "issue:44263", + "issue:44273", + "issue:44295", + "issue:44297", + "issue:44315", + "issue:44351", + "issue:44360", + "issue:44393", + "issue:44485", + "issue:44492", + "issue:44556", + "issue:44623", + "issue:44655", + "issue:44829", + "issue:44861", + "issue:44908", + "issue:44945", + "issue:44998", + "issue:45071", + "issue:45092", + "issue:45230", + "issue:45290", + "issue:45310", + "issue:45341", + "issue:45468", + "issue:45478" + ], + "soft_pairs": [ + "issue:43408|issue:44315", + "issue:44297|issue:45230", + "issue:43576|issue:44908", + "issue:43352|issue:44393", + "issue:43976|issue:44393", + "issue:44908|issue:45071", + "issue:44295|issue:45230", + "issue:41669|issue:44351", + "issue:41669|issue:45341", + "issue:44492|issue:44945", + "issue:41669|issue:44492", + "issue:44315|issue:44829", + "issue:44393|issue:45478", + "issue:44077|issue:44861", + "issue:43827|issue:44908", + "issue:31795|issue:33453", + "issue:41669|issue:44908", + "issue:43723|issue:44393", + "issue:44655|issue:44861", + "issue:43576|issue:44273", + "issue:44360|issue:45468", + "issue:45230|issue:45310", + "issue:44273|issue:44908", + "issue:41669|issue:43704", + "issue:43519|issue:44485", + "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:44261|issue:45468", + "issue:43606|issue:45468", + "issue:42890|issue:43704", + "issue:42913|issue:43066", + "issue:44908|issue:44998", + "issue:44861|issue:45290", + "issue:44492|issue:44998", + "issue:44998|issue:45341", + "issue:43618|issue:44998", + "issue:44556|issue:44998" + ], + "prepared_review_unit_hash": "5f102cb6cb356432e6903e2d9e4fbfdc7872607a41d0045ea4f6baaca48c7ffe", + "component_budget": { + "node_count": 367, + "item_count": 367, + "soft_pair_count": 1370, + "serialized_chars": 706986, + "estimated_input_tokens": 176747, + "estimated_eval_tokens": 353750 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 46, + "serialized_chars": 40859, + "estimated_input_tokens": 10215, + "estimated_eval_tokens": 20686 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is a noisy mix of mostly unrelated Transformers v5 regressions. The only clearly recurring subtheme is tokenizer loading/behavior, so I\u2019d use that as the representative issue, but I would reject all soft edges as true duplicates.", + "confidence": 0.36, + "canonical_issue_reason": "Issue 43723 is the clearest and most central representative of the tokenizer-loading regression subcluster, which appears to be the strongest coherent theme in an otherwise noisy cluster.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 43723 is the best standalone representative because it states the bug concretely, maps to the main recurring theme, and is more actionable than the more generic adjacent tokenizer reports.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43408", + "right": "issue:44315", + "accept": false, + "reason": "Different problems: model-type mismatch warning vs. Liger Kernel not being applied in model_init." + }, + { + "left": "issue:44297", + "right": "issue:45230", + "accept": false, + "reason": "One is a specific tokenizer.save_pretrained mismatch; the other is a generic bug report with no concrete shared failure mode." + }, + { + "left": "issue:43576", + "right": "issue:44908", + "accept": false, + "reason": "Unrelated subsystems: CLI env command vs. inverse_sqrt scheduler kwargs handling." + }, + { + "left": "issue:43352", + "right": "issue:44393", + "accept": false, + "reason": "Different model families and different failures: Flash Attention 2 support vs. Qwen3-VL bounding-box output issues." + }, + { + "left": "issue:43976", + "right": "issue:44393", + "accept": false, + "reason": "Python version compatibility regression is unrelated to the Qwen3-VL output bug." + }, + { + "left": "issue:44908", + "right": "issue:45071", + "accept": false, + "reason": "Scheduler kwarg propagation and PretrainedConfig type checking are distinct bugs." + }, + { + "left": "issue:44295", + "right": "issue:45230", + "accept": false, + "reason": "Specific position_ids buffer access issue does not match a generic bug report." + }, + { + "left": "issue:41669", + "right": "issue:44351", + "accept": false, + "reason": "Import-star cleanup/performance issue is not the same as the HybridCache import error." + }, + { + "left": "issue:41669", + "right": "issue:45341", + "accept": false, + "reason": "Broad import cleanup is unrelated to a testing_utils bug." + }, + { + "left": "issue:44492", + "right": "issue:44945", + "accept": false, + "reason": "A cache-strategy typo is unrelated to incorrect pipeline-parallel LLM output." + }, + { + "left": "issue:41669", + "right": "issue:44492", + "accept": false, + "reason": "These are different code changes and different symptoms: import cleanup vs. cache-strategy typo." + }, + { + "left": "issue:44315", + "right": "issue:44829", + "accept": false, + "reason": "Model_init/Liger Kernel application is not the same underlying bug as flash_attention_3 training degeneration." + }, + { + "left": "issue:44393", + "right": "issue:45478", + "accept": false, + "reason": "Same broad area only by model family name; the concrete failure modes are different Qwen bugs." + }, + { + "left": "issue:44077", + "right": "issue:44861", + "accept": false, + "reason": "Optional post_init validation and _tied_weights_keys list handling are separate model-internals issues." + }, + { + "left": "issue:43827", + "right": "issue:44908", + "accept": false, + "reason": "Docs still referencing pipeline() is unrelated to scheduler kwargs being ignored." + }, + { + "left": "issue:31795", + "right": "issue:33453", + "accept": false, + "reason": "Documentation confusion around forward args is not the same as tokenizer loading regression." + }, + { + "left": "issue:41669", + "right": "issue:44908", + "accept": false, + "reason": "Import cleanup is unrelated to scheduler behavior." + }, + { + "left": "issue:43723", + "right": "issue:44393", + "accept": false, + "reason": "Tokenizer loading in v5 is unrelated to Qwen3-VL bounding-box output errors." + }, + { + "left": "issue:44655", + "right": "issue:44861", + "accept": false, + "reason": "Pipeline save_pretrained problems do not match _get_tied_weight_keys crashing on a list." + }, + { + "left": "issue:43576", + "right": "issue:44273", + "accept": false, + "reason": "transformers env command breakage is not the same as lazy loading failures." + }, + { + "left": "issue:44360", + "right": "issue:45468", + "accept": false, + "reason": "A DSA indexer ReLU discussion and Gemma4AudioRelPositionalEncoding are unrelated." + }, + { + "left": "issue:45230", + "right": "issue:45310", + "accept": false, + "reason": "A generic bug report cannot be treated as the same concrete from_pretrained error." + }, + { + "left": "issue:44273", + "right": "issue:44908", + "accept": false, + "reason": "Lazy loading and scheduler kwargs are different areas with different failure modes." + }, + { + "left": "issue:41669", + "right": "issue:43704", + "accept": false, + "reason": "Import-star removal performance issue is unrelated to VRAM leakage across dataloader threads." + }, + { + "left": "issue:43519", + "right": "issue:44485", + "accept": false, + "reason": "Timestamp calculation in Qwen3VL does not match a GLM-5 RoPE implementation bug." + }, + { + "left": "issue:43519", + "right": "issue:44908", + "accept": false, + "reason": "Processor timestamp math and scheduler kwargs are unrelated." + }, + { + "left": "issue:31795", + "right": "issue:34689", + "accept": false, + "reason": "Documentation ambiguity around forward args is not the same as model loading regression for Llama 3.2 Vision." + }, + { + "left": "issue:31795", + "right": "issue:37428", + "accept": false, + "reason": "Forward-doc confusion is unrelated to the flash_attention_utils import failure." + }, + { + "left": "issue:44315", + "right": "issue:45092", + "accept": false, + "reason": "Liger Kernel application in model_init and remote-code meta-init incompatibility are different loading paths." + }, + { + "left": "issue:43704", + "right": "issue:44485", + "accept": false, + "reason": "VRAM leakage in threaded inference is unrelated to RoPE implementation concerns." + }, + { + "left": "issue:44485", + "right": "issue:45468", + "accept": false, + "reason": "Different model families and different bugs: GLM-5 RoPE vs. Gemma4Audio positional encoding." + }, + { + "left": "issue:42371", + "right": "issue:43704", + "accept": false, + "reason": "TF32 API settings and VRAM leaks are unrelated." + }, + { + "left": "issue:43502", + "right": "issue:43519", + "accept": false, + "reason": "local_files_only network leakage and Qwen3VL timestamp calculation are unrelated." + }, + { + "left": "issue:43519", + "right": "issue:44623", + "accept": false, + "reason": "Processor timestamp calculation is not the same as processor.save_pretrained missing files." + }, + { + "left": "issue:43519", + "right": "issue:43704", + "accept": false, + "reason": "Timestamp calculation bug and VRAM leak are different failure modes." + }, + { + "left": "issue:44263", + "right": "issue:45468", + "accept": false, + "reason": "torch.split return-value issue is unrelated to Gemma4AudioRelPositionalEncoding." + }, + { + "left": "issue:44261", + "right": "issue:45468", + "accept": false, + "reason": "MLA q_a_layernorm precision/config issue is unrelated to Gemma4Audio positional encoding." + }, + { + "left": "issue:43606", + "right": "issue:45468", + "accept": false, + "reason": "CPU offload device mismatch is not the same as Gemma4AudioRelPositionalEncoding." + }, + { + "left": "issue:42890", + "right": "issue:43704", + "accept": false, + "reason": "Missing set_seed in tests is unrelated to VRAM leakage in dataloader threads." + }, + { + "left": "issue:42913", + "right": "issue:43066", + "accept": false, + "reason": "Tokenizer behavior differences and wrong decoder type are related only loosely, not the same underlying bug." + }, + { + "left": "issue:44908", + "right": "issue:44998", + "accept": false, + "reason": "Scheduler kwarg handling and a nonsense issue title are clearly unrelated." + }, + { + "left": "issue:44861", + "right": "issue:45290", + "accept": false, + "reason": "Tied weight key handling is unrelated to apply_chat_template crashing on tool-call assistant messages." + }, + { + "left": "issue:44492", + "right": "issue:44998", + "accept": false, + "reason": "Cache-strategy typo and a nonsense issue title are unrelated." + }, + { + "left": "issue:44998", + "right": "issue:45341", + "accept": false, + "reason": "Nonsense title and a testing_utils bug are not the same issue." + }, + { + "left": "issue:43618", + "right": "issue:44998", + "accept": false, + "reason": "CLIPOutput attentions regression is unrelated to the nonsense issue title." + }, + { + "left": "issue:44556", + "right": "issue:44998", + "accept": false, + "reason": "Checkpoint reload regression across versions is unrelated to the nonsense issue title." + } + ] + }, + "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": 13, + "review_unit_index": 42, + "review_unit_count": 42, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-15354-367", + "nodes": [ + "issue:31515", + "issue:31795", + "issue:33453", + "issue:36296", + "issue:36683", + "issue:38617", + "issue:43298", + "issue:43525", + "issue:43572", + "issue:44230", + "issue:44623", + "issue:44861", + "issue:45310" + ], + "soft_pairs": [ + "issue:43525|issue:44861", + "issue:43572|issue:44861", + "issue:31515|issue:38617", + "issue:44230|issue:44623", + "issue:31515|issue:36296", + "issue:31515|issue:31795", + "issue:44230|issue:45310", + "issue:31515|issue:33453", + "issue:31515|issue:36683", + "issue:43298|issue:44230" + ], + "prepared_review_unit_hash": "eea735345ed49354c57614eab97c57d5a078991791107e80839aae5cccf288ba", + "component_budget": { + "node_count": 367, + "item_count": 367, + "soft_pair_count": 1370, + "serialized_chars": 706986, + "estimated_input_tokens": 176747, + "estimated_eval_tokens": 353750 + }, + "budget": { + "node_count": 13, + "item_count": 13, + "soft_pair_count": 10, + "serialized_chars": 10321, + "estimated_input_tokens": 2581, + "estimated_eval_tokens": 5418 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This set is not a true duplicate cluster: it mixes unrelated loading, config, docs, serialization, and feature-request issues across different code paths. No single issue cleanly represents the whole group.", + "confidence": 0.95, + "canonical_issue_reason": "No canonical issue fits the cluster: the reports describe different underlying bugs and requests rather than one shared regression.