diff --git a/analysis/current/analysis-report-hybrid.json b/analysis/current/analysis-report-hybrid.json index cc7dbab16ddf1d95dce61020331cff9629049acc..7b11d5e1bde566bd596460f8552e0f9b0b1a16e7 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": "20260523T180021Z", - "generated_at": "2026-05-23T18:07:26Z", + "snapshot_id": "20260524T000134Z", + "generated_at": "2026-05-24T00:08:54Z", "evidence_quality": "full", "llm_enrichment": true, "meta_bugs": [ @@ -1708,81 +1708,6 @@ } ] }, - { - "cluster_id": "cluster-43656-2", - "summary": "Cluster of 2 related pull requests linked by soft_similarity.", - "status": "open", - "confidence": 0.5, - "canonical_issue_number": null, - "canonical_pr_number": 43656, - "issue_numbers": [], - "pr_numbers": [ - 43656, - 43842 - ], - "evidence_types": [ - "soft_similarity" - ], - "pr_comparisons": [ - { - "left_pr_number": 43656, - "right_pr_number": 43842, - "code_similarity": 0.405, - "size_similarity": 0.5, - "file_overlap": 1.0, - "area_overlap": 0.0, - "patch_similarity": 0.036, - "shared_filenames": [ - "src/transformers/cli/serve.py" - ], - "shared_file_areas": [ - { - "filename": "src/transformers/cli/serve.py", - "left_ranges": [ - [ - 11, - 18 - ], - [ - 30, - 36 - ], - [ - 315, - 323 - ], - [ - 665, - 671 - ], - [ - 931, - 937 - ], - [ - 1843, - 1849 - ], - [ - 1868, - 1874 - ] - ], - "right_ranges": [ - [ - 54, - 61 - ], - [ - 587, - 593 - ] - ] - } - ] - } - ] - }, { "cluster_id": "cluster-46100-2", "summary": "Cluster of 2 related pull requests linked by soft_similarity.", @@ -2135,15 +2060,6 @@ "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-43656-2", - "canonical_pr_number": 43656, - "duplicate_pr_numbers": [ - 43842 - ], - "target_issue_number": null, - "reason": "PRs in cluster-43656-2 are treated as duplicates because they share soft_similarity evidence." - }, { "cluster_id": "cluster-43698-3", "canonical_pr_number": 43779, diff --git a/analysis/current/analysis-report-hybrid.llm-reviews.json b/analysis/current/analysis-report-hybrid.llm-reviews.json index 87f86cab7ee3744f73ed14aba033596dc7067515..c3c683f067f876f556e4de43aa4e093c27a022e5 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": "20260523T180021Z", - "generated_at": "2026-05-23T18:07:26Z", + "snapshot_id": "20260524T000134Z", + "generated_at": "2026-05-24T00:08:54Z", "model": "gpt-5.4-mini?service_tier=flex", "reviews": [ { @@ -10,7 +10,7 @@ "component_index": 1, "component_count": 17, "review_unit_index": 1, - "review_unit_count": 50, + "review_unit_count": 49, "status": "reviewed", "reason": null, "source": "llm", @@ -93,8 +93,8 @@ "issue:34634|issue:35545", "issue:30277|issue:43638", "issue:43596|issue:43638", - "issue:43975|issue:44457", "issue:43039|issue:45161", + "issue:43975|issue:44457", "issue:44292|issue:44457", "issue:34928|issue:44805", "issue:24643|issue:30277", @@ -108,8 +108,8 @@ "issue:43425|issue:44292", "issue:34928|issue:41720", "issue:43828|issue:45161", - "issue:44918|issue:45161", "issue:43716|issue:45685", + "issue:44918|issue:45161", "issue:43828|issue:44292", "issue:43541|issue:43716", "issue:44292|issue:44918", @@ -138,367 +138,331 @@ "issue:43866|issue:44863", "issue:44292|issue:46015" ], - "prepared_review_unit_hash": "c4f8ff4d3a668d957995d8b04d26dc1bc3cc62f6b253e1c173c20ab24c50c774", + "prepared_review_unit_hash": "ff874a85474cdc886c0f715b028ee4b8a25dcae056743f81e04870d26bc5b7d1", "component_budget": { "node_count": 426, "item_count": 426, - "soft_pair_count": 1572, - "serialized_chars": 814130, - "estimated_input_tokens": 203533, - "estimated_eval_tokens": 407322 + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 69, - "serialized_chars": 48400, - "estimated_input_tokens": 12100, - "estimated_eval_tokens": 24456 + "serialized_chars": 48406, + "estimated_input_tokens": 12102, + "estimated_eval_tokens": 24460 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Mostly heterogeneous cluster. The only solid duplicate pair is the Bert + DeepSpeed ZeRO-3 index-out-of-bounds reports (43596/43638); the Qwen3_5Moe rotary-embedding pair is related but not strong enough to merge conservatively. Everything else looks like broad false positives across unrelated models or subsystems.", - "confidence": 0.69, - "canonical_issue_reason": "Issue 43596 is the cleanest representative of the Bert + DeepSpeed ZeRO-3 index-out-of-bounds duplicate pair, with the broader wording making it the best canonical issue.", + "summary": "The cluster is mostly a mix of unrelated issue reports, with only a couple of true duplicate-style overlaps. The clearest duplicates are the DeepSpeed ZeRO-3 BERT index error pair and the AutoTokenizer wrong-token-ID regression pair.", + "confidence": 0.86, + "canonical_issue_reason": "issue:45920 is the broadest and most representative report of the AutoTokenizer v5 regression, covering multiple affected model families rather than a single model family.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 43596 is the best overall issue representative because it captures the clearest concrete duplicate in the set and has the broadest, least noisy wording.", + "best_issue_reason": "issue:45920 best captures the underlying tokenizer regression and is open, broader, and more suitable as the main tracking issue.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44322", "right": "issue:45464", "accept": false, - "reason": "Different failure modes: Qwen3_5 config attribute error vs chat/completions streaming API failure." + "reason": "Different failures: missing config attribute vs streaming chat/completions API behavior." }, { "left": "issue:45237", "right": "issue:45685", "accept": false, - "reason": "Both involve GPU/backend problems, but the concrete bugs are unrelated." + "reason": "Different models and backends; one is AMD GPU support for GPT-OSS, the other is an MPS histogram kernel issue in MoE code." }, { "left": "issue:43716", "right": "issue:45237", "accept": false, - "reason": "Mistral image dtype mismatch is unrelated to GPT-OSS AMD GPU support." + "reason": "Different code paths: image preprocessor dtype mismatch vs GPT-OSS running on AMD GPUs." }, { "left": "issue:43828", "right": "issue:45237", "accept": false, - "reason": "Different concrete problems: autocast dtype mismatch vs AMD GPU runtime failure." + "reason": "Both are model/backend errors, but the concrete bugs are different dtype/autocast handling vs AMD GPU execution." }, { "left": "issue:45237", "right": "issue:45507", "accept": false, - "reason": "Different model families and symptoms; no shared code-path evidence." + "reason": "Different underlying bug: AMD GPU runtime failure vs invalid method call in GraniteMoEHybrid." }, { "left": "issue:44292", "right": "issue:45237", "accept": false, - "reason": "Both are Qwen/GPT-OSS runtime issues, but not the same bug or fix." + "reason": "Different model/runtime failures; NVFP4 execution error is not the same as AMD GPU incompatibility." }, { "left": "issue:30277", "right": "issue:35545", "accept": false, - "reason": "Deepspeed collective mismatch and ONNX export error are unrelated." + "reason": "DeepSpeed collective mismatch and ONNX export failure are unrelated bugs." }, { "left": "issue:35545", "right": "issue:42915", "accept": false, - "reason": "ModernBERT ONNX export and Qwen3Moe FP8 failure are different issues." + "reason": "Different model/export path and error type; ONNX export is unrelated to FineGrainedFP8 config failure." }, { "left": "issue:35545", "right": "issue:41720", "accept": false, - "reason": "ONNX export and Qwen3 auto device mapping are different code paths." + "reason": "Different subsystems and symptoms: export error vs device-mapping cuda assert." }, { "left": "issue:45901", "right": "issue:46015", "accept": false, - "reason": "Table QA crash and outdated RAG examples are unrelated." + "reason": "One is a task crash, the other is stale docs; same broad area but not the same bug." }, { "left": "issue:44509", "right": "issue:44512", "accept": false, - "reason": "Both are docs regressions after v5, but they target different removed features and are not the same change." + "reason": "Both are v5 docs cleanups, but they concern different removed doc references and should stay separate." }, { "left": "issue:35545", "right": "issue:36010", "accept": false, - "reason": "ONNX export and GenerationMixin import failure are unrelated." + "reason": "Different problems: ONNX export versus missing GenerationMixin import." }, { "left": "issue:44805", "right": "issue:45823", "accept": false, - "reason": "Different symptoms and different model families; no same underlying bug evidence." + "reason": "Different failures: mask/tensor shape mismatch vs Gemma4 device placement mismatch." }, { "left": "issue:44918", "right": "issue:46165", "accept": false, - "reason": "TRL SFT unpacking failure and training-resume/trust_remote_code issue are distinct." + "reason": "Different training failures with different triggers; not the same code-path bug." }, { "left": "issue:43638", "right": "issue:44805", "accept": false, - "reason": "Different tensor-shape failures in different contexts." - }, - { - "left": "issue:43638", - "right": "issue:46165", - "accept": false, - "reason": "ZeRO-3 Bert index error and training-resume issue are unrelated." + "reason": "Both are IndexError reports, but one is DeepSpeed ZeRO-3 BERT init and the other is a mask shape mismatch." }, { "left": "issue:43596", "right": "issue:44805", "accept": false, - "reason": "Both mention index/shape errors, but they are different bugs in different paths." + "reason": "Same superficial error class, but different underlying failure modes and code paths." }, { "left": "issue:44805", "right": "issue:45161", "accept": false, - "reason": "Mask shape mismatch and GPT-OSS MoE tensor-parallel issue are unrelated." + "reason": "Mask-shape mismatch is unrelated to GPT-OSS tensor-parallel not working." }, { "left": "issue:44560", "right": "issue:46165", "accept": false, - "reason": "Video StopIteration and model-training resume are unrelated." + "reason": "Different model/runtime issue from a training resume/trust-remote-code problem." }, { "left": "issue:43596", "right": "issue:46165", "accept": false, - "reason": "Different bug categories with no shared concrete code path." + "reason": "DeepSpeed ZeRO-3 BERT index error is unrelated to resume-training with remote code." }, { "left": "issue:44805", "right": "issue:44918", "accept": false, - "reason": "Different failure modes in different trainer/input handling paths." + "reason": "Different symptoms and paths: masking error vs unpacking Qwen3.5 embeddings in TRL SFT." }, { "left": "issue:43572", "right": "issue:46165", "accept": false, - "reason": "Config missing pad_token_idx is unrelated to training-resume behavior." + "reason": "Config missing pad_token_idx is unrelated to training resume behavior." }, { "left": "issue:34634", "right": "issue:35545", "accept": false, - "reason": "BarkProcessor voice_preset and ModernBERT ONNX export are unrelated." + "reason": "Processor voice preset bug is unrelated to ModernBERT ONNX export." }, { "left": "issue:30277", "right": "issue:43638", "accept": false, - "reason": "Different models and failure modes; only both mention DeepSpeed broadly." + "reason": "Both involve DeepSpeed, but one is collective mismatch and the other is BERT zero-sized index on load." }, { "left": "issue:43596", "right": "issue:43638", "accept": true, - "reason": "Same concrete DeepSpeed ZeRO-3 Bert index-out-of-bounds failure; the titles describe the same underlying bug with slightly different wording." + "reason": "Same DeepSpeed ZeRO-3 BERT index-out-of-bounds failure; the titles describe the same underlying loading/training bug." }, { - "left": "issue:43975", - "right": "issue:44457", + "left": "issue:43039", + "right": "issue:45161", "accept": false, - "reason": "DeepSeek detokenization and LoRA merge/save mismatch are unrelated." + "reason": "Different issues in different parts of the stack; Liger cross-entropy call is not the same as GPT-OSS TP failure." }, { - "left": "issue:43039", - "right": "issue:45161", + "left": "issue:43975", + "right": "issue:44457", "accept": false, - "reason": "Cross-entropy routing under Liger Kernel and GPT-OSS tensor-parallel failure are different issues." + "reason": "Different bugs: detokenization regression vs LoRA merge/save/reload mismatch." }, { "left": "issue:44292", "right": "issue:44457", "accept": false, - "reason": "Qwen-3-8B NVFP4 runtime error and LoRA merge/save inconsistency are unrelated." + "reason": "Different failures and models; NVFP4 runtime error is not a LoRA merge regression." }, { "left": "issue:34928", "right": "issue:44805", "accept": false, - "reason": "Activation checkpointing/FSDP recomputation issue is different from the mask-shape index error." + "reason": "Activation checkpointing/FSDP recompute mismatch is unrelated to the mask-shape IndexError." }, { "left": "issue:24643", "right": "issue:30277", "accept": false, - "reason": "Both involve DeepSpeed, but the concrete RuntimeErrors differ and do not look like one bug." + "reason": "Both mention DeepSpeed, but the concrete failures are different runtime errors." }, { "left": "issue:45859", "right": "issue:45902", "accept": false, - "reason": "Same Qwen3_5Moe module, but the packet does not strongly establish one exact underlying bug/fix; the manifestations differ enough to stay conservative." + "reason": "Same Qwen3.5 MoE embedding area, but the triggers differ (CPU offload vs meta-to-empty materialization), so this is not clearly the same bug." }, { "left": "issue:43638", "right": "issue:45161", "accept": false, - "reason": "Bert ZeRO-3 index error and GPT-OSS MoE TP issue are unrelated." + "reason": "BERT ZeRO-3 loading error and GPT-OSS tensor-parallel failure are unrelated." }, { "left": "issue:34928", "right": "issue:36331", "accept": false, - "reason": "Different trainer-related bugs: activation checkpointing mismatch vs compute_loss signature change." - }, - { - "left": "issue:43596", - "right": "issue:45161", - "accept": false, - "reason": "Different models and failure signatures; no same fix implied." - }, - { - "left": "issue:43716", - "right": "issue:45507", - "accept": false, - "reason": "Mistral-3 dtype mismatch and GraniteMoEHybrid invalid method call are unrelated." - }, - { - "left": "issue:39290", - "right": "issue:45081", - "accept": false, - "reason": "Gemma3 config attribute error with vLLM and Mistral regex patch crash are different bugs." - }, - { - "left": "issue:45823", - "right": "issue:45874", - "accept": false, - "reason": "Device mismatch under auto mapping and expensive reverse embedding lookup are different issues." - }, - { - "left": "issue:43425", - "right": "issue:44292", - "accept": false, - "reason": "Torch version incompatibility and Qwen-3-8B runtime error are not the same bug." + "reason": "Activation checkpointing/FSDP mismatch is unrelated to CustomTrainer compute_loss signature breakage." }, { - "left": "issue:34928", - "right": "issue:41720", + "left": "issue:43039", + "right": "issue:43828", "accept": false, - "reason": "FSDP activation-checkpointing mismatch and Qwen3 auto device mapping assert are unrelated." + "reason": "Different model/runtime bugs: Liger Kernel cross-entropy invocation vs autocast dtype mismatch." }, { - "left": "issue:43827", - "right": "issue:44512", + "left": "issue:30277", + "right": "issue:36331", "accept": false, - "reason": "Both are docs regressions, but they cover different removed commands/features and are not the same change." + "reason": "DeepSpeed collective mismatch and trainer compute_loss signature mismatch are unrelated." }, { "left": "issue:45237", "right": "issue:46015", "accept": false, - "reason": "AMD GPU runtime failure and outdated RAG examples are unrelated." + "reason": "Model runtime failure vs docs cleanup; not the same bug or change." }, { "left": "issue:34928", "right": "issue:36010", "accept": false, - "reason": "Different training/checkpointing bug versus import error." + "reason": "FSDP activation-checkpointing error is unrelated to GenerationMixin import failure." }, { "left": "issue:43039", "right": "issue:44805", "accept": false, - "reason": "Liger Kernel cross-entropy issue and mask-shape error are unrelated." + "reason": "Different underlying failures: Liger kernel path vs tensor mask shape mismatch." }, { "left": "issue:41762", "right": "issue:43596", "accept": false, - "reason": "Both mention DeepSpeed, but Gemma3 ZeRO-3 loading and Bert ZeRO-3 index errors are different concrete bugs." + "reason": "Both are DeepSpeed-related, but Gemma3 ZeRO-3 loading and BERT ZeRO-3 index errors are different bugs." }, { "left": "issue:39290", "right": "issue:41720", "accept": false, - "reason": "Different model-specific errors with no shared code-path evidence." + "reason": "Different model/backend issues: Gemma3/vLLM attribute error vs Qwen3 auto device map cuda assert." }, { "left": "issue:43901", "right": "issue:44512", "accept": false, - "reason": "TextClassificationPipeline docs and transformers run command docs are separate documentation issues." + "reason": "Different docs issues referencing different removed API surface." }, { "left": "issue:45237", "right": "issue:45901", "accept": false, - "reason": "GPT-OSS AMD GPU failure and table-question-answering crash are unrelated." + "reason": "GPT-OSS AMD runtime failure is unrelated to table-question-answering crashes." }, { "left": "issue:44841", "right": "issue:45823", "accept": false, - "reason": "Processor failure for Voxtral and Gemma4 device mismatch are different issues." + "reason": "Different processor/model failures with no shared concrete bug." }, { "left": "issue:43039", "right": "issue:44560", "accept": false, - "reason": "Liger Kernel cross-entropy issue and Qwen3-vl video StopIteration are unrelated." + "reason": "Liger Kernel cross-entropy and Qwen3-vl video stop-iteration are unrelated." }, { "left": "issue:43975", "right": "issue:46015", "accept": false, - "reason": "DeepSeek detokenization bug and RAG docs staleness are unrelated." + "reason": "Tokenizer detokenization regression is unrelated to outdated RAG examples." }, { "left": "issue:45812", "right": "issue:45920", - "accept": false, - "reason": "Both are tokenizer regressions, but the affected model sets are broad enough that they may be separate AutoTokenizer bugs rather than one concrete duplicate." + "accept": true, + "reason": "Same AutoTokenizer wrong-token-ID regression in v5; 45920 simply broadens the affected model list beyond Granite." }, { "left": "issue:44863", "right": "issue:46015", "accept": false, - "reason": "NemotronH checkpoint loading and outdated RAG docs are unrelated." + "reason": "NemotronH checkpoint loading failure is unrelated to docs/example updates." }, { "left": "issue:43866", "right": "issue:44863", "accept": false, - "reason": "Checkpoint corruption and model-loading implementation failure are different problems." + "reason": "Ovis2 checkpoint corruption and NemotronH checkpoint loading are different problems." }, { "left": "issue:44292", "right": "issue:46015", "accept": false, - "reason": "Qwen-3-8B NVFP4 runtime error and outdated RAG docs are unrelated." + "reason": "NVFP4 runtime error is unrelated to outdated RAG docs." } ] }, "evaluator_result": { - "accept": false, - "feedback": "The packet does support the Bert/DeepSpeed ZeRO-3 duplicate pair, but the analyst overstates the uniqueness of that duplicate: the packet also has a second accepted soft edge for the Qwen3_5Moe rotary-embedding issues (45859/45902). Because the reasoning says 43596 is the representative of the \"only clearly duplicated concrete bug in the set,\" the summary is not fully grounded. The Qwen3_5Moe acceptance also feels a bit stronger than the packet alone guarantees, so I would keep this conservative and reject." + "accept": true, + "feedback": "The summary is grounded in the packet and stays conservative. The two accepted overlaps are well-supported by the titles and descriptions: the DeepSpeed ZeRO-3 BERT index errors (43596/43638) and the AutoTokenizer wrong-token-ID regression (45812/45920). The rest of the pairwise rejections are cautious and consistent with the packet." }, "evaluator_used": true, - "retried": true, + "retried": false, "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null @@ -508,7 +472,7 @@ "component_index": 1, "component_count": 17, "review_unit_index": 2, - "review_unit_count": 50, + "review_unit_count": 49, "status": "reviewed", "reason": null, "source": "llm", @@ -606,268 +570,268 @@ "issue:43645|issue:45399", "issue:43404|issue:45005" ], - "prepared_review_unit_hash": "ba3a2331265e47a9f0093b301b79dfe337850868cc5887814ef6b129e60c429a", + "prepared_review_unit_hash": "0c9974d3f32ac16fb6d7cd08ef94fa643f3dc5220aded1163b773100684cc3ae", "component_budget": { "node_count": 426, "item_count": 426, - "soft_pair_count": 1572, - "serialized_chars": 814130, - "estimated_input_tokens": 203533, - "estimated_eval_tokens": 407322 + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 39, - "serialized_chars": 39131, - "estimated_input_tokens": 9783, - "estimated_eval_tokens": 19822 + "serialized_chars": 39128, + "estimated_input_tokens": 9782, + "estimated_eval_tokens": 19820 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This set is a heterogeneous grab bag of unrelated Transformers/vLLM regressions (model loading, quantization, dtype/device mismatches, chat/task failures, config issues). None of the proposed soft pairs look like safe duplicates, so the cluster should not be collapsed further.", - "confidence": 0.88, - "canonical_issue_reason": "issue:45698 is the most actionable representative here: a concrete, reproducible custom-module loading regression after save_pretrained/from_pretrained, with active discussion and inbound references.", + "summary": "This cluster is mostly a grab bag of distinct Transformers regressions and model-specific failures rather than true duplicates. The best umbrella issue is the broad remote_code regression report, but I did not find any soft pair that looks like the same concrete bug.", + "confidence": 0.63, + "canonical_issue_reason": "Issue 45020 is the broadest and most central report here: it describes recent-version breakage for models using `remote_code`, which is the closest umbrella for several compatibility-style regressions in the set.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45698 is a good cluster anchor because it is specific, user-facing, and still under discussion; among these mostly unrelated issues it is the clearest representative bug report.", + "best_issue_reason": "Issue 45020 is the best global representative because it captures a class of recent Transformers compatibility breakages affecting custom/model-loading paths, not just one model instance.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:39290", "right": "issue:43531", "accept": false, - "reason": "Different models and failures: Gemma3TextConfig missing attribute vs Qwen3-MoE sliding_window behavior." + "reason": "Both mention `sliding_window`, but they are different models with different failure modes; not the same concrete bug." }, { "left": "issue:43854", "right": "issue:43866", "accept": false, - "reason": "One is a model-loading test failure, the other is a corrupted checkpoint claim; different root causes." + "reason": "Unrelated model-loading problem vs checkpoint corruption; no shared bug path." }, { "left": "issue:43792", "right": "issue:45901", "accept": false, - "reason": "Whisper runtime failure and table QA crash are unrelated tasks/code paths." + "reason": "Different tasks and model families; both are crashes, but not the same underlying issue." }, { "left": "issue:41093", "right": "issue:43638", "accept": false, - "reason": "Both are index-related, but one is a mask shape mismatch while the other is a zero-sized-tensor index error under ZeRO-3." + "reason": "Both are index-related errors under deepspeed, but one is a mask shape mismatch and the other is a zero3/Bert init failure." }, { "left": "issue:43425", "right": "issue:45070", "accept": false, - "reason": "Torch-version incompatibility and a pydantic PretrainedConfig regression are different bugs." + "reason": "Generic Torch version incompatibility is not the same as the specific pydantic `PretrainedConfig` regression." }, { "left": "issue:43596", "right": "issue:45800", "accept": false, - "reason": "Empty-index error during ZeRO-3/Bert loading is unrelated to torch 2.4.1 vs transformers 5.8.0 incompatibility." + "reason": "Different root causes: zero3/Bert loading index error vs a torch/transformers version mismatch." }, { "left": "issue:43975", "right": "issue:45901", "accept": false, - "reason": "Detokenization bug and table-question-answering crash are different features and failure modes." + "reason": "Detokenization bug and table QA crash are unrelated." }, { "left": "issue:43792", "right": "issue:45070", "accept": false, - "reason": "Whisper runtime failure does not match the pydantic field regression." + "reason": "Completely different issue classes: Whisper runtime failure vs config/model-schema regression." }, { "left": "issue:45362", "right": "issue:45464", "accept": false, - "reason": "Both involve Qwen3.5 chat, but one is a crash in transformers chat while the other is a streaming chat/completions API failure on a different model size." + "reason": "Same model family, but one is a chat crash and the other is a streaming chat/completions failure; too little evidence of one code-path bug." }, { "left": "issue:44292", "right": "issue:45901", "accept": false, - "reason": "NVFP4 model loading error and table QA crash are unrelated." + "reason": "`init_empty_weights` argument handling is unrelated to table-question-answering crashes." }, { "left": "issue:39290", "right": "issue:45823", "accept": false, - "reason": "Gemma3/vLLM sliding_window error is unrelated to Gemma4 device-mismatch during forward." + "reason": "Different models and different runtime errors; no shared bug." }, { "left": "issue:45464", "right": "issue:45685", "accept": false, - "reason": "Chat streaming API failure and an MPS histogram/int bug in MoE code are different code paths." + "reason": "Streaming chat API failure is unrelated to an MPS histogram-on-Int error." }, { "left": "issue:43404", "right": "issue:43866", "accept": false, - "reason": "Tied lm_head weights issue and a corrupted checkpoint report are not the same bug." + "reason": "Untied weights in Mistral3 is unrelated to a corrupted Ovis2 checkpoint." }, { "left": "issue:43541", "right": "issue:45685", "accept": false, - "reason": "Mixtral torch-dynamo grouped_mm tracing error is unrelated to the MPS histogram Int failure." + "reason": "Torch dynamo grouped_mm failure and MPS histogram failure are unrelated." }, { "left": "issue:43541", "right": "issue:43828", "accept": false, - "reason": "Both mention runtime/dtype problems, but the affected model paths and concrete errors differ." + "reason": "Both are dtype-related in MoE contexts, but the model paths and failure mechanisms differ." }, { "left": "issue:44841", "right": "issue:45084", "accept": false, - "reason": "Processor failure and non-template-node compilation error are unrelated." + "reason": "Processor load failure and template-node compilation failure are unrelated." }, { "left": "issue:36010", "right": "issue:39290", "accept": false, - "reason": "ImportError for GenerationMixin is unrelated to the Gemma3/vLLM sliding_window regression." + "reason": "ImportError for `GenerationMixin` and a Gemma3 sliding-window attribute error are unrelated." }, { "left": "issue:45800", "right": "issue:45823", "accept": false, - "reason": "Torch/transformers version incompatibility and Gemma4 device mismatch are different issues." + "reason": "Torch/Transformers version incompatibility is not the same bug as Gemma4 device-map mismatch." }, { "left": "issue:41093", "right": "issue:41762", "accept": false, - "reason": "Both