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 45310 is the most concrete recent from_pretrained regression, but it is still only one narrow problem and does not represent the rest of the mixed cluster.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43525", + "right": "issue:44861", + "accept": false, + "reason": "Different bugs: Llama4Config missing pad_token_id vs _get_tied_weight_keys crashing on list-valued _tied_weights_keys." + }, + { + "left": "issue:43572", + "right": "issue:44861", + "accept": false, + "reason": "StableLmConfig pad_token_idx regression is unrelated to tied-weight key handling." + }, + { + "left": "issue:31515", + "right": "issue:38617", + "accept": false, + "reason": "One is slow checkpoint loading; the other is an ImportError for layer_type_validation. Different failure modes and code paths." + }, + { + "left": "issue:44230", + "right": "issue:44623", + "accept": false, + "reason": "FP8 Qwen3-VL/MoE inference support is a feature request; processor.save_pretrained missing files is a serialization bug." + }, + { + "left": "issue:31515", + "right": "issue:36296", + "accept": false, + "reason": "from_pretrained checkpoint slowness is unrelated to tensor-parallel training behavior." + }, + { + "left": "issue:31515", + "right": "issue:31795", + "accept": false, + "reason": "Checkpoint loading performance and forward API documentation confusion are distinct issues." + }, + { + "left": "issue:44230", + "right": "issue:45310", + "accept": false, + "reason": "FP8 support request for Qwen3-VL/MoE is not the same as a Qwen3.5 MoE from_pretrained error." + }, + { + "left": "issue:31515", + "right": "issue:33453", + "accept": false, + "reason": "Slow from_pretrained checkpoint loading and tokenizer loading regression are different loader problems." + }, + { + "left": "issue:31515", + "right": "issue:36683", + "accept": false, + "reason": "Checkpoint load slowness does not match the Gemma3Config missing vocab_size attribute bug." + }, + { + "left": "issue:43298", + "right": "issue:44230", + "accept": false, + "reason": "Continuous batching audio-model support and FP8 inference support are separate feature gaps, not one defect." + } + ] + }, + "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": 13, + "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": "ff47fbaf803419612d8855ebedada103273e93ac4d94c9442c704ff31af5379c", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1590, + "estimated_input_tokens": 398, + "estimated_eval_tokens": 1052 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1590, + "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 separate feature-request issues about adding support for more architectures, but they target different technologies (Flash Attention 2 vs GGUF) and do not describe the same underlying change.", + "confidence": 0.96, + "canonical_issue_reason": "No canonical duplicate issue: the requests are about different capabilities and code paths, not the same bug or feature.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 26350 is the better representative issue only by discussion activity and age, but it is not a duplicate of 33260.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:26350", + "right": "issue:33260", + "accept": false, + "reason": "Different feature requests: one is for Flash Attention 2 support, the other for GGUF support. They are related only at a broad 'more architectures' theme, 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": 4, + "component_count": 13, + "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": "adc762a3c4bfe0ed98700f7c411845279288a0eab9a9651160efbb067a233c1e", + "component_budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 6, + "serialized_chars": 4455, + "estimated_input_tokens": 1114, + "estimated_eval_tokens": 2484 + }, + "budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 6, + "serialized_chars": 4455, + "estimated_input_tokens": 1114, + "estimated_eval_tokens": 2484 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are templated translation-request issues, not a clear duplicate set. The Turkish-specific issue is concrete, while the others are generic placeholders or separate language targets, so they should not be merged as the same underlying request.", + "confidence": 0.91, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": "issue:27088 is the most concrete and specific item in the cluster, with an explicit Turkish translation request; the others are generic template variants with no distinct target language shown.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:27088", + "right": "issue:44370", + "accept": false, + "reason": "Different or unspecified language targets; similar template wording is not enough to treat them as the same issue." + }, + { + "left": "issue:27088", + "right": "issue:44365", + "accept": false, + "reason": "This is a generic placeholder-style translation request, not the same concrete Turkish request." + }, + { + "left": "issue:27088", + "right": "issue:43834", + "accept": false, + "reason": "Template similarity only; no evidence these are the same underlying translation task." + }, + { + "left": "issue:43834", + "right": "issue:44370", + "accept": false, + "reason": "Both look like generic template instances, but they do not identify the same concrete language target or work item." + }, + { + "left": "issue:43834", + "right": "issue:44365", + "accept": false, + "reason": "High textual similarity alone is insufficient; no specific shared translation target is established." + }, + { + "left": "issue:44365", + "right": "issue:44370", + "accept": false, + "reason": "These appear to be separate template-style requests rather than one identical issue." + } + ] + }, + "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": 13, + "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-36817-13", + "nodes": [ + "issue:36817", + "issue:37015", + "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:41211|issue:43671", + "issue:42503|issue:43837", + "issue:42503|issue:43671", + "issue:42503|issue:42738", + "issue:42503|issue:45027", + "issue:42738|issue:45027", + "issue:43671|issue:43837", + "issue:43671|issue:45027", + "issue:43671|issue:45177", + "issue:43671|issue:44121", + "issue:43671|issue:43909", + "issue:43837|issue:43909", + "issue:43837|issue:45322", + "issue:43837|issue:44121", + "issue:43909|issue:44205", + "issue:44121|issue:45322", + "issue:43837|issue:45027", + "issue:43837|issue:45177" + ], + "prepared_review_unit_hash": "eb25d1614f619d939c8ee4daa842bc13ddbb473e669bdab817440b9c484fc3bd", + "component_budget": { + "node_count": 13, + "item_count": 13, + "soft_pair_count": 25, + "serialized_chars": 16008, + "estimated_input_tokens": 4002, + "estimated_eval_tokens": 8260 + }, + "budget": { + "node_count": 13, + "item_count": 13, + "soft_pair_count": 25, + "serialized_chars": 16008, + "estimated_input_tokens": 4002, + "estimated_eval_tokens": 8260 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "This is a loose cluster of model-addition requests, not a true duplicate set. The items cover distinct models and often different modalities/tasks (BERT variants, ASR/TTS, OCR, sparse transformers, vision-language encoders), so the soft similarities are mostly generic wording rather than the same underlying change.", + "confidence": 0.95, + "canonical_issue_reason": "If forced to pick an anchor, issue:37015 is the broadest and most active generic model-add request, but it is not a true duplicate of the others.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:37015 has the strongest discussion activity and is the most representative generic request in this otherwise non-duplicate set.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:36817", + "right": "issue:41211", + "accept": false, + "reason": "Different model targets (EuroBert config vs DEIMv2); shared theme is only generic model addition." + }, + { + "left": "issue:36817", + "right": "issue:43671", + "accept": false, + "reason": "EuroBert config request and Qwen3-TTS support are different model families and tasks." + }, + { + "left": "issue:37015", + "right": "issue:43837", + "accept": false, + "reason": "NeoBERT and Qwen3-ASR are unrelated model additions; not the same underlying change." + }, + { + "left": "issue:37015", + "right": "issue:44121", + "accept": false, + "reason": "NeoBERT vs circuit-sparsity model request; different architecture and code path." + }, + { + "left": "issue:37015", + "right": "issue:43909", + "accept": false, + "reason": "NeoBERT and LFM2.5 Audio are different model families and modalities." + }, + { + "left": "issue:37015", + "right": "issue:42503", + "accept": false, + "reason": "NeoBERT and ModernVBERT are distinct model requests, not one duplicate." + }, + { + "left": "issue:37015", + "right": "issue:45027", + "accept": false, + "reason": "NeoBERT and Voxtral TTS are separate model support requests with different implementation needs." + }, + { + "left": "issue:41211", + "right": "issue:43671", + "accept": false, + "reason": "DEIMv2 and Qwen3-TTS are unrelated requests; only broad model-add similarity." + }, + { + "left": "issue:42503", + "right": "issue:43837", + "accept": false, + "reason": "ModernVBERT models and Qwen3-ASR are different tasks and code paths." + }, + { + "left": "issue:42503", + "right": "issue:43671", + "accept": false, + "reason": "ModernVBERT and Qwen3-TTS are not the same model or feature." + }, + { + "left": "issue:42503", + "right": "issue:42738", + "accept": false, + "reason": "ModernVBERT models vs generic BERT-like RoPE support; related theme, but not the same change." + }, + { + "left": "issue:42503", + "right": "issue:45027", + "accept": false, + "reason": "ModernVBERT and Voxtral-TTS are different model support requests." + }, + { + "left": "issue:42738", + "right": "issue:45027", + "accept": false, + "reason": "BERT-like RoPE support and Voxtral-TTS are not duplicate changes." + }, + { + "left": "issue:43671", + "right": "issue:43837", + "accept": false, + "reason": "Qwen3-TTS and Qwen3-ASR are different modalities and implementation paths; same family name is insufficient." + }, + { + "left": "issue:43671", + "right": "issue:45027", + "accept": false, + "reason": "Qwen3-TTS vs Voxtral-TTS are separate TTS model support requests." + }, + { + "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": "TTS support and weight-sparse transformer support are different feature/code changes." + }, + { + "left": "issue:43671", + "right": "issue:43909", + "accept": false, + "reason": "Qwen3-TTS and LFM2.5 Audio are both audio-related but clearly different models and integrations." + }, + { + "left": "issue:43837", + "right": "issue:43909", + "accept": false, + "reason": "Qwen3-ASR and LFM2.5 Audio are distinct audio model requests, not duplicates." + }, + { + "left": "issue:43837", + "right": "issue:45322", + "accept": false, + "reason": "ASR support and EUPE encoder support are different model families and tasks." + }, + { + "left": "issue:43837", + "right": "issue:44121", + "accept": false, + "reason": "Qwen3-ASR and circuit-sparsity transformer support are unrelated changes." + }, + { + "left": "issue:43909", + "right": "issue:44205", + "accept": false, + "reason": "Audio model request vs SAM3-LiteText; different model types and implementation work." + }, + { + "left": "issue:44121", + "right": "issue:45322", + "accept": false, + "reason": "Weight-sparse transformer request and EUPE encoder request are different underlying changes." + }, + { + "left": "issue:43837", + "right": "issue:45027", + "accept": false, + "reason": "Qwen3-ASR and Voxtral-TTS are different speech models and code paths." + }, + { + "left": "issue:43837", + "right": "issue:45177", + "accept": false, + "reason": "Qwen3-ASR and DeepSeek-OCR2 are unrelated model additions." + } + ] + }, + "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": 13, + "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": "7cce7d1db6d3ce2f9ee54f6d2ba6dfb053d8aa4790ff7aea72db216b46d22c12", + "component_budget": { + "node_count": 6, + "item_count": 6, + "soft_pair_count": 8, + "serialized_chars": 6293, + "estimated_input_tokens": 1574, + "estimated_eval_tokens": 3404 + }, + "budget": { + "node_count": 6, + "item_count": 6, + "soft_pair_count": 8, + "serialized_chars": 6293, + "estimated_input_tokens": 1574, + "estimated_eval_tokens": 3404 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "False-positive cluster: the issues span unrelated bugs across different models and subsystems, with only superficial keyword overlap.", + "confidence": 0.98, + "canonical_issue_reason": "No canonical issue stands out; the reports cover distinct problems (normalization behavior, Mask2Former loss handling, missing kwargs propagation, a type-hint typo, a quantization attribute typo, and a tokenizer regex warning).", + "canonical_pr_reason": null, + "best_issue_reason": "No issue is a strong representative of the cluster. If forced, 44031 is the broadest user-facing regression report, but it still does not align with the others.