are index errors, but one is a mask-length mismatch and the other is an out-of-bounds access when loading Gemma3 with ZeRO-3." + "reason": "Both involve zero3 index errors, but the specific shapes and loading paths differ enough to treat them as separate bugs." }, { "left": "issue:43572", "right": "issue:45070", "accept": false, - "reason": "Missing pad_token_idx in StableLmConfig is unrelated to the pydantic PretrainedConfig field breakage." + "reason": "Both are config-related v5 regressions, but they concern different fields and different breakage modes." }, { "left": "issue:43296", "right": "issue:43366", "accept": false, - "reason": "vLLM model-load failure and a request for GGUF support are not the same concrete bug." + "reason": "vLLM loading failure and GGUF architecture support are different problems." }, { "left": "issue:43541", "right": "issue:45161", "accept": false, - "reason": "Mixtral torch-dynamo grouped_mm error and GPT-OSS tensor-parallel failure are different models and problems." + "reason": "Both are MoE-related, but one is grouped_mm tracing and the other is tensor-parallel-only GPT-OSS behavior." }, { "left": "issue:44387", "right": "issue:45005", "accept": false, - "reason": "Int4 quantization OOM and tied-weights translation regression are unrelated." + "reason": "Quantization OOM and tied-weights regressions are different issues." }, { "left": "issue:43828", "right": "issue:45464", "accept": false, - "reason": "Autocast dtype mismatch in Phi-tiny-MoE is unrelated to Qwen3.5 streaming API failure." + "reason": "Autocast dtype mismatch is unrelated to a streaming chat API failure." }, { "left": "issue:44291", "right": "issue:45070", "accept": false, - "reason": "init_empty_weights argument mismatch and pydantic PretrainedConfig breakage are different regressions." + "reason": "Unexpected `_is_hf_initialized` argument during init is a different path than the pydantic config-field regression." }, { "left": "issue:43366", "right": "issue:45464", "accept": false, - "reason": "GGUF GPT-OSS support request and Qwen3.5 streaming failure are unrelated." + "reason": "GGUF gpt-oss support and Qwen streaming inference are unrelated." }, { "left": "issue:45020", "right": "issue:45800", "accept": false, - "reason": "Remote_code breakage across recent versions is broader, but not the same bug as the torch/transformers compatibility report." + "reason": "`remote_code` breakage is broader, but this pair does not point to the same specific bug as the torch/Transformers version incompatibility report." }, { "left": "issue:43531", "right": "issue:45070", "accept": false, - "reason": "Sliding_window handling in Qwen3-MoE and a pydantic field regression do not match." + "reason": "Sliding-window handling and pydantic config regression are distinct." }, { "left": "issue:43541", "right": "issue:44589", "accept": false, - "reason": "Grouped_mm tracing failure and missing Float8 storage type are different runtime errors." + "reason": "Grouped_mm tracing failure and missing Float8 storage object are unrelated." }, { "left": "issue:41093", "right": "issue:43596", "accept": false, - "reason": "Both are index errors, but one is a mask shape mismatch and the other is a zero-sized tensor access in ZeRO-3 Bert loading." + "reason": "Both are index errors with zero3, but they occur in different contexts and likely different code paths." }, { "left": "issue:43054", "right": "issue:43994", "accept": false, - "reason": "Siglip2 embedding quality regression and nonsensical results for a different model are not the same bug." + "reason": "Same model family, but one is about poorer embeddings and the other about nonsensical pipeline output; not enough to merge as one bug." }, { "left": "issue:45542", "right": "issue:45800", "accept": false, - "reason": "TensorBoard-without-TensorFlow backend error is unrelated to torch/transformers incompatibility." + "reason": "TensorBoard/TensorFlow backend setup issue is unrelated to torch version incompatibility." }, { "left": "issue:45759", "right": "issue:45902", "accept": false, - "reason": "Composite-config unwrap issue and uninitialized rotary embedding memory are different code paths." + "reason": "Config unwrapping for composite Qwen models and uninitialized rotary-embedding memory are different bugs." }, { "left": "issue:39186", "right": "issue:44918", "accept": false, - "reason": "FSDP weight-dimensionality error and Qwen3.5 embedding unpacking failure are unrelated." + "reason": "FSDP 2-D weight error and Qwen embedding unpacking in TRL are unrelated." }, { "left": "issue:43716", "right": "issue:44918", "accept": false, - "reason": "Mistral-3 image/text dtype mismatch and Qwen3.5 embedding unpacking are different issues." + "reason": "Mistral3 image-preprocessor dtype mismatch and Qwen3.5 embedding handling are different failures." }, { "left": "issue:40990", "right": "issue:44960", "accept": false, - "reason": "High perplexity on gpt-oss-20b and a generic GLM5 issue do not describe the same bug." + "reason": "Perplexity regression on GPT-OSS and GLM5 are unrelated." }, { "left": "issue:45632", "right": "issue:45698", "accept": false, - "reason": "Both touch custom-code loading, but one is a cache-path collision while the other is loading the wrong module after save_pretrained." + "reason": "Both involve custom-module loading, but one is a cache-path collision and the other is loading the wrong module after save/load; not clearly the same bug." }, { "left": "issue:43645", "right": "issue:45399", "accept": false, - "reason": "Notebook custom-model initialization and flash-attn fallback blocking are unrelated." + "reason": "Custom models in notebooks and flash-attn fallback gating are unrelated." }, { "left": "issue:43404", "right": "issue:45005", "accept": false, - "reason": "Both mention tied weights, but they affect different model families and different concrete loading paths." + "reason": "Both mention tied weights, but they target different model classes and APIs; not enough to treat as one concrete bug." } ] }, @@ -883,7 +847,7 @@ "component_index": 1, "component_count": 17, "review_unit_index": 3, - "review_unit_count": 50, + "review_unit_count": 49, "status": "reviewed", "reason": null, "source": "llm", @@ -941,8 +905,8 @@ "issue:45902" ], "soft_pairs": [ - "issue:44530|issue:45005", "issue:42915|issue:45399", + "issue:44530|issue:45005", "issue:43873|issue:45005", "issue:44863|issue:45005", "issue:44360|issue:44512", @@ -956,8 +920,8 @@ "issue:43828|issue:44928", "issue:41628|issue:45800", "issue:36331|issue:41762", - "issue:43950|issue:45902", "issue:43872|issue:44589", + "issue:43950|issue:45902", "issue:43866|issue:44403", "issue:43054|issue:45701", "issue:41553|issue:43883", @@ -976,244 +940,244 @@ "issue:43881|issue:43883", "issue:43782|issue:44960" ], - "prepared_review_unit_hash": "0e495c9627ec08d60b1db3fdb6b231f2aa07b1c58bc241c3e21fe3a7e3675405", + "prepared_review_unit_hash": "ad5fe355f962ae403aa08d6e61eb3f300e8d32782e5671b79981a29f12c07c01", "component_budget": { "node_count": 426, "item_count": 426, - "soft_pair_count": 1572, - "serialized_chars": 814130, - "estimated_input_tokens": 203533, - "estimated_eval_tokens": 407322 + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 34, - "serialized_chars": 37513, - "estimated_input_tokens": 9379, - "estimated_eval_tokens": 19014 + "serialized_chars": 37510, + "estimated_input_tokens": 9378, + "estimated_eval_tokens": 19012 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Most candidates are broad semantic matches, not true duplicates. The only clear duplicate cluster is the Transformers 5.x non-persistent buffer corruption regression (43950/44534, with 45902 as a specific manifestation). A second plausible duplicate pair is the SigLIP2 implementation discrepancy (43054/43493), but the rest are too vague or touch different concrete bugs.", - "confidence": 0.68, - "canonical_issue_reason": "43950 is the cleanest canonical issue: it describes the core Transformers 5.x non-persistent buffer corruption regression, and it has the strongest duplicate-style evidence among the candidates.", + "summary": "This cluster is mostly heterogeneous. The only clear duplicate subcluster is the Transformers v5 non-persistent-buffer corruption regression; the rest are separate loading/quantization/model-specific bugs.", + "confidence": 0.91, + "canonical_issue_reason": "Issue 43950 is the best representative of the one clear duplicate family here: it states the non-persistent-buffer corruption regression directly and broadly enough to cover the closely related reports.", "canonical_pr_reason": null, - "best_issue_reason": "43950 is the best overall issue to represent the cluster because it is specific, technical, and has the clearest same-bug follow-ups (44534 and 45902).", + "best_issue_reason": "43950 is the strongest overall issue candidate because it is the clearest, most general description of the recurring buffer-corruption bug. That said, the wider cluster is not a single coherent duplicate set.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:44530", - "right": "issue:45005", + "left": "issue:42915", + "right": "issue:45399", "accept": false, - "reason": "Both mention model behavior in v5, but they are different bugs: Qwen3.5 cache/group-type failure vs translation-model tied-weights handling." + "reason": "Different failure modes: FP8 config failure on Qwen3Moe vs flash-attn fallback gating when fa2 is absent." }, { - "left": "issue:42915", - "right": "issue:45399", + "left": "issue:44530", + "right": "issue:45005", "accept": false, - "reason": "Different failure modes and code paths: FineGrainedFP8Config load failure vs flash-attn2 fallback gating." + "reason": "Both involve model/runtime behavior in v5, but one is a Qwen3.5 cache crash and the other is a translation-model tied-weights issue; not the same bug." }, { "left": "issue:43873", "right": "issue:45005", "accept": false, - "reason": "Quantized offloading behavior is unrelated to tied-weights issues." + "reason": "Quantization/offloading behavior is too broad; these report different concrete problems and code paths." }, { "left": "issue:44863", "right": "issue:45005", "accept": false, - "reason": "Checkpoint loading for NemotronH is a different bug than tied-weights handling on translation models." + "reason": "NemotronH checkpoint loading and tied-weights handling are unrelated failures." }, { "left": "issue:44360", "right": "issue:44512", "accept": false, - "reason": "ReLU in DSA indexer and a docs typo are unrelated." + "reason": "A DSA indexer ReLU discussion is unrelated to a documentation typo about the removed run command." }, { "left": "issue:44918", "right": "issue:45507", "accept": false, - "reason": "Qwen3.5 embedding unpacking with TRL and GraniteMoEHybrid invalid method are different concrete issues." + "reason": "Different models and subsystems: Qwen3.5 SFT embedding unpacking vs GraniteMoEHybrid invalid method calls." }, { "left": "issue:44492", "right": "issue:44512", "accept": false, - "reason": "A cache-strategy typo is unrelated to removed run-command documentation." + "reason": "One is a cache-strategy typo, the other is stale docs about a removed command; not the same underlying issue." }, { "left": "issue:43975", "right": "issue:44292", "accept": false, - "reason": "Incorrect detokenization and NVFP4 runtime error are different problems." + "reason": "Different model-specific inference/tokenization failures with no shared code-path evident." }, { "left": "issue:44292", "right": "issue:45507", "accept": false, - "reason": "Qwen3-8B-NVFP4 runtime failure does not match GraniteMoEHybrid invalid-method behavior." + "reason": "NVFP4 runtime error and GraniteMoEHybrid invalid method are unrelated." }, { "left": "issue:43950", "right": "issue:44534", "accept": true, - "reason": "Same underlying regression: Transformers 5.x non-persistent buffers being corrupted/junk-filled on load." + "reason": "Same regression: non-persistent buffers getting corrupted/junk-filled during loading in Transformers v5." }, { "left": "issue:43644", "right": "issue:43950", - "accept": false, - "reason": "AutoConfig missing fields with trust_remote_code is a different bug from non-persistent buffer corruption." + "accept": true, + "reason": "These describe the same non-persistent-buffer junk/corruption regression in Transformers 5.x." }, { "left": "issue:39692", "right": "issue:43054", "accept": false, - "reason": "Docs example errors and degraded SigLIP2 embeddings are not the same concrete bug." + "reason": "Both concern SigLIP2, but one is a broken docs example and the other is degraded text embeddings; not the same bug." }, { "left": "issue:43828", "right": "issue:44928", "accept": false, - "reason": "Autocast dtype mismatch on Phi-tiny-MoE and Qwen3.5 RLHF gradient explosion are different failures." + "reason": "Different problems: autocast dtype mismatch in Phi-tiny-MoE vs Qwen3.5 embedding unpacking in TRL SFT." }, { "left": "issue:41628", "right": "issue:45800", "accept": false, - "reason": "AutoImageProcessor import failure and torch/transformers incompatibility are related only at a broad ecosystem level, not the same bug." + "reason": "Import failure for AutoImageProcessor and torch/transformers version incompatibility are separate issues." }, { "left": "issue:36331", "right": "issue:41762", "accept": false, - "reason": "Unexpected trainer kwarg and Gemma3 ZeRO-3 loading IndexError are unrelated." - }, - { - "left": "issue:43950", - "right": "issue:45902", - "accept": true, - "reason": "Same non-persistent-buffer materialization bug: inv_freq reads uninitialized memory after meta -> to_empty(cuda), matching the regression in 43950." + "reason": "Training custom loss API mismatch is unrelated to Gemma3 loading under ZeRO-3." }, { "left": "issue:43872", "right": "issue:44589", "accept": false, - "reason": "Different bitsandbytes/storage incompatibilities with different error