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:37934", + "right": "issue:43668", + "accept": false, + "reason": "Different problem domains: Llama4 normalization semantics vs a ModernBERT config type-hint typo." + }, + { + "left": "issue:42936", + "right": "issue:43240", + "accept": false, + "reason": "Mask2Former ignore_value handling is separate from kwargs not being forwarded into the loss function." + }, + { + "left": "issue:42936", + "right": "issue:44031", + "accept": false, + "reason": "Unrelated subsystems: Mask2Former loss/ignore_value vs tokenizer regex warnings after a version bump." + }, + { + "left": "issue:42936", + "right": "issue:43813", + "accept": false, + "reason": "No shared code path; one is a segmentation loss bug, the other is a typo in a quantization attribute name." + }, + { + "left": "issue:42936", + "right": "issue:43668", + "accept": false, + "reason": "Different artifact types and bugs: runtime loss behavior vs a config annotation mistake." + }, + { + "left": "issue:43240", + "right": "issue:43668", + "accept": false, + "reason": "Missing loss kwargs forwarding is not the same as an incorrect type hint in ModernBERTConfig." + }, + { + "left": "issue:43668", + "right": "issue:44031", + "accept": false, + "reason": "A type-hint bug and a tokenizer regex warning regression are unrelated." + }, + { + "left": "issue:43668", + "right": "issue:43813", + "accept": false, + "reason": "ModernBERT config typing issue vs a literal typo in a different conversion path." + } + ] + }, + "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": 13, + "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": "7f26a9e50cf93f54877d8d9b20dbe86c1c27e39deaa15fed3c9c8d86e8ebc753", + "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 are not duplicates: one is a generic request for block decoding, the other is a feature request for DeepSeek v3.2 support. They do not describe the same underlying bug or change.", + "confidence": 0.92, + "canonical_issue_reason": "Issue #41084 is the broader, more general request around block decoding and is the better representative of that topic.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue #41084 is the best issue candidate because it is the more foundational/general problem statement, while #43037 is model-specific.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:41084", + "right": "issue:43037", + "accept": false, + "reason": "The titles and likely scopes differ: block decoding versus DeepSeek v3.2 support. They are not the same bug or change and there is no explicit linkage or shared target." + } + ] + }, + "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": 13, + "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": "c24bd1fb854b312ac387e1c1df5f53f35cd2822c20b28cbdc1dbcd54acd2979e", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1583, + "estimated_input_tokens": 396, + "estimated_eval_tokens": 1048 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1583, + "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 two distinct Trainer-related feature requests: one about `remove_unused_column` customizability, the other about FLOPS estimation for custom trainers. They share only a broad subsystem and are not duplicates.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": "No single issue cleanly represents both topics; they are unrelated enough that neither should be treated as the canonical issue for the pair.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43598", + "right": "issue:43600", + "accept": false, + "reason": "Different underlying requests: one changes input-column pruning behavior, the other adjusts FLOPS estimation for custom trainers. Same area (Trainer) but not the same bug or 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": 13, + "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": "1dd9098894c31059fd5cb3a2fddeaef92e33859d1b73baa60e2353c63c5e07ef", + "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": "These are related only at a very high level (Qwen auto-class support), but they target different model families and different AutoModel APIs, so they are not duplicates.", + "confidence": 0.94, + "canonical_issue_reason": "Issue #43684 is broader and more general: it requests registration for Qwen3-Omni across core auto classes, making it the better representative if this cluster is grouped at all.", + "canonical_pr_reason": null, + "best_issue_reason": "#43684 is the better canonical issue because it describes a broader registration gap, while #44405 is a narrower task for a specific sequence-classification head on Qwen3.5.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43684", + "right": "issue:44405", + "accept": false, + "reason": "Different underlying changes: one asks for AutoModel/AutoModelForConditionalGeneration registration for Qwen3-Omni, the other asks for AutoModelForSequenceClassification support for Qwen3.5. They involve different model variants and different auto-class paths, so they should not be merged as the same bug/request." + } + ] + }, + "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": 13, + "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": "ca92f410b4da238dd16083ef795d5da4b6dcb52957138ef7dc6ad2bbffa8c25b", + "component_budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 6, + "serialized_chars": 4262, + "estimated_input_tokens": 1066, + "estimated_eval_tokens": 2388 + }, + "budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 6, + "serialized_chars": 4262, + "estimated_input_tokens": 1066, + "estimated_eval_tokens": 2388 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Three issues share the same generic title \"huggingface \" and look like near-duplicate placeholders; one issue titled \"Fett\" is unrelated. I would keep the earliest huggingface report as the representative issue, but confidence is limited because the bodies contain no actionable detail.", + "confidence": 0.57, + "canonical_issue_reason": "issue:43845 is the earliest of the near-identical \"huggingface \" reports and is a reasonable representative for the cluster.", + "canonical_pr_reason": null, + "best_issue_reason": "43845 is the best issue to retain because it is the earliest among the repeated huggingface-titled reports; however, the cluster is too underspecified to prove an exact duplicate set.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43835", + "right": "issue:43845", + "accept": false, + "reason": "Different titles; \"Fett\" does not look like the same underlying report as the huggingface placeholder issues." + }, + { + "left": "issue:43835", + "right": "issue:43846", + "accept": false, + "reason": "Different titles and no shared details beyond generic similarity." + }, + { + "left": "issue:43835", + "right": "issue:43859", + "accept": false, + "reason": "Different titles; no evidence these describe the same problem." + }, + { + "left": "issue:43845", + "right": "issue:43846", + "accept": false, + "reason": "The titles match, but the bodies are too sparse to confirm the same underlying bug with confidence." + }, + { + "left": "issue:43845", + "right": "issue:43859", + "accept": false, + "reason": "Same generic title, but the available text is too minimal to safely treat them as the same issue." + }, + { + "left": "issue:43846", + "right": "issue:43859", + "accept": false, + "reason": "Same generic title, but no concrete evidence shows they are the same underlying report." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 11, + "component_count": 13, + "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": "19f232c873586bd704c8948bd44c826959351073a84ace2172e5815a0e44ba47", + "component_budget": { + "node_count": 5, + "item_count": 5, + "soft_pair_count": 9, + "serialized_chars": 5791, + "estimated_input_tokens": 1448, + "estimated_eval_tokens": 3152 + }, + "budget": { + "node_count": 5, + "item_count": 5, + "soft_pair_count": 9, + "serialized_chars": 5791, + "estimated_input_tokens": 1448, + "estimated_eval_tokens": 3152 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are mostly spammy or placeholder issues with no evidence they describe the same underlying bug or request. None of the soft similarity links are strong enough to treat as duplicates.", + "confidence": 0.93, + "canonical_issue_reason": "issue:44418 is the only item that resembles a real user-facing request (a documentation enhancement) rather than obvious spam, so it is the least-bad issue to anchor the cluster if one must be chosen.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:44418 is the best standalone representative because it is the only non-spam-looking issue; the others are generic spam or unrelated text and should not be merged with it.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44371", + "right": "issue:44572", + "accept": false, + "reason": "Both titles are generic spam placeholders; no concrete shared bug or change is described." + }, + { + "left": "issue:44371", + "right": "issue:44573", + "accept": false, + "reason": "Generic spam titles with no actionable content; not the same underlying issue." + }, + { + "left": "issue:44371", + "right": "issue:44418", + "accept": false, + "reason": "One is spam, the other is a documentation enhancement suggestion; unrelated." + }, + { + "left": "issue:44371", + "right": "issue:44937", + "accept": false, + "reason": "Spam placeholder versus a specific quoted phrase; no duplicate signal." + }, + { + "left": "issue:44418", + "right": "issue:44572", + "accept": false, + "reason": "Documentation enhancement request and spam placeholder are not the same issue." + }, + { + "left": "issue:44418", + "right": "issue:44573", + "accept": false, + "reason": "Documentation enhancement request versus spam placeholder; no shared bug/change." + }, + { + "left": "issue:44572", + "right": "issue:44573", + "accept": false, + "reason": "Both are spam placeholders, but there is no evidence they refer to the same concrete problem." + }, + { + "left": "issue:44572", + "right": "issue:44937", + "accept": false, + "reason": "Spam placeholder versus a different quoted-text title; unrelated." + }, + { + "left": "issue:44573", + "right": "issue:44937", + "accept": false, + "reason": "Spam placeholder versus a different quoted-text title; not a duplicate." + } + ] + }, + "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": 13, + "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": "415c4aa3e00e24ce1931372625b1dcfeb68b94dc4d819f3e1ae3de82e9df4ae9", + "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": "The two closed issues have the same title and very similar body size/content, with no distinguishing links or filenames. They look like duplicate reports of the same Energy/N6 Arithmetic idea.", + "confidence": 0.98, + "canonical_issue_reason": "Issue #45145 is the better canonical because it has the same title as #45151 and slightly more inbound reference activity, suggesting it is the more connected duplicate target.", + "canonical_pr_reason": null, + "best_issue_reason": "#45145 is the best representative of this issue cluster since it appears equivalent to #45151 but has a bit more discussion/linkage.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45145", + "right": "issue:45151", + "accept": true, + "reason": "Same title, same apparent topic, and no evidence of a different underlying problem; these are duplicates." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The conclusions are grounded in the packet: the issues have identical titles, similar body lengths, and only soft similarity evidence, with no conflicting links or filenames. The canonical choice is also conservative and based on the slightly higher inbound reference count for #45145. The soft-edge accept is reasonable here." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 13, + "component_count": 13, + "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": "pull_request", + "component_index": 1, + "component_count": 23, + "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-39895-12", + "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:45587" + ], + "soft_pairs": [ + "pull_request:39895|pull_request:43424", + "pull_request:41855|pull_request:43823", + "pull_request:41855|pull_request:43424", + "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:43995", + "pull_request:43823|pull_request:44827", + "pull_request:43995|pull_request:45213", + "pull_request:43995|pull_request:44827", + "pull_request:44827|pull_request:45587", + "pull_request:44827|pull_request:45170", + "pull_request:44827|pull_request:45213", + "pull_request:43823|pull_request:45213", + "pull_request:43251|pull_request:43254" + ], + "prepared_review_unit_hash": "9bd24c94d579fabe4f58899fc0b189a5c31340ff58adfb148d48f1825e471544", + "component_budget": { + "node_count": 12, + "item_count": 12, + "soft_pair_count": 17, + "serialized_chars": 32981, + "estimated_input_tokens": 8246, + "estimated_eval_tokens": 16748 + }, + "budget": { + "node_count": 12, + "item_count": 12, + "soft_pair_count": 17, + "serialized_chars": 32981, + "estimated_input_tokens": 8246, + "estimated_eval_tokens": 16748 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Almost all pairs are unrelated across distinct model additions, refactors, tests, docs, or drafts. The only true duplicate is the loss-utils cross-entropy kwargs fix: 43251 and 43254 both target issue 43240 and patch the same function in the same code path.", + "confidence": 0.95, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 43251 is a good canonical representative for the only true duplicate pair because it and 43254 both target issue 43240 and modify src/transformers/loss/loss_utils.py in the same cross-entropy forwarding path; the choice is representative, not because it is materially more complete.", + "best_issue_reason": null, + "best_pr_reason": "PR 43251 is the best overall PR representative for deduping because it belongs to the only confirmed duplicate pair and is explicitly linked to issue 43240, making it a clean anchor for the shared loss-utils fix.", + "soft_edge_verdicts": [ + { + "left": "pull_request:39895", + "right": "pull_request:43424", + "accept": false, + "reason": "VideoPrism model addition vs executorch dynamic-shape export test; unrelated changes." + }, + { + "left": "pull_request:41855", + "right": "pull_request:43823", + "accept": false, + "reason": "Tokenizer method addition vs new MobileLLM model implementation; different features and code paths." + }, + { + "left": "pull_request:41855", + "right": "pull_request:43424", + "accept": false, + "reason": "Tokenizer API work vs executorch export testing; not the same underlying bug or change." + }, + { + "left": "pull_request:42668", + "right": "pull_request:43823", + "accept": false, + "reason": "Processor auto-construction robustness vs new model integration; unrelated." + }, + { + "left": "pull_request:43251", + "right": "pull_request:43823", + "accept": false, + "reason": "Cross-entropy kwargs fix vs MobileLLM model addition; no shared code-path problem." + }, + { + "left": "pull_request:43251", + "right": "pull_request:44827", + "accept": false, + "reason": "Loss utility change vs Mistral4 test fixes; different concrete issues." + }, + { + "left": "pull_request:43424", + "right": "pull_request:44827", + "accept": false, + "reason": "Executorch export test vs Mistral4 test fixes; both are tests but not the same fix." + }, + { + "left": "pull_request:43424", + "right": "pull_request:43823", + "accept": false, + "reason": "Executorch exportability test vs MobileLLM model implementation; unrelated." + }, + { + "left": "pull_request:43823", + "right": "pull_request:43995", + "accept": false, + "reason": "MobileLLM model addition vs Falcon output-interface refactor; different model families and change types." + }, + { + "left": "pull_request:43823", + "right": "pull_request:44827", + "accept": false, + "reason": "New MobileLLM model vs Mistral4 test fixes; no common bug/change." + }, + { + "left": "pull_request:43995", + "right": "pull_request:45213", + "accept": false, + "reason": "Falcon refactor vs draft model-creation skill content; not duplicates." + }, + { + "left": "pull_request:43995", + "right": "pull_request:44827", + "accept": false, + "reason": "Falcon output-collection refactor vs Mistral4 test fixes; unrelated." + }, + { + "left": "pull_request:44827", + "right": "pull_request:45587", + "accept": false, + "reason": "Mistral4 test fixes vs continuous batching docs update; different subsystems and purposes." + }, + { + "left": "pull_request:44827", + "right": "pull_request:45170", + "accept": false, + "reason": "Mistral4 test fixes vs CLIP-like weight-renaming cleanup; unrelated." + }, + { + "left": "pull_request:44827", + "right": "pull_request:45213", + "accept": false, + "reason": "Mistral4 test fixes vs draft model-creation skill; shared paths are incidental, not the same change." + }, + { + "left": "pull_request:43823", + "right": "pull_request:45213", + "accept": false, + "reason": "MobileLLM model addition vs model-creation skill draft; no shared underlying bug or feature." + }, + { + "left": "pull_request:43251", + "right": "pull_request:43254", + "accept": true, + "reason": "Both PRs patch src/transformers/loss/loss_utils.py to forward supported kwargs to nn.functional.cross_entropy for issue 43240; they are near-identical variants of the same fix." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "The duplicate-pair acceptance is grounded, but the canonical PR reasoning overstates the evidence: both 43251 and 43254 explicitly target issue 43240 and patch the same function, so the packet does not support claiming 43251 is 'more complete' than 43254. The rest of the soft-edge rejects look conservative." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 2, + "component_count": 23, + "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": "5eea437ee5dac868dd94391308c8598c023d2279d6c26568c4557a0784104e08", + "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 are related VibeVoice PRs, but they are not the same change: #40546 adds the base VibeVoice implementation, while #42781 adds a separate VibeVoice Realtime extension with additional acoustic-tokenizer and generation changes. They should not be treated as duplicates.", + "confidence": 0.89, + "canonical_issue_reason": null, + "canonical_pr_reason": "#40546 is the foundational, more complete VibeVoice implementation and has substantial review activity; it is the better representative of the main model addition.", + "best_issue_reason": null, + "best_pr_reason": "#40546 is the best cluster representative because it covers the core VibeVoice codepath and docs, whereas #42781 is a narrower follow-up for realtime support.", + "soft_edge_verdicts": [ + { + "left": "pull_request:40546", + "right": "pull_request:42781", + "accept": false, + "reason": "Same model family, but different concrete scope: base VibeVoice implementation vs. realtime variant with extra tokenizer/generation work. Not the same mergeable 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": 3, + "component_count": 23, + "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": "89411a8e362a683daa60f23c7aafa79d12a1e1ef588cc0fcbb617fe6f7cf58df", + "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 cluster contains two unrelated pull requests: one adds covariate forecasting for TimesFM, the other adds DINOv3 image classification support. They touch different models, code paths, and documentation, 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": "pull_request:40755", + "right": "pull_request:41224", + "accept": false, + "reason": "Different model families and different concrete features: TimesFM covariate regression vs DINOv3 image classification support. No shared bug/change target, and they would not plausibly be one merged 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": 4, + "component_count": 23, + "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": "83aaab8646470ad87eff0522aecb453ed5557310ff76056dd9f5eca14805867e", + "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": "Both PRs are for adding DEIMv2, but #44339 is the substantially more complete integration (auto mappings, converter, modular files, loss/utils updates, config checks) whereas #41356 is an earlier, narrower model/image-processor/test drop-in. They overlap in files and target the same issue, but they are not a clean duplicate pair.", + "confidence": 0.88, + "canonical_issue_reason": null, + "canonical_pr_reason": "#44339 is the better canonical PR because it is the fuller end-state implementation: it adds the model plus conversion tooling, auto-mapping wiring, loss/util updates, modular code, docs, and broader test coverage.", + "best_issue_reason": null, + "best_pr_reason": "#44339 is the best PR for the cluster because it most completely represents the DEIMv2 addition and integration into Transformers.", + "soft_edge_verdicts": [ + { + "left": "pull_request:41356", + "right": "pull_request:44339", + "accept": false, + "reason": "Shared DEIMv2 scope and issue target, but #44339 is a broader follow-up/superset rather than the same concrete change; the overlap is too coarse to treat them 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": 5, + "component_count": 23, + "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-41797-2", + "nodes": [ + "pull_request:41797", + "pull_request:45186" + ], + "soft_pairs": [ + "pull_request:41797|pull_request:45186" + ], + "prepared_review_unit_hash": "ade0862fa5b40ee50a77ef3df4cf3e247c883fdfb6bee5f5a9b8bac09c191aae", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 6723, + "estimated_input_tokens": 1681, + "estimated_eval_tokens": 3618 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 6723, + "estimated_input_tokens": 1681, + "estimated_eval_tokens": 3618 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Both pull requests add entirely different model integrations (DeepSeek OCR vs. Isaac). They share only the usual Transformers boilerplate and registry updates, so they are not the same underlying change and should not be clustered as duplicates.", + "confidence": 0.94, + "canonical_issue_reason": null, + "canonical_pr_reason": "No single PR is a true canonical for this pair because they target different models and separate codepaths; if forced, the DeepSeek OCR PR is the earlier, more established integration but it is not a duplicate of Isaac.", + "best_issue_reason": null, + "best_pr_reason": "PR 41797 is the better representative only by convention: it is the earlier, more complete-looking model addition with broader review activity. However, it is not equivalent to 45186 and should not be treated as a duplicate of it.", + "soft_edge_verdicts": [ + { + "left": "pull_request:41797", + "right": "pull_request:45186", + "accept": false, + "reason": "Different model additions for different architectures (DeepSeek OCR vs. Isaac). Shared files are only autogenerated registries/docs that appear in many model-integration PRs, not evidence of the same 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": 6, + "component_count": 23, + "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-42166-2", + "nodes": [ + "pull_request:42166", + "pull_request:45426" + ], + "soft_pairs": [ + "pull_request:42166|pull_request:45426" + ], + "prepared_review_unit_hash": "0e3ce74c15c13905f748f482878fa9c11bf16b4a3107f6b7950a1f80fc152023", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5408, + "estimated_input_tokens": 1352, + "estimated_eval_tokens": 2960 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5408, + "estimated_input_tokens": 1352, + "estimated_eval_tokens": 2960 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Both items are feature PRs adding different models (InternVLFlash vs AXK1). They overlap only in shared registration/auto-discovery files, not in the underlying code change, so they are not duplicates.