surfaces." + "reason": "bitsandbytes constructor incompatibility and missing Float8 storage type are different failures." + }, + { + "left": "issue:43950", + "right": "issue:45902", + "accept": false, + "reason": "The latter is about Qwen3_5Moe rotary embedding uninitialized memory after meta->to_empty, not the same as generic non-persistent-buffer corruption." }, { "left": "issue:43866", "right": "issue:44403", "accept": false, - "reason": "A corrupted checkpoint and loading noise are not the same bug." + "reason": "Checkpoint corruption for Ovis2 and generic loader noise are not the same underlying bug." }, { "left": "issue:43054", "right": "issue:45701", "accept": false, - "reason": "SigLIP2 embedding quality regression and tokenization changes across versions are too broad to treat as the same bug." + "reason": "SigLIP2 embedding quality vs tokenization changes are not the same defect." }, { "left": "issue:41553", "right": "issue:43883", "accept": false, - "reason": "Bad AutoTokenizer error messaging and missing all_tied_weights_keys are unrelated." + "reason": "Bad AutoTokenizer error message for Voxtral and missing all_tied_weights_keys for Molmo are unrelated." }, { "left": "issue:38175", "right": "issue:43054", "accept": false, - "reason": "Both involve SigLIP2 output quality, but the specific failure modes are not concrete enough to call the same bug." + "reason": "Unexpected zero probabilities and degraded SigLIP2 text embeddings are different model behaviors." }, { "left": "issue:44910", "right": "issue:45859", "accept": false, - "reason": "Flash-attention illegal memory access from 3D position_ids is different from CPU offload incompatibility in Qwen3_5MoeTextRotaryEmbedding." + "reason": "Flash-Attention illegal access from 3D position_ids is unrelated to CPU offload incompatibility in Qwen3_5Moe rotary embeddings." }, { "left": "issue:45663", "right": "issue:45902", "accept": false, - "reason": "Gemma4 FSDP2 shared_kv_states KeyError is unrelated to Qwen3_5MoeTextRotaryEmbedding buffer materialization." + "reason": "Gemma4 FSDP2 shared_kv_states rebuilding and Qwen3_5Moe rotary embedding memory initialization are unrelated." }, { "left": "issue:41553", "right": "issue:43881", "accept": false, - "reason": "Voxtral tokenizer error messaging and glm-4v-9b loading failure are different issues." + "reason": "Different model-loading failures with different root causes." }, { "left": "issue:42947", "right": "issue:44387", "accept": false, - "reason": "LoRA gradient checkpointing ineffectiveness and int4 quantization OOM are different regressions." + "reason": "Gradient checkpointing ineffectiveness with LoRA and int4 reserved-memory/OOM regression are not the same bug." }, { "left": "issue:43054", "right": "issue:43493", - "accept": true, - "reason": "Same SigLIP2 implementation discrepancy family: worse text embeddings line up with a broader HF-vs-JAX implementation mismatch." + "accept": false, + "reason": "Different SigLIP2 concerns: embedding quality versus HF/JAX implementation discrepancy." }, { "left": "issue:44403", "right": "issue:44451", "accept": false, - "reason": "Loading noise and inability to load a specific model are not the same underlying bug." + "reason": "Generic loader noise and a specific model load failure are not the same underlying issue." }, { "left": "issue:43975", "right": "issue:45237", "accept": false, - "reason": "Detokenization regression and AMD GPU runtime failure are unrelated." + "reason": "Incorrect detokenization in v5 and AMD GPU model failure are unrelated." }, { "left": "issue:43643", "right": "issue:45020", "accept": false, - "reason": "Trust-remote-code missing fields and broad remote_code breakage are too general to merge as one concrete bug." + "reason": "Missing fields from AutoConfig with trust_remote_code is too specific to merge with the broader remote_code breakage report." }, { "left": "issue:44403", "right": "issue:44488", "accept": false, - "reason": "Generic loading noise and model-specific loading failure are different issues." + "reason": "Generic loading noise is not the same as a specific model load failure." }, { "left": "issue:42947", "right": "issue:43856", "accept": false, - "reason": "Gradient checkpointing/LoRA and Qwen3 MoE memory usage are different problems." + "reason": "LoRA gradient checkpointing and Qwen3 MoE memory usage are different training-memory issues." }, { "left": "issue:41762", "right": "issue:42915", "accept": false, - "reason": "Gemma3 ZeRO-3 loading and Qwen3Moe FP8 config failure are unrelated." + "reason": "Gemma3 ZeRO-3 load failure and Qwen3Moe FP8 failure are unrelated." }, { "left": "issue:43881", "right": "issue:43883", "accept": false, - "reason": "glm-4v-9b loading failure and Molmo missing all_tied_weights_keys are different model-specific bugs." + "reason": "Two separate loading errors with different symptoms and likely different code paths." }, { "left": "issue:43782", "right": "issue:44960", "accept": false, - "reason": "Qwen3VL weight_only load error and GLM5 are unrelated." + "reason": "Qwen3VL weight_only loading error and GLM5 issue are unrelated." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is grounded in the packet and stays mostly conservative. The 43950/44534/45902 grouping is well supported by the titles, and the rejected pairs are handled cautiously. The only mildly speculative point is the accepted 43054/43493 edge, but it is still phrased as a plausible pair rather than a clear cluster, so it does not overstate the evidence." + "feedback": "The summary is grounded in the packet and stays conservative: it correctly treats the cluster as mostly heterogeneous while identifying the non-persistent-buffer corruption issues as the only clear duplicate family. The accepted soft-edge pairs are well supported by the issue titles and wording, and the rejected pairs avoid over-merging distinct bugs." }, "evaluator_used": true, "retried": false, @@ -1226,7 +1190,7 @@ "component_index": 1, "component_count": 17, "review_unit_index": 4, - "review_unit_count": 50, + "review_unit_count": 49, "status": "reviewed", "reason": null, "source": "llm", @@ -1294,8 +1258,8 @@ "issue:45081|issue:45399", "issue:45020|issue:45356", "issue:43742|issue:45399", - "issue:43404|issue:43950", "issue:43827|issue:44509", + "issue:43404|issue:43950", "issue:43638|issue:45823", "issue:42915|issue:45020", "issue:43792|issue:44220", @@ -1318,20 +1282,20 @@ "issue:43366|issue:43828", "issue:45325|issue:45902" ], - "prepared_review_unit_hash": "c203cbed20ac2b2ce2714e44ae563dad9f9717e5e65939072991c1d1f608ee78", + "prepared_review_unit_hash": "58b4d1f2ad7e2e4078b8d393052617350d4a72674f8154e6933c5ff9e3bb0d1e", "component_budget": { "node_count": 426, "item_count": 426, - "soft_pair_count": 1572, - "serialized_chars": 814130, - "estimated_input_tokens": 203533, - "estimated_eval_tokens": 407322 + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 33, - "serialized_chars": 38075, + "serialized_chars": 38073, "estimated_input_tokens": 9519, "estimated_eval_tokens": 19294 }, @@ -1340,204 +1304,204 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "A few clear duplicate pairs emerged around identical error signatures and stale docs, but most soft-similarity pairs are only thematically related and should stay separate. No pull requests are present in the cluster.", - "confidence": 0.66, - "canonical_issue_reason": "Best canonical issue: issue:44291. It states the root cause and version window most precisely for the shared `_is_hf_initialized` TypeError, making it the cleanest anchor for that duplicate pair.", + "summary": "Mostly a mixed bag of unrelated runtime regressions; only a few soft pairs look like true duplicates, concentrated around shared error paths in docs, audio feature extraction, kwarg propagation, mask-shape mismatches, and Qwen video regressions.", + "confidence": 0.73, + "canonical_issue_reason": "Issue 44291 is the clearest, most root-cause-focused report of the `_is_hf_initialized` regression and subsumes the closely related bitsandbytes/init-empty-weights failure mode.", "canonical_pr_reason": null, - "best_issue_reason": "Issue:44291 is the strongest representative because it is specific, actionable, and directly subsumes the closely matching bitsandbytes crash report.", + "best_issue_reason": "Best representative of the cluster: it names a concrete, reproducible framework regression with a specific cause and version boundary, making it the most actionable canonical issue.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45685", "right": "issue:45823", "accept": false, - "reason": "Both are runtime/model configuration failures, but the symptoms and code paths are unrelated." + "reason": "Different failure modes: MPS int-histogram limitation vs a PLE device-mismatch forward bug." }, { "left": "issue:43824", "right": "issue:43872", "accept": false, - "reason": "ImportError for a missing class and a bitsandbytes constructor TypeError are different bugs." + "reason": "Both are load/runtime errors, but one is a Qwen2.5-VL import problem and the other is a bitsandbytes kwarg incompatibility." }, { "left": "issue:44991", "right": "issue:45399", "accept": false, - "reason": "Tokenizer loading regression vs flash-attn fallback gating; not the same failure." + "reason": "Tokenizer load failure for a specific model versus flash-attn fallback gating; only broad version-regression theme overlaps." }, { "left": "issue:45020", "right": "issue:45823", "accept": false, - "reason": "Broad remote_code regression and Gemma4 device mismatch are distinct issues." + "reason": "Remote-code compatibility regression is too broad to match a specific device-map forward mismatch." }, { "left": "issue:42491", "right": "issue:44368", "accept": false, - "reason": "Qwen3-MoE LoRA compatibility and a tie_word_embeddings warning are not the same bug." + "reason": "Both mention Qwen/LoRA warnings, but one is a version-compatibility failure and the other is a config warning about tied embeddings." }, { "left": "issue:44661", "right": "issue:45020", "accept": false, - "reason": "Tokenizer mapping validation failure is separate from the remote_code regression." + "reason": "Generic model-addition/tokenizer-mapping bug is not the same concrete problem as remote_code breakage." }, { "left": "issue:43828", "right": "issue:45823", "accept": false, - "reason": "Autocast dtype mismatch and device_map device mismatch are different code-path failures." + "reason": "Dtype mismatch under autocast is unrelated to the Gemma4 PLE device-placement bug." }, { "left": "issue:45081", "right": "issue:45399", "accept": false, - "reason": "Mistral regex patch crash is unrelated to flash-attn fallback checks." + "reason": "Mistral tokenizer regex crash and flash-attn fallback gating are different code paths." }, { "left": "issue:45020", "right": "issue:45356", "accept": false, - "reason": "These are both tokenizer-related regressions, but they affect different models and different bugs." + "reason": "Remote-code regressions are too broad; Kimi-K2.5 tokenizer codec handling is a separate tokenizer-specific bug." }, { "left": "issue:43742", "right": "issue:45399", "accept": false, - "reason": "MobileLLM load key error and flash-attn fallback gating are unrelated." - }, - { - "left": "issue:43404", - "right": "issue:43950", - "accept": false, - "reason": "Untied lm_head weights vs non-persistent buffer corruption are different mechanisms." + "reason": "MobileLLM load key error is unrelated to flash-attn fallback checks." }, { "left": "issue:43827", "right": "issue:44509", "accept": true, - "reason": "Same stale v5 docs issue: both point to removed pipeline/task references that still remain in documentation." + "reason": "Same docs regression: both report stale pipeline-task references removed in v5." + }, + { + "left": "issue:43404", + "right": "issue:43950", + "accept": false, + "reason": "Tied lm_head warning and non-persistent buffer corruption are different model-state issues." }, { "left": "issue:43638", "right": "issue:45823", "accept": false, - "reason": "ZeRO-3/Bert index error and Gemma4 forward device mismatch are distinct." + "reason": "ZeRO-3 zero-size index error is unrelated to Gemma4 PLE device mismatch." }, { "left": "issue:42915", "right": "issue:45020", "accept": false, - "reason": "LoRA checkpoint compatibility across versions is not the same as the remote_code loading regression." + "reason": "Specific Qwen3-MoE hf4.x/hf5.x incompatibility is not the same as generic remote_code breakage." }, { "left": "issue:43792", "right": "issue:44220", - "accept": false, - "reason": "Whisper failing to run and a specific fbank helper issue are too indirect to merge as one bug." + "accept": true, + "reason": "Both point to the same audio feature-extraction path causing Whisper to fail." }, { "left": "issue:43257", "right": "issue:43866", "accept": false, - "reason": "DeepSpeed weight conversion on Qwen3 MoE and Ovis2 checkpoint corruption are unrelated." + "reason": "Qwen3 MoE weight conversion under accelerate+deepspeed does not match a corrupted Ovis2 checkpoint report." }, { "left": "issue:43872", "right": "issue:44291", "accept": true, - "reason": "Same concrete `_is_hf_initialized` TypeError during `init_empty_weights`/model loading." + "reason": "Same underlying `_is_hf_initialized` argument regression, just surfaced through different load/init paths." }, { "left": "issue:42915", "right": "issue:43872", "accept": false, - "reason": "LoRA model compatibility and the bitsandbytes kwarg crash are different issues." + "reason": "Different bugs: Qwen3-MoE cross-version compatibility versus the `_is_hf_initialized` kwarg issue." }, { "left": "issue:41553", "right": "issue:44991", "accept": false, - "reason": "Voxtral error-message quality and est-roberta tokenizer loading failure do not share the same bug." + "reason": "Bad Voxtral tokenizer error message is not the same as a tokenizer load failure for EMBEDDIA/est-roberta." }, { "left": "issue:43940", "right": "issue:45313", "accept": false, - "reason": "Both involve ZeRO-3 weight loading, but the model families and concrete failures differ." + "reason": "Both are ZeRO-3 weight-loading failures, but the model variants and missing-parameter patterns are different enough to avoid merging." }, { "left": "issue:36331", "right": "issue:41093", "accept": false, - "reason": "Unexpected trainer kwarg and mask/tensor shape mismatch are unrelated." + "reason": "Unexpected compute_loss kwarg and mask-shape IndexError are unrelated training/runtime failures." }, { "left": "issue:43854", "right": "issue:44863", "accept": false, - "reason": "Different model-loading failures for different architectures." + "reason": "GLM-4.7 test-loading failure and NemotronH checkpoint-loading failure are separate model-specific bugs." }, { "left": "issue:43257", "right": "issue:45305", "accept": false, - "reason": "Weight conversion/loading and gradient averaging under GAS are separate defects." + "reason": "DeepSpeed is shared, but weight-conversion failure and gradient-averaging bug are different code paths." }, { "left": "issue:42947", "right": "issue:45923", "accept": false, - "reason": "Both touch gradient checkpointing, but one is ineffective LoRA behavior and the other is a missing config flag." + "reason": "Gradient checkpointing/LoRA behavior is not the same bug as a missing supports_gradient_checkpointing flag on a remote-code model." }, { "left": "issue:44928", "right": "issue:45588", "accept": false, - "reason": "RLHF NaNs from position_ids and a flash-attn crash on s_aux=None are distinct problems." + "reason": "NaN gradient explosion in RLHF is unrelated to flash-attention crash on sink-less models." }, { "left": "issue:43872", "right": "issue:45538", "accept": false, - "reason": "bitsandbytes kwarg crash and CLIPTokenizer max-length handling are unrelated." + "reason": "The `model_max_length` CLIPTokenizer issue is unrelated to the `_is_hf_initialized` regression." }, { "left": "issue:41093", "right": "issue:44805", "accept": true, - "reason": "Same mask-shape IndexError; 44805 is the later, more detailed manifestation of the same bug." + "reason": "These are the same mask/tensor shape mismatch error, with only the report details differing." }, { "left": "issue:41553", "right": "issue:44488", "accept": false, - "reason": "Voxtral loading UX and cjvt/sleng-bert tokenizer loading are different model-specific issues." + "reason": "Voxtral tokenizer error-message quality and sleng-bert loading failure are different tokenizer issues." }, { "left": "issue:44479", "right": "issue:44560", "accept": true, - "reason": "Same v5.3.0 video-input regression in the Qwen3-VL family; 44560 is a specific StopIteration manifestation." + "reason": "Both describe the same 5.3.0 video-input regression across Qwen VL-family models." }, { - "left": "issue:43366", - "right": "issue:43828", + "left": "issue:44530", + "right": "issue:44863", "accept": false, - "reason": "GGUF gpt-oss support and Phi-tiny-MoE autocast dtype mismatch are unrelated." + "reason": "PagedAttentionCache linear_attention crash and NemotronH checkpoint loading are unrelated." }, { "left": "issue:45325", "right": "issue:45902", "accept": false, - "reason": "Still-image position_ids scaling and uninitialized-memory rotary state are different rotary/initialization bugs." + "reason": "Still-image rope scaling bug and uninitialized `inv_freq` after `to_empty(cuda)` are separate rotary-embedding regressions." } ] }, "evaluator_result": { "accept": true, - "feedback": "The cluster review is grounded in the packet: the canonical issue choice is well-supported by the matching `_is_hf_initialized` TypeError, the accepted duplicate pairs are limited to clearly matching titles/error signatures, and the rejected pairs stay conservative. The summary also correctly notes that no PRs are present." + "feedback": "The summary is grounded in the packet and stays conservative about the cluster being mostly unrelated runtime regressions. The accepted soft pairs are also reasonably supported by the titles/body metadata, especially the docs, mask-shape, _is_hf_initialized, and Qwen video-regression matches." }, "evaluator_used": true, "retried": false, @@ -1550,7 +1514,7 @@ "component_index": 1, "component_count": 17, "review_unit_index": 5, - "review_unit_count": 50, + "review_unit_count": 49, "status": "reviewed", "reason": null, "source": "llm", @@ -1640,232 +1604,220 @@ "issue:42491|issue:45276", "issue:43708|issue:44075" ], - "prepared_review_unit_hash": "515b59d586bef3495f823c10a6c13eec3c12354d688318dc86aa202a35bf3cad", + "prepared_review_unit_hash": "c50670ff3f2b4fa7f3aeb05f25bcbd1f57a7215ba757cf783875489d9e8724ca", "component_budget": { "node_count": 426, "item_count": 426, - "soft_pair_count": 1572, - "serialized_chars": 814130, - "estimated_input_tokens": 203533, - "estimated_eval_tokens": 407322 + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 }, "budget": { "node_count": 47, "item_count": 47, "soft_pair_count": 32, - "serialized_chars": 36622, - "estimated_input_tokens": 9156, - "estimated_eval_tokens": 18568 + "serialized_chars": 36625, + "estimated_input_tokens": 9157, + "estimated_eval_tokens": 18570 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The cluster is mixed, but there are two clear duplicate subgroups: Transformers 5 non-persistent buffer corruption and current-version BERT checkpoint load failures. Most other soft pairs are broader-similarity matches rather than true duplicates.", - "confidence": 0.68, - "canonical_issue_reason": "No single issue covers the whole heterogeneous cluster; #43646 is the best broad anchor because it captures a generic Transformers 5 custom-model initialization regression and overlaps directly with a near-duplicate report.", + "summary": "This set is mostly heterogeneous, but a few true duplicate pockets stand out. The strongest canonical thread is the Transformers v5 non-persistent-buffer regression, with a near-verbatim follow-up and a couple of adjacent custom-init reports; a few other pairs are only loosely similar and should stay separate.", + "confidence": 0.76, + "canonical_issue_reason": "Issue 43644 is the cleanest representative of the main duplicate pocket: it states a specific Transformers 5.0.0 regression about non-persistent buffers, and 44534 is a near-verbatim later report of the same bug.", "canonical_pr_reason": null, - "best_issue_reason": "#43646 is the best overall issue to keep as the umbrella report: it is concrete, broadly framed, and likely to subsume related Transformers 5 initialization failures better than the model-specific reports.", + "best_issue_reason": "43644 is the best standalone issue to keep as the canonical bug report because it is concrete, narrowly scoped, and has the clearest duplicate signal among the candidates.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45663", "right": "issue:45823", "accept": false, - "reason": "Different bugs: Gemma4 FSDP2 training KeyError vs Gemma4 PLE device mismatch in auto device mapping." + "reason": "Different Gemma4 failures: FSDP2 shared_kv_states KeyError vs device_map=\"auto\" PLE device mismatch." }, { "left": "issue:43366", "right": "issue:43531", "accept": false, - "reason": "Different problem classes: GGUF/gpt-oss support request vs Qwen3-MoE sliding_window behavior." + "reason": "Unrelated problems: gpt-oss GGUF support vs Qwen3-MoE sliding_window behavior." }, { "left": "issue:44155", "right": "issue:45584", "accept": false, - "reason": "Different code paths and symptoms: AudioFlamingo3 batch leakage vs Whisper empty-transcription generation failure." + "reason": "Different model paths and symptoms: AudioFlamingo3 batched leakage vs Whisper empty-transcription generation." }, { "left": "issue:43950", "right": "issue:45834", "accept": false, - "reason": "Different regressions: non-persistent buffer corruption vs Kosmos2.5 OCR long-input index error." + "reason": "Non-persistent buffer corruption and Kosmos2.5 long-OCR index errors are different bugs." }, { "left": "issue:44560", "right": "issue:45685", "accept": false, - "reason": "Unrelated model/runtime issues: Qwen3-vl StopIteration vs MPS histogram not implemented for Int." + "reason": "Qwen3-VL video StopIteration and an MPS histogram_int limitation are unrelated." }, { "left": "issue:43824", "right": "issue:43957", "accept": false, - "reason": "Different failures: missing export/import for a specific class vs meta-device loading breaking some models." + "reason": "ImportError for a missing class and meta-device loading regressions are distinct code paths." }, { "left": "issue:43550", "right": "issue:44910", "accept": false, - "reason": "Different models and code paths: Bamba-9B-v2 SDPA/torch.compile failure vs Qwen3.5 flash-attention packed-sequence bug." + "reason": "Different attention/compile failures on different models; not the same concrete bug." }, { "left": "issue:42915", "right": "issue:43957", "accept": false, - "reason": "Only broad similarity around model loading; the FP8/Qwen3MoE failure and meta-device loading issue are not the same bug." + "reason": "Qwen3Moe FineGrainedFP8Config failure is separate from generic torch.device(\"meta\") loading issues." }, { "left": "issue:43366", "right": "issue:44589", "accept": false, - "reason": "Different bugs: GGUF gpt-oss support vs Float8 storage lookup failure." + "reason": "GGUF gpt-oss support and missing Float8 storage are unrelated loading problems." }, { "left": "issue:43645", "right": "issue:43646", "accept": true, - "reason": "Same underlying Transformers 5 custom-model initialization regression; the notebook-specific report and the general report are the same bug family." + "reason": "Both describe the same Transformers v5 custom-model initialization regression; the notebook detail in 43645 is incidental." }, { "left": "issue:44560", "right": "issue:44918", "accept": false, - "reason": "Different symptoms: video StopIteration in Qwen3-vl-embedding vs TRL SFT unpacking failure for Qwen3.5 embeddings." + "reason": "Different Qwen3.5/Qwen3-VL failures with different triggers and stacks." }, { "left": "issue:42617", "right": "issue:43366", "accept": false, - "reason": "Not the same bug: 3d_parallel.py runtime issue vs GGUF architecture support request." + "reason": "3d_parallel.py execution failure is not the same bug as gpt-oss GGUF support." }, { "left": "issue:36010", "right": "issue:42915", "accept": false, - "reason": "Completely different failures: GenerationMixin import breakage vs Qwen3-MoE FP8/device-mapping failure." + "reason": "GenerationMixin import failure and Qwen3MoE FP8 issues are unrelated." }, { "left": "issue:44661", "right": "issue:45823", "accept": false, - "reason": "Different layers: tokenizer mapping/add-new-model-like validation vs Gemma4 runtime device mismatch." + "reason": "Tokenizer-mapping registration bug and Gemma4 device mismatch are different problems." }, { "left": "issue:44560", "right": "issue:44805", "accept": false, - "reason": "Different issues: Qwen3-vl video StopIteration vs generic mask-shape IndexError." + "reason": "Qwen3-VL video StopIteration is not the same as a generic mask shape mismatch." }, { "left": "issue:43632", "right": "issue:45834", "accept": false, - "reason": "Different regressions: _is_hf_initialized flag breakage vs Kosmos2.5 OCR long-input index error." + "reason": "_is_hf_initialized regression does not match the Kosmos2.5 long-input index error." }, { "left": "issue:38175", "right": "issue:43994", - "accept": false, - "reason": "Same model family but not clearly the same bug from the titles alone; zero probabilities and nonsensical outputs may be related, but evidence is too weak to merge confidently." + "accept": true, + "reason": "Both concern broken SigLIP2-base-patch16-224 inference producing nonsensical/zero outputs through the same model/processor path." }, { "left": "issue:43638", "right": "issue:44661", "accept": false, - "reason": "Different failures: DeepSpeed ZeRO-3 Bert indexing vs tokenizer mapping restrictions for add-new-model-like." + "reason": "Zero3 Bert initialization and tokenizer-mapping registration are not the same underlying issue." }, { "left": "issue:43596", "right": "issue:44661", "accept": false, - "reason": "Different model-loading paths and symptoms; Bert zero3 init index error is not the same as tokenizer mapping validation." + "reason": "Bert Zero3 index error is a different failure mode from the tokenizer mapping bug." }, { "left": "issue:44661", "right": "issue:44805", "accept": false, - "reason": "Unrelated: tokenizer mapping validation vs generic mask/tensor shape mismatch." + "reason": "Tokenizer mapping validation and mask/tensor shape mismatch are unrelated." }, { "left": "issue:44877", "right": "issue:45030", "accept": false, - "reason": "Both involve config validation, but the affected models and exact failures differ; not the same underlying bug." + "reason": "Strict config blocking granite_speech and glm4v config validation are separate issues." }, { "left": "issue:43317", "right": "issue:44756", "accept": false, - "reason": "Different issues: gpu+cpu offload dequantized loading failure vs mmap/OOM concern on Strix Halo." + "reason": "dequantized model offload failure and mmap/OOM avoidance on Strix Halo are not the same bug." }, { "left": "issue:43644", "right": "issue:44534", "accept": true, - "reason": "Same underlying v5 regression: non-persistent buffers being filled/corrupted during model loading." + "reason": "These are near-verbatim reports of the same Transformers v5 non-persistent-buffer junk regression." }, { "left": "issue:38175", "right": "issue:39692", "accept": false, - "reason": "Related SigLIP2 area, but one is output-quality/runtime behavior and the other is documentation/example errors; not the same bug." + "reason": "Both mention SigLIP2, but 39692 is a docs/example problem with additional quantization issues, not clearly the same concrete bug as 