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": "No single canonical PR fits this cluster: the PRs introduce different models with different implementations and test/docs surfaces.", + "best_issue_reason": null, + "best_pr_reason": "No global best PR is appropriate here because the PRs are unrelated feature additions; the overlap is limited to shared model registry plumbing.", + "soft_edge_verdicts": [ + { + "left": "pull_request:42166", + "right": "pull_request:45426", + "accept": false, + "reason": "Reject: these are different model-addition PRs for different code paths and cannot plausibly be merged into one duplicate 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": 7, + "component_count": 23, + "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-42227-4", + "nodes": [ + "pull_request:42227", + "pull_request:43291", + "pull_request:43665", + "pull_request:44815" + ], + "soft_pairs": [ + "pull_request:42227|pull_request:43291", + "pull_request:42227|pull_request:43665", + "pull_request:42227|pull_request:44815", + "pull_request:43291|pull_request:43665", + "pull_request:43291|pull_request:44815", + "pull_request:43665|pull_request:44815" + ], + "prepared_review_unit_hash": "98c092ccb827bdf8fac102a0302171bd8565b7b3c8a930b4b1b511b61e369372", + "component_budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 6, + "serialized_chars": 10170, + "estimated_input_tokens": 2543, + "estimated_eval_tokens": 5342 + }, + "budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 6, + "serialized_chars": 10170, + "estimated_input_tokens": 2543, + "estimated_eval_tokens": 5342 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are four unrelated PRs that only share loose topical similarity (mostly Transformers/Whisper/test changes). They do not look like duplicates of the same underlying bug or change.", + "confidence": 0.89, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 42227 is the cleanest canonical candidate: it has an explicit issue target, a focused code-path fix in the ASR pipeline, and a clear bug description.", + "best_issue_reason": null, + "best_pr_reason": "PR 42227 is the most representative and mergeable-looking change in the set; the others are either unrelated test edits or a different FP8/dequantization effort.", + "soft_edge_verdicts": [ + { + "left": "pull_request:42227", + "right": "pull_request:43291", + "accept": false, + "reason": "Both mention Whisper, but one fixes ASR return_language plumbing while the other changes Whisper tokenizer/test behavior. Different code paths and different bugs." + }, + { + "left": "pull_request:42227", + "right": "pull_request:43665", + "accept": false, + "reason": "42227 is a Whisper pipeline fix; 43665 is unrelated model test edits for CLIP/ViT. No shared underlying issue." + }, + { + "left": "pull_request:42227", + "right": "pull_request:44815", + "accept": false, + "reason": "Whisper return-language plumbing is unrelated to FP8/dequantization/model-loading work. Different subsystem and bug class." + }, + { + "left": "pull_request:43291", + "right": "pull_request:43665", + "accept": false, + "reason": "Tokenizer/test changes for Whisper do not match CLIP/ViT test edits. Similar in being tests, but not the same change or bug." + }, + { + "left": "pull_request:43291", + "right": "pull_request:44815", + "accept": false, + "reason": "Whisper decode/test cleanup and FP8 dequant fix are unrelated code paths and should not be merged as duplicates." + }, + { + "left": "pull_request:43665", + "right": "pull_request:44815", + "accept": false, + "reason": "The shared similarity is only that both touch tests somewhere, but the actual changes target unrelated model families and problems." + } + ] + }, + "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": 23, + "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": "ff018e71bbef2f94bf22e7703b445616f369c8290cf00ecac795da660087aeec", + "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": "Two unrelated GitHub Actions maintenance PRs: one upgrades many workflow actions for Node 24 compatibility, the other bumps a smaller set of workflow/action versions, mainly docker/build-push-action. They overlap in files but do not look like the same fix or change.", + "confidence": 0.92, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #42975 is the better canonical representative because it is the broader, more clearly scoped workflow-action upgrade aimed at Node 24 compatibility, touching many of the same automation files.", + "best_issue_reason": null, + "best_pr_reason": "PR #42975 is the best overall representative of the cluster: it has the clearest stated purpose, the widest workflow coverage, and summarizes the common theme of action-version maintenance.", + "soft_edge_verdicts": [ + { + "left": "pull_request:42975", + "right": "pull_request:42976", + "accept": false, + "reason": "Both are workflow dependency bumps, but they upgrade different actions for different reasons: #42975 targets Node 24 compatibility across many workflows, while #42976 mainly updates docker/build-push-action to v6. Shared workflow filenames alone are not enough to treat them as 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": 9, + "component_count": 23, + "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": "6e144bf9570e2f5eb42f778e4a10f9079c2d1dc41c36d893b1121a5bb28f73a6", + "component_budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 6, + "serialized_chars": 14136, + "estimated_input_tokens": 3534, + "estimated_eval_tokens": 7324 + }, + "budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 6, + "serialized_chars": 14136, + "estimated_input_tokens": 3534, + "estimated_eval_tokens": 7324 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These four PRs are separate model-integration additions (ViT NEPA, Molmo, Molmo2, and CTSM). They share some boilerplate paths like docs and auto-registration, but they do not describe the same bug or the same code change.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": "No canonical PR: the four PRs target different model families and different implementation files, so none is a duplicate representative.", + "best_issue_reason": null, + "best_pr_reason": "PR 43451 is the broadest single integration, with the most files and review activity, but it is still not a duplicate of the others.", + "soft_edge_verdicts": [ + { + "left": "pull_request:42978", + "right": "pull_request:43448", + "accept": false, + "reason": "ViT NEPA vs Molmo: different model integrations and different code paths; overlap is only in shared framework boilerplate." + }, + { + "left": "pull_request:42978", + "right": "pull_request:43451", + "accept": false, + "reason": "ViT NEPA vs Molmo2 are unrelated model additions, not the same fix or feature." + }, + { + "left": "pull_request:42978", + "right": "pull_request:45490", + "accept": false, + "reason": "ViT NEPA and CTSM add different models with different implementations; not mergeable as one duplicate PR." + }, + { + "left": "pull_request:43448", + "right": "pull_request:43451", + "accept": false, + "reason": "Molmo vs Molmo2 are distinct model families; shared docs/auto-registration does not make them duplicates." + }, + { + "left": "pull_request:43448", + "right": "pull_request:45490", + "accept": false, + "reason": "Molmo and CTSM are separate model integrations touching different model code." + }, + { + "left": "pull_request:43451", + "right": "pull_request:45490", + "accept": false, + "reason": "Molmo2 and CTSM are different model additions with no shared underlying bug or concrete code-path 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": 23, + "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-43488-3", + "nodes": [ + "pull_request:43488", + "pull_request:43532", + "pull_request:45476" + ], + "soft_pairs": [ + "pull_request:43488|pull_request:43532", + "pull_request:43488|pull_request:45476", + "pull_request:43532|pull_request:45476" + ], + "prepared_review_unit_hash": "c49b0af034f5601017c196ac612076acbebef736d291e1e6baaa8e38bd72d053", + "component_budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 3, + "serialized_chars": 7469, + "estimated_input_tokens": 1868, + "estimated_eval_tokens": 3992 + }, + "budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 3, + "serialized_chars": 7469, + "estimated_input_tokens": 1868, + "estimated_eval_tokens": 3992 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "All three items are unrelated placeholder PRs that only overlap on an incidental test file. They do not appear to fix the same underlying bug or change, so no soft-edge merge is warranted.", + "confidence": 0.95, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 43532 is the most central of the three: it touches multiple CI/workflow-related files and has the highest discussion/review activity, while the other two are narrower and look even more synthetic.", + "best_issue_reason": null, + "best_pr_reason": "PR 43532 is the best representative of the cluster because it has the broadest change surface and connects to both other PRs via weak similarity, but the overlap is still incidental rather than a shared fix.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43488", + "right": "pull_request:43532", + "accept": false, + "reason": "These are different placeholder changes: one is about setup/version/dependency metadata, the other is about workflow/comment CI plumbing. Shared vit test presence is incidental and not the same code-path problem." + }, + { + "left": "pull_request:43488", + "right": "pull_request:45476", + "accept": false, + "reason": "One PR changes packaging/dependency setup, the other adds a CI workflow caller and unrelated tokenizer/test noise. No concrete shared bug or mergeable combined fix is evident." + }, + { + "left": "pull_request:43532", + "right": "pull_request:45476", + "accept": false, + "reason": "Both touch CI-adjacent files, but they address different areas: comment/scheduled/slack workflow logic versus a new PR CI caller workflow plus tokenizer/test edits. The overlap is too broad and superficial 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": 11, + "component_count": 23, + "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": "814e77809d208172e5bc54330b49f4b5c45becffc7de3b8b6ac40df6a6bb140c", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 6083, + "estimated_input_tokens": 1521, + "estimated_eval_tokens": 3298 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 6083, + "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 feature PRs: one adds a Promptable Concept Segmentation pipeline, the other adds a Promptable Visual Segmentation pipeline. They share documentation and pipeline infrastructure changes, but target different tasks/models and are not the same underlying change.", + "confidence": 0.93, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #43613 is the better canonical representative for the cluster because it is the broader pipeline addition, with more model integrations and test coverage, while #43612 is a narrower sibling feature.", + "best_issue_reason": null, + "best_pr_reason": "PR #43613 is the best representative overall for this cluster: it has the wider scope, touches more related code paths, and is the more central of the two additions.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43612", + "right": "pull_request:43613", + "accept": false, + "reason": "Do not merge: these add different pipelines for different segmentation tasks. The overlap is mostly shared docs/registry plumbing, not the same concrete 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": 12, + "component_count": 23, + "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": "7525d53a7ac22053874b457dedc8a829a6000c4690c14e75d5d8a9b5e3169d11", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4292, + "estimated_input_tokens": 1073, + "estimated_eval_tokens": 2402 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4292, + "estimated_input_tokens": 1073, + "estimated_eval_tokens": 2402 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Two unrelated open PRs in `trainer.py`: one changes gradient-accumulation loss scaling behavior, the other refactors signature-column filtering for custom trainers/PEFT models. They share a file and subsystem but not the same underlying bug or change.", + "confidence": 0.96, + "canonical_issue_reason": null, + "canonical_pr_reason": "No canonical PR: these are distinct feature/refactor PRs with different code paths and goals, not duplicates of one another.", + "best_issue_reason": null, + "best_pr_reason": "Neither PR is a duplicate candidate; if a representative is needed, `#43651` is slightly more central to trainer training-step behavior, but it should not be treated as the same change as `#43663`.