38175." }, { "left": "issue:42491", "right": "issue:43720", "accept": false, - "reason": "Different subsystems and failures: Qwen3 MoE LoRA compatibility vs BitNet packed-weight loading in accelerate." + "reason": "LoRA cross-version compatibility for qwen3_moe and BitNet packed-weight loading are different code paths." }, { "left": "issue:41720", "right": "issue:42915", "accept": false, - "reason": "Same model family only loosely; cuda assert on A800 and FP8 config failure are distinct problems." + "reason": "cudaErrorAssert on A800 with auto device mapping is unrelated to FineGrainedFP8Config failure." }, { "left": "issue:44960", "right": "issue:44977", "accept": false, - "reason": "Different models and behaviors: GLM5 generic issue vs Qwen3.5 flash-attention generation bug." + "reason": "GLM5 and Qwen3.5 flash-attention generation failures are different model-specific bugs." }, { "left": "issue:45313", "right": "issue:45357", "accept": false, - "reason": "Both are Qwen3.5, but one is ZeRO-3 language_model loading and the other is save_pretrained visual-encoder key regression." + "reason": "ZeRO-3 language_model weight loading and save_pretrained visual-key corruption are separate issues." }, { "left": "issue:44451", "right": "issue:44488", "accept": true, - "reason": "Same underlying model-loading regression on current versions; both reports say BERT-family checkpoints no longer load." - }, - { - "left": "issue:42915", - "right": "issue:44756", - "accept": false, - "reason": "Different bugs: Qwen3-MoE FP8 failure vs mmap/OOM on Strix Halo." + "reason": "Both report the same current-version regression loading specific BERT checkpoints, likely the same model-loading breakage." }, { "left": "issue:42491", "right": "issue:45276", "accept": false, - "reason": "Different problems: Qwen3 MoE LoRA hf5 compatibility vs Gemma4 resize_token_embeddings not updating tied structures." - }, - { - "left": "issue:43708", - "right": "issue:44075", - "accept": false, - "reason": "Unrelated trainer scheduling bug vs optimizer argument handling." + "reason": "qwen3_moe LoRA incompatibility and Gemma4 resize_token_embeddings behavior are unrelated." } ] }, "evaluator_result": { "accept": true, - "feedback": "The evaluation is grounded in the packet, and the soft-edge verdicts are generally conservative. The three accepted duplicate pairs are plausible from the issue titles/summaries, and the rest are mostly appropriately rejected as different bug classes." + "feedback": "The canonical 43644/44534 duplicate pocket is well grounded, and the other accepted pairs are phrased cautiously enough to stay within the packet evidence. The rejected pairs also look appropriately conservative. No obvious overstatement stands out." }, "evaluator_used": true, "retried": false, @@ -1878,7 +1830,7 @@ "component_index": 1, "component_count": 17, "review_unit_index": 6, - "review_unit_count": 50, + "review_unit_count": 49, "status": "reviewed", "reason": null, "source": "llm", @@ -1940,12 +1892,12 @@ "issue:43257|issue:43854", "issue:45488|issue:45812", "issue:44488|issue:44991", - "issue:39692|issue:43994", "issue:43644|issue:43645", + "issue:39692|issue:43994", "issue:43856|issue:43873", "issue:42491|issue:43931", - "issue:43856|issue:44387", "issue:45362|issue:45406", + "issue:43856|issue:44387", "issue:44960|issue:46050", "issue:42491|issue:45292", "issue:44910|issue:45588", @@ -1959,8 +1911,8 @@ "issue:44402|issue:44488", "issue:42947|issue:45704", "issue:43873|issue:44387", - "issue:45663|issue:45759", "issue:44291|issue:44756", + "issue:45663|issue:45759", "issue:45663|issue:45859", "issue:44155|issue:45823", "issue:43335|issue:45663", @@ -1970,17 +1922,17 @@ "issue:43854|issue:45216", "issue:45276|issue:45292", "issue:42947|issue:43381", - "issue:44164|issue:44756", - "issue:43381|issue:44928" + "issue:43381|issue:44928", + "issue:44164|issue:44756" ], - "prepared_review_unit_hash": "231ab0da021d42b6f5107dc3745dcd4ba821fd02899eb33289bdcaa7a5e4a1ce", + "prepared_review_unit_hash": "84bf4590b2367a88d6d0a96b6b051b222e0300a2072204da53864b5ce195b28a", "component_budget": { "node_count": 426, "item_count": 426, - "soft_pair_count": 1572, - "serialized_chars": 814130, - "estimated_input_tokens": 203533, - "estimated_eval_tokens": 407322 + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 }, "budget": { "node_count": 48, @@ -1995,234 +1947,234 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly heterogeneous one-off bug reports with only one clear duplicate-like pair around `resize_token_embeddings` not updating output embeddings. No pull requests are present.", - "confidence": 0.78, - "canonical_issue_reason": "`issue:45292` is the cleanest, model-agnostic statement of the `resize_token_embeddings` regression and serves well as the duplicate target for the more specific Gemma4 report.", + "summary": "This cluster is mostly a grab-bag of unrelated Transformers bugs. The only credible duplicate subcluster is around `resize_token_embeddings` not updating model-specific output embeddings, where one issue is the broader generic report and the other is a Gemma4-specific manifestation.", + "confidence": 0.64, + "canonical_issue_reason": "Issue 45292 is the best anchor because it is the broad, model-agnostic version of the only apparent duplicate family (`resize_token_embeddings` not affecting `output_embeddings`).", "canonical_pr_reason": null, - "best_issue_reason": "`issue:45292` is the best representative because it names the core bug directly and broadly, without tying it to one model variant.", + "best_issue_reason": "Issue 45292 is the strongest representative overall: it is the most general and concise description of the shared `resize_token_embeddings`/`output_embeddings` regression, while the nearby issues are narrower model-specific variants.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43854", "right": "issue:43931", "accept": false, - "reason": "Both are model-loading failures, but they involve different models and different symptoms, so the root cause is not clearly shared." + "reason": "Different bugs: GLM-4.7 unit-test load failure vs Qwen3-VL weight-shape mismatch." }, { "left": "issue:43257", "right": "issue:43854", "accept": false, - "reason": "Qwen3 MoE weight conversion under accelerate/deepspeed is a different bug from a GLM model-loading failure." + "reason": "Different model-loading failures with unrelated root causes (Qwen3 MoE conversion vs GLM-4.7 test load)." }, { "left": "issue:45488", "right": "issue:45812", "accept": false, - "reason": "Both concern tokenizer behavior, but one is a DeepSeek byte-level pretokenizer regression and the other is a Granite token-id regression." + "reason": "Both are tokenizer regressions, but they target different model families and different failure modes (`LlamaTokenizer` pre-tokenizer override vs Granite token IDs)." }, { "left": "issue:44488", "right": "issue:44991", "accept": false, - "reason": "These are separate tokenizer/load failures for different models, with no indication of the same underlying defect." + "reason": "Different tokenizer-load problems for different models; no clear shared code path." }, { - "left": "issue:39692", - "right": "issue:43994", + "left": "issue:43644", + "right": "issue:43645", "accept": false, - "reason": "Both mention SigLIP2, but one is a docs example mismatch/quantization issue and the other is incorrect inference results; likely different causes." + "reason": "Both are Transformers 5.0 regressions, but one is junk non-persistent buffers and the other is custom-model initialization in Jupyter; unrelated." }, { - "left": "issue:43644", - "right": "issue:43645", + "left": "issue:39692", + "right": "issue:43994", "accept": false, - "reason": "Both are Transformers 5.0 regressions, but one is about junk non-persistent buffers and the other about notebook model initialization." + "reason": "Same model family (SigLIP2) but different reported issues: docs/example errors and quantization failure vs nonsensical inference results." }, { "left": "issue:43856", "right": "issue:43873", "accept": false, - "reason": "Both touch Qwen3 MoE/quantization memory behavior, but they describe different failure modes and code paths." + "reason": "Different problems: Qwen3 MoE training memory efficiency vs quantization/offloading behavior." }, { "left": "issue:42491", "right": "issue:43931", "accept": false, - "reason": "A Qwen3 MoE LoRA incompatibility on hf5.x is unrelated to a Qwen3-VL weight-shape mismatch." + "reason": "Unrelated issues: LoRA compatibility across HF versions vs Qwen3-VL load-time shape mismatch." }, { - "left": "issue:43856", - "right": "issue:44387", + "left": "issue:45362", + "right": "issue:45406", "accept": false, - "reason": "Both are memory-related, but one is MoE training efficiency and the other is int4 quantization increasing reserved CUDA memory." + "reason": "Different model/runtime crashes: Qwen3.5 chat failure vs Gemma4Processor missing `_tokenizer` in serve." }, { - "left": "issue:45362", - "right": "issue:45406", + "left": "issue:43856", + "right": "issue:44387", "accept": false, - "reason": "Different model families and different crash mechanisms; not the same bug." + "reason": "Both mention memory, but one is Qwen3 MoE training efficiency and the other is int4 quantization reserved-memory/OOM; not the same bug." }, { "left": "issue:44960", "right": "issue:46050", "accept": false, - "reason": "GLM5 is too vague to match a specific sinq save/load corruption bug." + "reason": "Different model families and failure modes; GLM5 vs sinq save/load corruption." }, { "left": "issue:42491", "right": "issue:45292", "accept": false, - "reason": "LoRA model compatibility on hf5.x is not the same issue as `resize_token_embeddings` failing to update output embeddings." + "reason": "Different root causes; Qwen3 MoE HF4.x/HF5.x incompatibility is not the same as a generic `resize_token_embeddings` regression." }, { "left": "issue:44910", "right": "issue:45588", "accept": false, - "reason": "Both involve flash attention, but one is a packed-sequence/position_ids crash and the other is an `s_aux=None` AttributeError." + "reason": "Both involve attention code, but one is Qwen3.5 flash-attn illegal memory access from 3D `position_ids`, the other is `s_aux=None` handling in sink-less models." }, { "left": "issue:44164", "right": "issue:44291", "accept": false, - "reason": "One is about `extra_state` in save/from_pretrained, the other about `init_empty_weights` rejecting `_is_hf_initialized`." + "reason": "Both are loading/saving internals, but one is `extra_state` serialization and the other is `init_empty_weights` argument compatibility." }, { "left": "issue:36010", "right": "issue:39404", "accept": false, - "reason": "An import-path regression for `GenerationMixin` is unrelated to Whisper pipeline `return_language` behavior." + "reason": "Completely different issues: import path breakage vs Whisper pipeline `return_language` regression." }, { "left": "issue:45276", "right": "issue:45335", "accept": false, - "reason": "They are very similar, but the titles point to different model-specific embedding locations and no concrete shared code path is shown." + "reason": "Similar API symptom, but different model families and target embeddings (`Gemma4` output/embed tokens vs `t5gemma` decoder embed tokens); not clearly the same bug." }, { "left": "issue:43404", "right": "issue:43454", "accept": false, - "reason": "Both are lm_head weight-tying bugs, but in different model classes with no evidence of the same underlying implementation fault." + "reason": "Both are weight-tying issues, but they affect different model classes and are not clearly the same concrete code-path problem." }, { "left": "issue:43425", "right": "issue:43957", "accept": false, - "reason": "Torch version incompatibility is a different problem from meta-device model loading breakage." + "reason": "Unrelated compatibility/loading bugs: Torch 2.10 vs meta-device loading regressions." }, { "left": "issue:41720", "right": "issue:44155", "accept": false, - "reason": "A Qwen3 auto-device-map CUDA assert is unrelated to AudioFlamingo3 token/embedding leakage during batched inference." + "reason": "Different subsystems and failures: Qwen3 auto device mapping CUDA assert vs AudioFlamingo3 batched inference leakage." }, { "left": "issue:39692", "right": "issue:44960", "accept": false, - "reason": "SigLIP2 docs errors and GLM5 loading are unrelated issues." + "reason": "Different model families and different problems; SigLIP2 docs/example issues are not the same as GLM5 model loading." }, { "left": "issue:44402", "right": "issue:44488", "accept": false, - "reason": "Both are model loading/tokenizer problems, but the first is a vocab-size mismatch and the second is a different model load failure." + "reason": "Both are token/vocab-related model-loading issues, but they concern different models and different underlying tokenizer problems." }, { "left": "issue:42947", "right": "issue:45704", "accept": false, - "reason": "Gradient checkpointing under PEFT LoRA is unrelated to a T5 RMSNorm memory leak." + "reason": "Different bugs: PEFT LoRA gradient checkpointing vs T5 accidentally using apex RMSNorm." }, { "left": "issue:43873", "right": "issue:44387", "accept": false, - "reason": "Offloading/quantization behavior and int4 reserved-memory OOM are related only broadly, not the same concrete bug." + "reason": "Both involve quantization/memory, but one is offloading not working and the other is increased CUDA reserved memory under int4; separate issues." }, { - "left": "issue:45663", - "right": "issue:45759", + "left": "issue:44291", + "right": "issue:44756", "accept": false, - "reason": "Gemma4 FSDP2 attention-state rebuilding and Qwen multimodal config unwrapping are different code paths." + "reason": "Different concerns: `init_empty_weights` type error vs Strix Halo mmap/OOM." }, { - "left": "issue:44291", - "right": "issue:44756", + "left": "issue:45663", + "right": "issue:45759", "accept": false, - "reason": "`init_empty_weights` argument handling is unrelated to disabling mmap for Strix Halo OOM." + "reason": "Different model architecture bugs: Gemma4 FSDP2 `shared_kv_states` vs Qwen 3.5/3.6 `text_config` unwrap behavior." }, { "left": "issue:45663", "right": "issue:45859", "accept": false, - "reason": "Gemma4 FSDP2 KeyError and Qwen3_5 rotary embedding CPU-offload incompatibility are distinct bugs." + "reason": "Different code paths: Gemma4 FSDP2 attention state vs Qwen3_5 MoE rotary embedding CPU offload." }, { "left": "issue:44155", "right": "issue:45823", "accept": false, - "reason": "AudioFlamingo3 batch leakage is unrelated to Gemma4 device-map mismatches." + "reason": "No shared underlying bug; AudioFlamingo3 track leakage is unrelated to Gemma4 PLE device mismatch." }, { "left": "issue:43335", "right": "issue:45663", "accept": false, - "reason": "SwitchTransformers sparse-layer construction and Gemma4 shared-kv-state rebuilding are unrelated." + "reason": "Different model families and failures; SwitchTransformers sparse-layer creation is unrelated to Gemma4 FSDP2 key errors." }, { "left": "issue:44292", "right": "issue:44912", "accept": false, - "reason": "Different quantization formats and different failure modes; not the same issue." + "reason": "Both involve quantization, but one is a Qwen-3 NVFP4 runtime error and the other is MXFP4 load fallback; not the same issue." }, { "left": "issue:44387", "right": "issue:44756", "accept": false, - "reason": "Both mention memory, but one is quantization OOM and the other is mmap-related OOM avoidance." + "reason": "Both mention memory, but one is int4 reserved-memory/OOM and the other is mmap on Strix Halo; unrelated." }, { "left": "issue:43957", "right": "issue:44756", "accept": false, - "reason": "Meta-device loading failures are unrelated to mmap being disabled to avoid OOM." + "reason": "Different bugs: meta-device loading regressions vs Strix Halo mmap/OOM." }, { "left": "issue:43854", "right": "issue:45216", "accept": false, - "reason": "A GLM model-load problem is not the same as the Qwen3.5 save_pretrained checkpoint regression." + "reason": "Different load/save bugs on different models: GLM-4.7 unit-test loading vs Qwen3.5 checkpoint save regression." }, { "left": "issue:45276", "right": "issue:45292", "accept": true, - "reason": "Same underlying `resize_token_embeddings` regression: the Gemma4-specific report is a concrete instance of the broader `output_embeddings` bug." + "reason": "Same underlying `resize_token_embeddings` regression: both report output embeddings not being updated, with 45276 adding a Gemma4-specific manifestation." }, { "left": "issue:42947", "right": "issue:43381", "accept": false, - "reason": "LoRA gradient checkpointing ineffectiveness is unrelated to gradient checkpointing being unusable in eval mode." + "reason": "Both involve gradient checkpointing language, but one is ineffective LoRA checkpointing and the other is eval-mode incompatibility." }, { - "left": "issue:44164", - "right": "issue:44756", + "left": "issue:43381", + "right": "issue:44928", "accept": false, - "reason": "Save/load `extra_state` handling is unrelated to mmap being disabled for OOM." + "reason": "Different checkpointing/SDPA problems; eval-mode checkpointing is unrelated to Qwen3.5 3D `position_ids` NaN behavior." }, { - "left": "issue:43381", - "right": "issue:44928", + "left": "issue:44164", + "right": "issue:44756", "accept": false, - "reason": "Eval-mode gradient checkpointing and Qwen3.5 3D position_ids/SDPA issues are different problems." + "reason": "Serialization `extra_state` handling is unrelated to mmap/OOM on Strix Halo." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is grounded in the packet and stays conservative. The single accepted duplicate-like pair, `issue:45276` / `issue:45292`, is well supported by the titles, and the other soft-edge rejections avoid overclaiming shared root causes. The note that no pull requests are present is also consistent with the packet." + "feedback": "The summary is well grounded in the packet: most issues are clearly unrelated by model/failure mode, and the one accepted duplicate pair (45276/45292) is a conservative, defensible match. The soft-edge rejections are generally cautious rather than overstated. One minor omission is that 45335 also looks adjacent to the same resize_token_embeddings family, but that does not materially undermine the overall assessment." }, "evaluator_used": true, "retried": false, @@ -2235,7 +2187,7 @@ "component_index": 1, "component_count": 17, "review_unit_index": 7, - "review_unit_count": 50, + "review_unit_count": 49, "status": "reviewed", "reason": null, "source": "llm", @@ -2243,6 +2195,7 @@ "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-426", "nodes": [ + "issue:32090", "issue:36032", "issue:39404", "issue:43257", @@ -2256,7 +2209,6 @@ "issue:43708", "issue:43716", "issue:43828", - "issue:43856", "issue:43901", "issue:43906", "issue:43931", @@ -2300,15 +2252,15 @@ "issue:43645|issue:44164", "issue:43828|issue:43957", "issue:45292|issue:45335", - "issue:43425|issue:43828", "issue:44402|issue:44451", + "issue:43425|issue:43828", "issue:43257|issue:44863", "issue:44683|issue:45823", "issue:44589|issue:44756", "issue:39404|issue:45701", "issue:43425|issue:45823", - "issue:43611|issue:43645", "issue:43425|issue:43716", + "issue:43611|issue:43645", "issue:44530|issue:45084", "issue:43906|issue:44530", "issue:39404|issue:44912", @@ -2324,228 +2276,234 @@ "issue:44164|issue:45759", "issue:36032|issue:43452", "issue:44671|issue:44991", - "issue:43856|issue:44756" + "issue:32090|issue:39404" ], - "prepared_review_unit_hash": "d4c1878db55df5fba2fcd564915f15c40451a823f853f0a2b9648c827dd19705", + "prepared_review_unit_hash": "6c8dabfe29e2b70254f899c5f1167e8a9f026d7bdf91ee577ce66076b3ad40d0", "component_budget": { "node_count": 426, "item_count": 426, - "soft_pair_count": 1572, - "serialized_chars": 814130, - "estimated_input_tokens": 203533, - "estimated_eval_tokens": 407322 + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 32, - "serialized_chars": 37355, - "estimated_input_tokens": 9339, - "estimated_eval_tokens": 18934 + "serialized_chars": 37406, + "estimated_input_tokens": 9352, + "estimated_eval_tokens": 18960 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a set of separate Transformers regression reports, many around v5 model loading and multimodal/model-specific edge cases. Only a couple of pairs look like true duplicates; most are just thematically similar.", - "confidence": 0.61, - "canonical_issue_reason": "Issue 43611 is the broadest representative of the cluster\u2019s main theme: Transformers 5.0 breaking model loading/initialization for models that rely on special config/model-class handling.", + "summary": "This cluster is heterogeneous and mostly contains distinct Transformers regressions across model loading, tokenization, attention backends, and training/runtime edge cases. One clear duplicate-like pair appears around `resize_token_embeddings` not propagating to output embeddings.", + "confidence": 0.62, + "canonical_issue_reason": "`issue:43611` is the best broad representative: it describes a general Transformers 5.0 model-loading regression, which matches the dominant theme of the cluster better than the narrower model-specific reports.", "canonical_pr_reason": null, - "best_issue_reason": "43611 is the best single issue to anchor the cluster because it is general, clearly phrased, and closest to the recurring loading-regression pattern seen across the set.", + "best_issue_reason": "`issue:43611` is the most suitable anchor issue because it is broad, clearly regression-oriented, and centrally aligned with the repeated v5 loading breakages in this set.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45684", "right": "issue:45698", "accept": false, - "reason": "Both involve custom save/load behavior, but one is about permissions propagation and the other about loading the wrong custom module; different bugs." + "reason": "Both involve custom-model save/load, but one is about read-only permissions during `save_pretrained` and the other is about `from_pretrained` selecting the wrong custom module. Different bugs." }, { "left": "issue:44479", "right": "issue:44877", "accept": false, - "reason": "Both are v5 regressions in model/config loading, but they affect different models and different failure modes." + "reason": "Different models and failures: Qwen VL video-input regression vs strict config blocking `granite_speech` loading." }, { "left": "issue:45588", "right": "issue:45663", "accept": false, - "reason": "Flash-attention sink handling and Gemma4 FSDP shared_kv_states are unrelated code paths." + "reason": "`s_aux=None` crash in flash attention for sink-less models is unrelated to Gemma4 FSDP2 `shared_kv_states` rebuilding." }, { "left": "issue:43646", "right": "issue:43708", "accept": false, - "reason": "Custom model initialization in notebooks versus Trainer resume/max_steps calculation are separate problems." + "reason": "Custom model initialization breakage and `resume_from_checkpoint` max-step miscalculation are separate code paths." }, { "left": "issue:43645", "right": "issue:44164", "accept": false, - "reason": "Notebook class-definition init issues and extra_state serialization/loading are not the same bug." + "reason": "Notebook custom model initialization and `extra_state` save/load handling are different issues." }, { "left": "issue:43828", "right": "issue:43957", "accept": false, - "reason": "Both are dtype/device mismatch reports, but they target different mechanisms and model families." + "reason": "Phi-tiny-MoE autocast dtype mismatch and meta-device loading breakage are different runtime problems." }, { "left": "issue:45292", "right": "issue:45335", "accept": true, - "reason": "These describe the same embedding-resize propagation bug, with 45335 being the t5gemma-specific instance of the broader resize_token_embeddings issue." + "reason": "Same underlying bug: `resize_token_embeddings` fails to update output embeddings; `t5gemma` is just a specific manifestation affecting `decoder.embed_tokens`." }, { - "left": "issue:43425", - "right": "issue:43828", + "left": "issue:44402", + "right": "issue:44451", "accept": false, - "reason": "Torch version incompatibility and an autocast dtype mismatch are different issues." + "reason": "The titles both concern tokenizer/model loading, but they do not point to the same concrete failure or shared code path with enough evidence." }, { - "left": "issue:44402", - "right": "issue:44451", + "left": "issue:43425", + "right": "issue:43828", "accept": false, - "reason": "A tokenizer vocab-size mismatch for one model is not clearly the same as a generic load failure for another model." + "reason": "Torch 2.10 compatibility is a different problem from the Phi-tiny-MoE autocast dtype mismatch." }, { "left": "issue:43257", "right": "issue:44863", "accept": false, - "reason": "Both are model-loading failures, but they concern different architectures and different failure causes." + "reason": "Qwen3 MOE weight conversion with accelerate+deepspeed and NemotronH checkpoint loading are different model-specific loading bugs." }, { "left": "issue:44683", "right": "issue:45823", "accept": false, - "reason": "Compiled flex_attention on torch>=2.9 and Gemma4 device-mapping mismatch are unrelated." + "reason": "Compiled flex_attention on torch >=2.9 is unrelated to Gemma4 PLE device mismatch during forward." }, { "left": "issue:44589", "right": "issue:44756", "accept": false, - "reason": "Float8 storage lookup failure and mmap/OOM on Strix Halo are unrelated runtime issues." + "reason": "Missing Float8 storage support and disabling mmap to avoid OOM are unrelated failure modes." }, { "left": "issue:39404", "right": "issue:45701", "accept": false, - "reason": "Whisper pipeline return_language and tokenizer-behavior changes are different regressions." + "reason": "Whisper pipeline `return_language` regression and tokenization changes are not the same bug." }, { "left": "issue:43425", "right": "issue:45823", "accept": false, - "reason": "Torch 2.10 incompatibility is not the same as Gemma4 PLE device mismatch." + "reason": "General torch version incompatibility is not the same concrete issue as Gemma4 PLE device mismatch." }, { - "left": "issue:43611", - "right": "issue:43645", + "left": "issue:43425", + "right": "issue:43716", "accept": false, - "reason": "Both mention custom model loading, but one is base_model_prefix handling and the other is notebook redefinition/initialization." + "reason": "Torch 2.10 incompatibility and Mistral-3 image preprocessor dtype mismatch are different problems." }, { - "left": "issue:43425", - "right": "issue:43716", + "left": "issue:43611", + "right": "issue:43645", "accept": false, - "reason": "Generic torch-version incompatibility and Mistral-3 dtype mismatch are distinct." + "reason": "Both are v5 regressions, but one is generic model loading with `base_model_prefix` while the other is custom model initialization in notebooks." }, { "left": "issue:44530", "right": "issue:45084", "accept": false, - "reason": "PagedAttentionCache linear_attention failure and non-template-node compilation are different code paths." + "reason": "PagedAttentionCache `linear_attention` crash and non-template-node compilation failure are different attention/compiler issues." }, { "left": "issue:43906", "right": "issue:44530", "accept": false, - "reason": "The isolated reproduction reference is too vague to treat as the same concrete bug as the Qwen3.5 cache crash." + "reason": "An isolated reproduction of another issue is not the same as the Qwen3.5 PagedAttentionCache crash." }, { "left": "issue:39404", "right": "issue:44912", "accept": false, - "reason": "Whisper return_language and MXFP4 quantization fallback are unrelated." + "reason": "Whisper pipeline language regression and MXFP4 quantization fallback