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43651", + "right": "pull_request:43663", + "accept": false, + "reason": "Both touch `trainer.py`, but one modifies loss scaling in `training_step` and the other adds an overrideable signature-column helper for data filtering. Different bugs/changes, not mergeable as 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": 13, + "component_count": 23, + "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": "aaae9ca124a545dcbb9947a6aed2b8214570914fb751060eaf028752f2cbe194", + "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": "All three PRs are small, overlapping fixes in `src/transformers/cli/serve.py` around `TypeAdapter`/typing evaluation at import time. PR 43842 is the cleanest and most directly scoped fix; 43836 looks like a close variant of the same bug; 43656 is similar but less clearly the same concrete failure mode.", + "confidence": 0.74, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 43842 is the most focused and explicit fix for the runtime `TypeAdapter` NameError, with minimal scope and a clear explanation tied to pydantic being absent.", + "best_issue_reason": null, + "best_pr_reason": "PR 43842 best represents the cluster because it directly addresses the concrete import-time `TypeAdapter` NameError with the smallest, most targeted change set.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43656", + "right": "pull_request:43842", + "accept": false, + "reason": "Both touch the same file and mention `TypeAdapter`, but 43656 appears to be a broader typing/annotation cleanup and has a different issue target; the evidence is not strong enough to say it is the same exact bug/fix." + }, + { + "left": "pull_request:43836", + "right": "pull_request:43842", + "accept": true, + "reason": "These are very likely the same underlying `TypeAdapter` NameError fix in `serve.py`: both target the same issue, both are tiny annotation-only changes, and both avoid runtime evaluation of `TypeAdapter`." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary stays within the packet evidence and is suitably conservative. PR 43842 is clearly the smallest, most targeted TypeAdapter fix. The 43836/43842 soft-accept is defensible because they share the same issue target and file and appear to be alternative solutions to the same import-time typing problem. The 43656/43842 reject is also conservative given the weaker linkage and different issue target." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 14, + "component_count": 23, + "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": 15, + "component_count": 23, + "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-43973-2", + "nodes": [ + "pull_request:43973", + "pull_request:45283" + ], + "soft_pairs": [ + "pull_request:43973|pull_request:45283" + ], + "prepared_review_unit_hash": "943436408822dc567b5d61dd36502cbe2f8f715d5c54c74363f63b0d61951814", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5195, + "estimated_input_tokens": 1299, + "estimated_eval_tokens": 2854 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5195, + "estimated_input_tokens": 1299, + "estimated_eval_tokens": 2854 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These two PRs are unrelated: one adds the new Lfm2Audio model package and docs, while the other adds Qwen3.5 GGUF loading support via integration mappings. They do not target the same bug or code path.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 43973 is the better canonical PR candidate for this cluster only because it is a self-contained model addition with an explicit issue target and substantial implementation/test/docs coverage.", + "best_issue_reason": null, + "best_pr_reason": "PR 43973 is the strongest representative of the cluster, but the cluster is not a duplicate set; PR 45283 is a separate GGUF-loading change for a different model family.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43973", + "right": "pull_request:45283", + "accept": false, + "reason": "Different features and code paths: Lfm2Audio model/docs/tests versus Qwen3.5 GGUF loading mappings. Shared transformer infrastructure is too broad to indicate 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": 16, + "component_count": 23, + "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:44059|pull_request:44722", + "pull_request:44085|pull_request:44086", + "pull_request:44098|pull_request:44101", + "pull_request:44018|pull_request:44071", + "pull_request:44017|pull_request:44722", + "pull_request:44017|pull_request:44073", + "pull_request:44017|pull_request:44161", + "pull_request:44068|pull_request:44071", + "pull_request:44071|pull_request:44086", + "pull_request:44059|pull_request:44071", + "pull_request:43997|pull_request:44073", + "pull_request:44017|pull_request:44071", + "pull_request:44059|pull_request:44068", + "pull_request:44073|pull_request:44101", + "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:44017|pull_request:44018", + "pull_request:43997|pull_request:44056", + "pull_request:44071|pull_request:44076", + "pull_request:44074|pull_request:44101", + "pull_request:43997|pull_request:44018", + "pull_request:44056|pull_request:44161", + "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": "cd98c8fbcb8a4ff605858352e22680c7229a3a18d58c4e2a3526abbd1ced83ea", + "component_budget": { + "node_count": 28, + "item_count": 28, + "soft_pair_count": 77, + "serialized_chars": 84503, + "estimated_input_tokens": 21126, + "estimated_eval_tokens": 42508 + }, + "budget": { + "node_count": 28, + "item_count": 28, + "soft_pair_count": 72, + "serialized_chars": 76936, + "estimated_input_tokens": 19234, + "estimated_eval_tokens": 38724 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a refactor series for standardized output tracing across many model implementations. Most pairs are not true duplicates because they touch different models/code paths. The only clear duplicate-like subclusters are GPT-Neo (#44018/#44068) and the GPT-J family (#44066/#44085/#44722), which overlap on the same model files and the same decorator-based output-tracing change.", + "confidence": 0.79, + "canonical_issue_reason": "Issue #43979 appears to be the umbrella tracking issue for the standardized output-tracing refactor across models, which explains the shared lineage of many PRs in this cluster.", + "canonical_pr_reason": "PR #44722 is the best canonical PR: it is the most complete representative of the GPT-J/CodeGen output-tracing refactor, directly targets the shared issue, and covers the same concrete decorator-based code path as the overlapping GPT-J PRs.", + "best_issue_reason": "#43979 is the best issue anchor because it is the shared tracker for the broader refactor work, even though it is an umbrella rather than a narrow bug report.", + "best_pr_reason": "#44722 is the strongest single PR to keep as the representative example of the shared change set because it is explicit, broader than the earlier GPT-J-only variants, and aligns with the same implementation path.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43996", + "right": "pull_request:44085", + "accept": false, + "reason": "Different models and files (CVT/FNet vs GPT-J); same umbrella refactor theme, but not the same concrete code path." + }, + { + "left": "pull_request:43996", + "right": "pull_request:44044", + "accept": false, + "reason": "Both are output-tracing refactors, but they target unrelated model implementations (CVT/FNet vs DeBERTa-v2)." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44085", + "accept": true, + "reason": "Both touch GPT-J output tracing in the same model file and appear to be overlapping/superset variants of the same change." + }, + { + "left": "pull_request:44007", + "right": "pull_request:44072", + "accept": false, + "reason": "RegNet vs EfficientNet: same style of refactor, but different model code paths." + }, + { + "left": "pull_request:44072", + "right": "pull_request:44722", + "accept": false, + "reason": "EfficientNet and GPT-J/CodeGen are unrelated model implementations; shared issue link is not enough." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44072", + "accept": false, + "reason": "GPT-J/CodeGen vs EfficientNet are separate model refactors, not the same underlying bug or change." + }, + { + "left": "pull_request:44013", + "right": "pull_request:44044", + "accept": false, + "reason": "MobileNetV2 and DeBERTa-v2 are different code paths despite similar output-tracing cleanup." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44086", + "accept": false, + "reason": "GPT-J/CodeGen refactor vs MGP-STR refactor; different model internals." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44071", + "accept": false, + "reason": "GPT-J/CodeGen and MPT are separate architectures; not the same change." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44068", + "accept": true, + "reason": "Both are GPT-Neo output-tracing refactors over the same model file/test set and look like duplicate/superset variants." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44068", + "accept": false, + "reason": "GPT-J/CodeGen vs GPT-Neo: similar refactor pattern, but different model code paths." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44066", + "accept": false, + "reason": "GPT-Neo and GPT-J/CodeGen are distinct model refactors." + }, + { + "left": "pull_request:44019", + "right": "pull_request:44722", + "accept": false, + "reason": "ResNet and GPT-J/CodeGen are unrelated; same issue target does not make them duplicates." + }, + { + "left": "pull_request:44086", + "right": "pull_request:44722", + "accept": false, + "reason": "MGP-STR and GPT-J/CodeGen are different model implementations and not mergeable as one change." + }, + { + "left": "pull_request:44019", + "right": "pull_request:44085", + "accept": false, + "reason": "ResNet vs GPT-J output tracing; different code paths." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44019", + "accept": false, + "reason": "GPT-Neo and ResNet are unrelated model refactors." + }, + { + "left": "pull_request:44019", + "right": "pull_request:44071", + "accept": false, + "reason": "ResNet vs MPT: same refactor theme, different concrete implementation." + }, + { + "left": "pull_request:44068", + "right": "pull_request:44116", + "accept": false, + "reason": "GPT-Neo and Flaubert/XLM are separate model code paths." + }, + { + "left": "pull_request:44068", + "right": "pull_request:44722", + "accept": false, + "reason": "GPT-Neo is a different model family from GPT-J/CodeGen." + }, + { + "left": "pull_request:44071", + "right": "pull_request:44722", + "accept": false, + "reason": "MPT and GPT-J/CodeGen are unrelated implementations." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44129", + "accept": false, + "reason": "GPT-Neo vs SpeechT5: different model families, only a shared refactor style." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44086", + "accept": false, + "reason": "GPT-Neo and MGP-STR are different code paths." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44085", + "accept": false, + "reason": "GPT-Neo and GPT-J are related only at a high level; not the same model change." + }, + { + "left": "pull_request:44068", + "right": "pull_request:44086", + "accept": false, + "reason": "GPT-Neo and MGP-STR are unrelated model refactors." + }, + { + "left": "pull_request:44071", + "right": "pull_request:44085", + "accept": false, + "reason": "MPT vs GPT-J are distinct model implementations." + }, + { + "left": "pull_request:44059", + "right": "pull_request:44086", + "accept": false, + "reason": "GPT-2 and MGP-STR are different models; not a duplicate." + }, + { + "left": "pull_request:44059", + "right": "pull_request:44722", + "accept": false, + "reason": "GPT-2 output-tracing refactor is separate from GPT-J/CodeGen." + }, + { + "left": "pull_request:44085", + "right": "pull_request:44086", + "accept": false, + "reason": "GPT-J and MGP-STR are unrelated code paths." + }, + { + "left": "pull_request:44098", + "right": "pull_request:44101", + "accept": false, + "reason": "ViLT and Flaubert/XLM are different model families." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44071", + "accept": false, + "reason": "GPT-Neo and MPT are separate implementations." + }, + { + "left": "pull_request:44017", + "right": "pull_request:44722", + "accept": false, + "reason": "SegFormer and GPT-J/CodeGen are unrelated model refactors." + }, + { + "left": "pull_request:44017", + "right": "pull_request:44073", + "accept": false, + "reason": "SegFormer vs VisualBert: same pattern, different code paths." + }, + { + "left": "pull_request:44017", + "right": "pull_request:44161", + "accept": false, + "reason": "SegFormer and LongT5 are different models; not a true duplicate." + }, + { + "left": "pull_request:44068", + "right": "pull_request:44071", + "accept": false, + "reason": "GPT-Neo and MPT are different architectures." + }, + { + "left": "pull_request:44071", + "right": "pull_request:44076", + "accept": false, + "reason": "MPT and ImageGPT are unrelated code paths." + }, + { + "left": "pull_request:44074", + "right": "pull_request:44101", + "accept": false, + "reason": "TextNet and Flaubert/XLM are different model implementations." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44056", + "accept": false, + "reason": "RegNet and MPNet are separate model refactors." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44018", + "accept": false, + "reason": "RegNet and GPT-Neo are unrelated." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44161", + "accept": false, + "reason": "RegNet and LongT5 are different model code paths." + }, + { + "left": "pull_request:44056", + "right": "pull_request:44161", + "accept": false, + "reason": "MPNet and LongT5 are unrelated model-specific changes." + }, + { + "left": "pull_request:43996", + "right": "pull_request:44161", + "accept": false, + "reason": "CVT/FNet and LongT5 are different implementations; shared refactor theme is not enough." + }, + { + "left": "pull_request:44010", + "right": "pull_request:44071", + "accept": false, + "reason": "SqueezeBert and MPT are different model code paths." + }, + { + "left": "pull_request:44024", + "right": "pull_request:44076", + "accept": false, + "reason": "FocalNet and ImageGPT are separate model implementations; same output-tracing style only." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "Grounded in the packet, and the soft-edge accepts are conservative: both accepted pairs are same-model, same-file overlaps that look mergeable into one PR for the same refactor. The rejections are appropriately cautious about different model code paths." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 16, + "component_count": 23, + "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": "8f6f4ead4fb877334034eba773bcf856d0142d83356f688fa2bd8971c0ecf2f5", + "component_budget": { + "node_count": 28, + "item_count": 28, + "soft_pair_count": 77, + "serialized_chars": 84503, + "estimated_input_tokens": 21126, + "estimated_eval_tokens": 42508 + }, + "budget": { + "node_count": 6, + "item_count": 6, + "soft_pair_count": 5, + "serialized_chars": 13067, + "estimated_input_tokens": 3267, + "estimated_eval_tokens": 6790 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "All items are model-specific refactors toward standardized output tracing/capture_outputs. They share a theme, but each PR targets a different architecture and code path, so they are not true duplicates.", + "confidence": 0.93, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 44129 is the best cluster representative: it has an explicit issue target, a clear output-tracing refactor scope, and enough context to anchor the shared theme without conflating distinct model implementations.", + "best_issue_reason": null, + "best_pr_reason": "PR 44129 is the strongest single representative because it is explicitly linked to an issue and cleanly expresses the standardized output capture refactor pattern seen across the cluster.", + "soft_edge_verdicts": [ + { + "left": "pull_request:44073", + "right": "pull_request:44074", + "accept": false, + "reason": "Both are output-tracing refactors, but they touch different model implementations (VisualBert vs TextNet) and different forward paths; they are related patterns, not the same concrete change." + }, + { + "left": "pull_request:44056", + "right": "pull_request:44073", + "accept": false, + "reason": "These are separate model-specific refactors in MPNet and VisualBert. Similar decorator/capture_outputs work is not enough to treat them as one duplicate PR." + }, + { + "left": "pull_request:44056", + "right": "pull_request:44074", + "accept": false, + "reason": "MPNet and TextNet are different code paths with independent refactors; they may share a migration pattern, but they are not the same underlying bug or change." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44010", + "accept": false, + "reason": "RegNet and SqueezeBert are unrelated model files and implementations. The diff theme is similar, but this is not one mergeable PR for the same fix." + }, + { + "left": "pull_request:44010", + "right": "pull_request:44129", + "accept": false, + "reason": "SqueezeBert and SpeechT5 each implement their own output-tracing refactor in separate model code paths; this is a shared migration style, not a duplicate 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": 17, + "component_count": 23, + "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: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": "e8084fb0789b2c511b36c9fcee64ab494a101e04502b2f87854312a4b2b24e80", + "component_budget": { + "node_count": 13, + "item_count": 13, + "soft_pair_count": 39, + "serialized_chars": 39782, + "estimated_input_tokens": 9946, + "estimated_eval_tokens": 20148 + }, + "budget": { + "node_count": 13, + "item_count": 13, + "soft_pair_count": 39, + "serialized_chars": 39782, + "estimated_input_tokens": 9946, + "estimated_eval_tokens": 20148 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These PRs are all model-specific implementations of the same broad output-tracing refactor theme, but they target different architectures and code paths. They do not look like duplicates of one another, so no soft edges should be merged.", + "confidence": 0.91, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 44003 is the strongest representative of the cluster because it makes the most infrastructure-like tracing changes and spans closely related Mamba/Falcon-Mamba paths, rather than only applying the pattern in one model.", + "best_issue_reason": null, + "best_pr_reason": "44003 is the clearest anchor PR for this cluster: it has the broadest tracing machinery changes and best captures the underlying refactor theme.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43998", + "right": "pull_request:44002", + "accept": false, + "reason": "Same refactor pattern, but different model implementations (`timm_backbone` vs `upernet`) and no shared concrete code path." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44000", + "accept": false, + "reason": "Both are output-tracing refactors, but they affect different models and different forward paths, so they are not the same change." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44001", + "accept": false, + "reason": "Different model families (`timm_backbone` vs `univnet`) with separate forward logic; similar cleanup, not the same bug or patch." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44003", + "accept": false, + "reason": "These touch unrelated model code paths (`timm_backbone` vs `mamba/falcon_mamba`), so they are not mergeable as one PR." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44029", + "accept": false, + "reason": "Shared theme only; `timm_backbone` and `rwkv` are separate architectures with different tracing behavior." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44001", + "accept": false, + "reason": "Both refactor output handling, but `mobilenet_v1` and `univnet` are distinct implementations with no shared underlying bug." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44002", + "accept": false, + "reason": "`mobilenet_v1` and `upernet` are separate model paths; this is the same pattern applied independently." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44025", + "accept": false, + "reason": "Different model stacks (`mobilenet_v1` vs `depth_anything/prompt_depth_anything`); no common code path." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44026", + "accept": false, + "reason": "These are unrelated encoder/decoder and classification-style model paths, despite both adding tracing helpers." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44027", + "accept": false, + "reason": "`mobilenet_v1` and `speech_encoder_decoder` are different subsystems; not the same underlying fix." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44001", + "accept": false, + "reason": "`vision_text_dual_encoder` and `univnet` have different forward paths and output structures, so this is not one duplicate change." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44002", + "accept": false, + "reason": "Same refactor motif, but different models and APIs; no shared concrete code path." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44003", + "accept": false, + "reason": "`vision_text_dual_encoder` and `mamba/falcon_mamba` are unrelated code paths, so they should not be merged together." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44025", + "accept": false, + "reason": "These modify different model families and different output contracts; only the general theme matches." + }, + { + "left": "pull_request:44001", + "right": "pull_request:44002", + "accept": false, + "reason": "`univnet` and `upernet` are separate model implementations; no shared bug or single PR target." + }, + { + "left": "pull_request:44001", + "right": "pull_request:44003", + "accept": false, + "reason": "Different architectures and different forward behavior; not the same underlying change." + }, + { + "left": "pull_request:44001", + "right": "pull_request:44025", + "accept": false, + "reason": "`univnet` and `depth_anything` are unrelated code paths, so this is not a duplicate." + }, + { + "left": "pull_request:44002", + "right": "pull_request:44029", + "accept": false, + "reason": "`upernet` and `rwkv` are separate model families; similar output-tracing work is not enough to merge them." + }, + { + "left": "pull_request:44002", + "right": "pull_request:44003", + "accept": false, + "reason": "These touch unrelated model internals and do not fix the same concrete issue." + }, + { + "left": "pull_request:44003", + "right": "pull_request:44004", + "accept": false, + "reason": "Both are refactors, but `mamba/falcon_mamba` and `codegen` are distinct code paths with different output semantics." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44004", + "accept": false, + "reason": "`timm_backbone` and `codegen` are unrelated model implementations; same theme, different changes." + }, + { + "left": "pull_request:44001", + "right": "pull_request:44004", + "accept": false, + "reason": "`univnet` and `codegen` are separate model families, so they are not the same underlying fix." + }, + { + "left": "pull_request:44002", + "right": "pull_request:44004", + "accept": false, + "reason": "`upernet` and `codegen` are unrelated forward paths and cannot plausibly be one merged PR." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44004", + "accept": false, + "reason": "`vision_text_dual_encoder` and `codegen` differ in both architecture and output handling; not a duplicate." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44026", + "accept": false, + "reason": "Different model families (`depth_anything` vs `vision_encoder_decoder`) and separate output-tracing changes." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44027", + "accept": false, + "reason": "`depth_anything` and `speech_encoder_decoder` are distinct code paths; these are parallel refactors, not one change." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44028", + "accept": false, + "reason": "`depth_anything` and `superpoint` are unrelated model implementations with different output structures." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44029", + "accept": false, + "reason": "These target different architectures (`depth_anything` vs `rwkv`), so they are not the same underlying bug or patch." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44030", + "accept": false, + "reason": "`depth_anything` and `dpr` are separate models; similar refactoring does not make them duplicates." + }, + { + "left": "pull_request:44026", + "right": "pull_request:44029", + "accept": false, + "reason": "`vision_encoder_decoder` and `rwkv` are different code paths with no shared fix." + }, + { + "left": "pull_request:44026", + "right": "pull_request:44027", + "accept": false, + "reason": "These update different encoder-decoder stacks and should be triaged separately." + }, + { + "left": "pull_request:44026", + "right": "pull_request:44028", + "accept": false, + "reason": "`vision_encoder_decoder` and `superpoint` are unrelated model implementations." + }, + { + "left": "pull_request:44026", + "right": "pull_request:44030", + "accept": false, + "reason": "Different subsystems (`vision_encoder_decoder` vs `dpr`); same pattern, not the same change." + }, + { + "left": "pull_request:44027", + "right": "pull_request:44028", + "accept": false, + "reason": "`speech_encoder_decoder` and `superpoint` do not share a concrete code path or bug." + }, + { + "left": "pull_request:44027", + "right": "pull_request:44029", + "accept": false, + "reason": "These are separate model families (`speech_encoder_decoder` vs `rwkv`) with independent tracing adjustments." + }, + { + "left": "pull_request:44027", + "right": "pull_request:44030", + "accept": false, + "reason": "`speech_encoder_decoder` and `dpr` are unrelated implementations, so not a duplicate pair." + }, + { + "left": "pull_request:44028", + "right": "pull_request:44029", + "accept": false, + "reason": "`superpoint` and `rwkv` are different model paths; no single concrete fix connects them." + }, + { + "left": "pull_request:44028", + "right": "pull_request:44030", + "accept": false, + "reason": "These are distinct model implementations with different output contracts, so they should not be merged." + }, + { + "left": "pull_request:44029", + "right": "pull_request:44030", + "accept": false, + "reason": "`rwkv` and `dpr` are separate architectures and do not appear to fix the same underlying issue." + } + ] + }, + "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": 23, + "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": "788c14159f465e1d89c72ab5c5ff680b9654cbe795fb988d0e53c93e3bb67f0b", + "component_budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 3, + "serialized_chars": 11072, + "estimated_input_tokens": 2768, + "estimated_eval_tokens": 5792 + }, + "budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 3, + "serialized_chars": 11072, + "estimated_input_tokens": 2768, + "estimated_eval_tokens": 5792 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "All three PRs are separate model-integration changes that only overlap in generic registry/docs wiring; they do not appear to be duplicates or the same underlying code-path fix.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 45075 is the strongest standalone candidate: it is a complete, non-draft model addition with converter, model, image-processing, processing, and test coverage, whereas the others are distinct integrations.", + "best_issue_reason": null, + "best_pr_reason": "PR 45075 is the best overall representative because it is the most production-ready and self-contained of the three model additions.", + "soft_edge_verdicts": [ + { + "left": "pull_request:44662", + "right": "pull_request:45401", + "accept": false, + "reason": "Different models and use cases (PenguinVL vision-language vs Voxtral TTS). Shared auto/TOC files are generic integration boilerplate, not evidence of the same change." + }, + { + "left": "pull_request:44662", + "right": "pull_request:45075", + "accept": false, + "reason": "Different model integrations (PenguinVL vs DeepSeek-OCR-2). Overlap is limited to standard Transformers registry/docs files, so they are not the same underlying bug/change." + }, + { + "left": "pull_request:45075", + "right": "pull_request:45401", + "accept": false, + "reason": "DeepSeek-OCR-2 OCR support and Voxtral TTS support are unrelated modalities and code paths; the overlap is only in common model-registration files." + } + ] + }, + "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": 23, + "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": "dea6054d98e767949526b3bc34d9ea3286ebd7ae67c804cc3b2d92bd1ef09084", + "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": "Both PRs target kernel/module fusion, but they are not the same change: #44979 introduces a low-level Module Fusion API, while #45363 implements n-to-1 kernel fusion through KernelConfig and integrates it into model loading/runtime. They look related and potentially dependent, not duplicate.", + "confidence": 0.91, + "canonical_issue_reason": null, + "canonical_pr_reason": "#45363 is the better cluster representative because it is the more user-facing, end-to-end feature implementation and touches the main integration path for KernelConfig-based fusion.", + "best_issue_reason": null, + "best_pr_reason": "#45363 has the clearest concrete scope and broader impact; #44979 is a foundational API piece rather than the full feature implementation.", + "soft_edge_verdicts": [ + { + "left": "pull_request:44979", + "right": "pull_request:45363", + "accept": false, + "reason": "Related fusion work, but different layers of the stack: one adds the Module Fusion API, the other wires n-to-1 kernel fusion through KernelConfig and integrations. Not the same underlying change and not something that would 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": 20, + "component_count": 23, + "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": "13b67ffbb1eec714308edaae58d66fd0bfdf86eb975552fcd9cdbf66b846017b", + "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 address the same Mistral regex AttributeError in `_patch_mistral_regex` by changing the same pre-tokenizer access path. PR 45317 is the more complete version because it includes tests in addition to the code fix, so it is the best canonical PR.", + "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 test coverage, making it the more complete representation of the change.", + "best_issue_reason": null, + "best_pr_reason": "PR 45317 is the best overall PR candidate: same concrete bug fix as 45086, but with added tests and slightly broader implementation coverage.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45086", + "right": "pull_request:45317", + "accept": true, + "reason": "Same underlying bug and same code path: both replace `backend_tokenizer.pre_tokenizer` with `tokenizer.pre_tokenizer` in `_patch_mistral_regex`. The second PR appears to be a superset with tests, so these look mergeable into one 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": 21, + "component_count": 23, + "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": "44422564ac9e4ff89f16e3bb3734c0b3ca8b2263e88db1dc8e6f7f6236f2895c", + "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-family refactors, but they target different concrete changes: one rewires NemotronH to inherit GraniteMoeHybrid and adds conversion mapping, while the other splits the family into NemotronH Dense/Sparse variants with docs, auto-mapping, and tests. They are not the same bug or change set.", + "confidence": 0.82, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": "PR #45569 is the better cluster representative because it is broader and more self-contained around the NemotronH family split, covering model registration, docs, and tests; #45115 is a narrower internal refactor/conversion-mapping change.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45115", + "right": "pull_request:45569", + "accept": false, + "reason": "Overlap is only at the shared NemotronH base model files. #45115 focuses on GraniteMoeHybrid inheritance and checkpoint conversion renames, while #45569 introduces separate Dense/Sparse model variants and documentation changes. Different concrete changes, so not a duplicate pair." + } + ] + }, + "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": 23, + "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": "82d7423e3ef463a26b389d7162745ae0f1b4e4b8fcf6a2efcdba87d1a4003768", + "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 but not duplicates: #45333 adds heterogeneous per-layer configuration support, while #45332 builds on that with broader heterogeneous modeling/cache integration. They share utility files, but they target different layers of the feature stack.", + "confidence": 0.92, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #45332 is the better representative because it covers the larger end-to-end heterogeneous model support path, including cache/modeling changes, while #45333 is mainly the configuration foundation.", + "best_issue_reason": null, + "best_pr_reason": "#45332 is the strongest overall PR in the cluster since it subsumes config support and extends into concrete model/runtime behavior, making it the most complete implementation candidate.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45332", + "right": "pull_request:45333", + "accept": false, + "reason": "Related but not the same change: #45333 is config plumbing for per-layer config, whereas #45332 adds heterogeneous model/runtime support and cache handling on top. They share subsystem files but would not plausibly be merged as 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": 23, + "component_count": 23, + "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": "5a60dc715725ed1187461e6d2d4b7d532f26cc41f9d956e0e924e2e7301d71c4", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5306, + "estimated_input_tokens": 1327, + "estimated_eval_tokens": 2910 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5306, + "estimated_input_tokens": 1327, + "estimated_eval_tokens": 2910 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Two unrelated typing-focused PRs: one is a broad repo-wide type-checking sweep, the other adds specific typing helpers and annotations for modeling_utils/PEFT integration. They overlap on a couple of shared files but do not fix the same concrete bug/change.", + "confidence": 0.94, + "canonical_issue_reason": null, + "canonical_pr_reason": "45425 is the more focused, concrete typing change centered on modeling_utils and PEFT-related protocols; 45415 is a broad type-checking cleanup across many unrelated files and should not be treated as the same change.", + "best_issue_reason": null, + "best_pr_reason": "45425 is the best representative of the cluster because it has a narrowly scoped, coherent purpose and a single code-path theme, unlike the large cross-file typing sweep in 45415.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45415", + "right": "pull_request:45425", + "accept": false, + "reason": "Both are typing-related, but 45415 is a broad static-type cleanup across many modules while 45425 adds specific protocols and annotations for modeling_utils/PEFT. Shared filenames alone do not indicate the same underlying fix, 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 + } + ] +} diff --git a/snapshots/20260423T000027Z/analysis-runs/hybrid-model-20260423t000027z/manifest.json b/snapshots/20260423T000027Z/analysis-runs/hybrid-model-20260423t000027z/manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..d44c229a8a89489254012b9d7db53fcbfcaaee99 --- /dev/null +++ b/snapshots/20260423T000027Z/analysis-runs/hybrid-model-20260423t000027z/manifest.json @@ -0,0 +1,14 @@ +{ + "analysis_id": "hybrid-model-20260423t000027z", + "artifacts": { + "hybrid": "snapshots/20260423T000027Z/analysis-runs/hybrid-model-20260423t000027z/analysis-report-hybrid.json", + "hybrid_reviews": "snapshots/20260423T000027Z/analysis-runs/hybrid-model-20260423t000027z/analysis-report-hybrid.llm-reviews.json" + }, + "channel": "canonical", + "model": null, + "published_at": "2026-04-23T00:09:12Z", + "repo": "huggingface/transformers", + "schema_version": 1, + "snapshot_id": "20260423T000027Z", + "variant": "hybrid" +} diff --git a/snapshots/20260423T000027Z/manifest.json b/snapshots/20260423T000027Z/manifest.json index 4d23bdf0ef9862927b551b4067409d957af527fb..37c3f99e94eaf3fcb65ef321d08402c554796c38 100644 --- a/snapshots/20260423T000027Z/manifest.json +++ b/snapshots/20260423T000027Z/manifest.json @@ -32,6 +32,24 @@ "timeline_events": 340 }, "extracted_at": "2026-04-23T00:00:27Z", + "published_analysis": { + "canonical_analysis_id": "hybrid-model-20260423t000027z", + "runs": { + "hybrid-model-20260423t000027z": { + "analysis_id": "hybrid-model-20260423t000027z", + "artifacts": { + "hybrid": "snapshots/20260423T000027Z/analysis-runs/hybrid-model-20260423t000027z/analysis-report-hybrid.json", + "hybrid_reviews": "snapshots/20260423T000027Z/analysis-runs/hybrid-model-20260423t000027z/analysis-report-hybrid.llm-reviews.json" + }, + "channel": "canonical", + "manifest_path": "snapshots/20260423T000027Z/analysis-runs/hybrid-model-20260423t000027z/manifest.json", + "model": null, + "published_at": "2026-04-23T00:09:12Z", + "variant": "hybrid" + } + }, + "schema_version": 1 + }, "repo": "huggingface/transformers", "snapshot_id": "20260423T000027Z", "watermark": {