are unrelated." }, { "left": "issue:44265", "right": "issue:44898", "accept": false, - "reason": "torch.export with torch_compilable_check and Perceiver resize/interpolation failure are not the same defect." + "reason": "`torch.export.export` failing with `torch_compilable_check` is a different path from Perceiver non-default-resolution handling." }, { "left": "issue:44479", "right": "issue:44805", "accept": false, - "reason": "A video-input regression and a mask shape mismatch are too different to merge." + "reason": "A Qwen VL video regression is not the same as the mask-shape indexing error." }, { "left": "issue:43901", "right": "issue:44509", "accept": false, - "reason": "Both are docs mismatches, but they cover different deprecated/removed pipeline surfaces." + "reason": "Both are docs-related, but they concern different pipeline/task deprecations and should not be merged as one bug." }, { "left": "issue:43994", "right": "issue:44079", "accept": false, - "reason": "Nonsensical outputs for SigLIP2 and ModelOutput key assignment are unrelated." + "reason": "Siglip2 producing bad outputs and `ModelOutput` key assignment when a previous key is `None` are different issues." + }, + { + "left": "issue:43425", + "right": "issue:44918", + "accept": false, + "reason": "Torch compatibility and TRL SFT embedding unpacking are unrelated." }, { "left": "issue:44589", "right": "issue:45084", "accept": false, - "reason": "Storage-object lookup and compilation-template-node errors are unrelated." + "reason": "Float8 storage lookup failure and compilation of non-template nodes are separate failures." }, { "left": "issue:43299", "right": "issue:43931", - "accept": true, - "reason": "Both point to the same Qwen3-VL MoE loading regression, with 43931 being a specific shape-mismatch manifestation of the broader 43299 issue." + "accept": false, + "reason": "Both concern Qwen3-VL loading, but one is a dev0 regression and the other is a checkpoint shape mismatch; not enough evidence of the same underlying bug." }, { "left": "issue:43596", "right": "issue:44683", "accept": false, - "reason": "Deepspeed zero3 init index error and compiled flex_attention failure are unrelated." + "reason": "Deepspeed ZeRO-3/BertModel indexing and compiled flex_attention on torch >=2.9 are unrelated." }, { "left": "issue:39404", "right": "issue:44671", "accept": false, - "reason": "Whisper pipeline behavior and CamemBERT masked-LM predictions are unrelated." + "reason": "Whisper pipeline behavior and CamemBERT masked-LM predictions are different model-specific regressions." }, { "left": "issue:44164", "right": "issue:45759", "accept": false, - "reason": "extra_state save/load handling and multimodal config unwrapping are separate model-loading bugs." + "reason": "`extra_state` save/load handling and `from_config` not unwrapping `text_config` are different model-loading behaviors." }, { "left": "issue:36032", "right": "issue:43452", "accept": false, - "reason": "Tokenizer special-token method conflict and gguf_file loading breakage are different failures." + "reason": "Tokenizer method-name conflict and `gguf_file` handling for tokenizer/model loading are different bugs." }, { "left": "issue:44671", "right": "issue:44991", "accept": false, - "reason": "CamemBERT prediction drift and tokenizer loading failure for a different model are not the same issue." + "reason": "CamemBERT prediction regression and tokenizer loading failure for EMBEDDIA/est-roberta are not the same issue." }, { - "left": "issue:43856", - "right": "issue:44756", + "left": "issue:32090", + "right": "issue:39404", "accept": false, - "reason": "Qwen3 MoE training memory usage and Strix Halo mmap OOM are different performance/memory problems." + "reason": "Trainer GPU broadcast `NoneType` error and Whisper `return_language` pipeline regression are unrelated." } ] }, "evaluator_result": { "accept": true, - "feedback": "The cluster summary stays grounded in the packet: it treats most pairs as thematically similar but not duplicates, and the two accepted pairs (45292/45335 and 43299/43931) are plausible same-bug/broader-instance matches. The anchor choice around 43611 is conservative and fits the dominant v5 loading/initialization theme." + "feedback": "The selection and summary are grounded in the packet: issue:43611 is a defensible broad representative for the repeated v5 model-loading regressions, and the lone accepted soft pair (issue:45292 / issue:45335) looks like the same resize_token_embeddings bug at a model-specific level. The rejected pairs are stated conservatively, with no clear overclaiming or unsupported mergeability." }, "evaluator_used": true, "retried": false, @@ -2558,7 +2516,7 @@ "component_index": 1, "component_count": 17, "review_unit_index": 8, - "review_unit_count": 50, + "review_unit_count": 49, "status": "reviewed", "reason": null, "source": "llm", @@ -2566,10 +2524,8 @@ "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-426", "nodes": [ - "issue:32090", "issue:34567", "issue:36032", - "issue:39404", "issue:40444", "issue:40990", "issue:42915", @@ -2589,6 +2545,7 @@ "issue:43645", "issue:43646", "issue:43716", + "issue:43856", "issue:43883", "issue:44164", "issue:44206", @@ -2602,6 +2559,7 @@ "issue:44661", "issue:44683", "issue:44737", + "issue:44756", "issue:44811", "issue:44849", "issue:44987", @@ -2615,12 +2573,12 @@ "issue:45987" ], "soft_pairs": [ - "issue:32090|issue:39404", "issue:44991|issue:45081", + "issue:43856|issue:44756", "issue:44164|issue:44479", "issue:44737|issue:45859", - "issue:43582|issue:43632", "issue:44661|issue:44683", + "issue:43582|issue:43632", "issue:44488|issue:45081", "issue:44488|issue:44987", "issue:43611|issue:43644", @@ -2638,253 +2596,243 @@ "issue:44464|issue:45987", "issue:43644|issue:43646", "issue:43452|issue:43482", + "issue:44849|issue:45325", "issue:36032|issue:43645", "issue:43317|issue:43482", - "issue:44849|issue:45325", "issue:44987|issue:44991", "issue:45084|issue:45507", "issue:43883|issue:44991", "issue:43262|issue:43377", "issue:43089|issue:45987", - "issue:43645|issue:44534", - "issue:43646|issue:44534" + "issue:43645|issue:44534" ], - "prepared_review_unit_hash": "390618722a8c13f66bc81ac7c32a257abfa4f9f901bed2d480c8f09b488fd82b", + "prepared_review_unit_hash": "557ab533fdfcc7ae3e096ad04e5a873b35329a28fe70599a929f573847b31618", "component_budget": { "node_count": 426, "item_count": 426, - "soft_pair_count": 1572, - "serialized_chars": 814130, - "estimated_input_tokens": 203533, - "estimated_eval_tokens": 407322 + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 }, "budget": { "node_count": 47, "item_count": 47, - "soft_pair_count": 33, - "serialized_chars": 37024, - "estimated_input_tokens": 9256, - "estimated_eval_tokens": 18768 + "soft_pair_count": 32, + "serialized_chars": 36655, + "estimated_input_tokens": 9164, + "estimated_eval_tokens": 18584 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This set is mostly a collection of adjacent Transformers regressions rather than true duplicates. The only pair that looks like the same underlying bug is the GGUF/from_pretrained loader regression; the rest are model- or feature-specific issues with similar symptoms but different code paths.", - "confidence": 0.73, - "canonical_issue_reason": "issue:43452 is the broader, more general report: it names the shared gguf_file/from_pretrained failure path and covers both AutoTokenizer and AutoModelForCausalLM. That makes it a better canonical duplicate target than the more model-specific issue:43482.", + "summary": "This set is mostly a grab bag of unrelated issue reports that share broad Transformers/version/model vocabulary but not the same underlying bug. The soft pairs are all false positives and should be kept separate.", + "confidence": 0.87, + "canonical_issue_reason": "No single canonical duplicate stands out: the items cover many distinct regressions across different models, loaders, and runtime paths.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43452 is the best overall representative because it is the most general and concrete duplicate target in the cluster, with the clearest shared failure mode and the strongest fit to the one accepted soft pair.", + "best_issue_reason": "Issue 43611 is the broadest representative of the Transformers v5 regression family here, but it is still only one specific loading bug and should not absorb the others.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:32090", - "right": "issue:39404", + "left": "issue:44991", + "right": "issue:45081", "accept": false, - "reason": "Different subsystems and failures: Trainer GPU broadcast TypeError vs Whisper pipeline return_language behavior." + "reason": "Both involve tokenizer/model loading regressions, but one is a tokenizer failure for EMBEDDIA/est-roberta and the other is a Mistral regex patch crash; different root causes and code paths." }, { - "left": "issue:44991", - "right": "issue:45081", + "left": "issue:43856", + "right": "issue:44756", "accept": false, - "reason": "Both are tokenizer-loading regressions, but they affect different models and appear to come from different loader paths." + "reason": "One is MoE training memory inefficiency; the other is mmap-related OOM on Strix Halo. Same broad area, not the same bug." }, { "left": "issue:44164", "right": "issue:44479", "accept": false, - "reason": "Unrelated bugs: save/from_pretrained extra_state handling vs Qwen2.5-VL video input regression." + "reason": "save/from_pretrained extra_state handling is unrelated to the Qwen video input regression." }, { "left": "issue:44737", "right": "issue:45859", "accept": false, - "reason": "Both involve device placement, but one is XLNet CPU work on every forward and the other is Qwen3_5Moe CPU offload compatibility." + "reason": "Both mention device placement, but XLNet CPU arange and Qwen3.5 MoE CPU-offload rotary embedding failures are different code paths." }, { - "left": "issue:43582", - "right": "issue:43632", + "left": "issue:44661", + "right": "issue:44683", "accept": false, - "reason": "Apple Silicon allocator TypeError and the _is_hf_initialized flag regression are unrelated." + "reason": "Tokenizer mapping/add-new-model-like logic is unrelated to compiled flex_attention failing on torch >= 2.9." }, { - "left": "issue:44661", - "right": "issue:44683", + "left": "issue:43582", + "right": "issue:43632", "accept": false, - "reason": "Different areas: tokenizer-mapping model registration vs compiled flex_attention on torch>=2.9." + "reason": "Apple Silicon allocator warmup TypeError and the _is_hf_initialized flag regression are separate issues." }, { "left": "issue:44488", "right": "issue:45081", "accept": false, - "reason": "Both mention loading failures, but one is a generic model load failure and the other is a specific Mistral regex patch crash." + "reason": "Both are loading failures, but one is a model/tokenizer load issue for a specific repo and the other is a Mistral tokenizer regex-patching crash." }, { "left": "issue:44488", "right": "issue:44987", "accept": false, - "reason": "Same broad symptom class, but different models and no evidence of the same code path." + "reason": "These are both model-loading failures, but they affect different models and likely different breakpoints in the load pipeline." }, { "left": "issue:43611", "right": "issue:43644", "accept": false, - "reason": "Both are v5 regressions, but one is base_model_prefix loading and the other is non-persistent buffer initialization." + "reason": "Both are Transformers v5 regressions, but one is base_model_prefix loading and the other is non-persistent buffer initialization; not the same bug." }, { "left": "issue:40990", "right": "issue:43421", "accept": false, - "reason": "Perplexity on a benchmark and tokenizer post-processor updates are unrelated." + "reason": "Perplexity on gpt-oss-20b is unrelated to runtime post-processor updates when special tokens change." }, { "left": "issue:43716", "right": "issue:45084", "accept": false, - "reason": "Different failures: Mistral image preprocessor dtype mismatch vs template compilation TypeError." + "reason": "Image preprocessor dtype mismatch and template compilation errors are different subsystems with different failure modes." }, { "left": "issue:44206", "right": "issue:44220", "accept": false, - "reason": "Both are audio-feature related, but the titles point to different failure modes and not the same concrete bug." + "reason": "Both are audio-related, but one is an unsupported center arg in a feature extractor and the other is a low-level fbank extraction issue." }, { "left": "issue:36032", "right": "issue:42915", "accept": false, - "reason": "Tokenizer special-token conflict and Qwen3Moe FP8 config failure are unrelated." + "reason": "T5 tokenizer add_special_tokens conflict and Qwen3Moe FP8 config failure are unrelated." }, { "left": "issue:43643", "right": "issue:44530", "accept": false, - "reason": "AutoConfig trust_remote_code missing fields and PagedAttentionCache linear_attention crash are different bugs." + "reason": "AutoConfig trust_remote_code field loss is not the same as PagedAttentionCache rejecting linear_attention groups." }, { "left": "issue:43883", "right": "issue:44488", "accept": false, - "reason": "Molmo tied-weights metadata and a model-loading failure for a different repo do not match." + "reason": "Missing all_tied_weights_keys in Molmo is unrelated to a model-loading failure for cjvt/sleng-bert." }, { "left": "issue:34567", "right": "issue:40444", "accept": false, - "reason": "TrainerState token counting and Qwen2.5-VL multi-image IterableDataset failure are unrelated." + "reason": "TrainerState token counting and Qwen2.5-VL iterable-dataset multi-image finetuning are distinct bugs." }, { "left": "issue:44811", "right": "issue:45584", "accept": false, - "reason": "Both involve Whisper, but one is batch_decode skip_special_tokens handling and the other is empty-transcription generation after align_special_tokens." + "reason": "Both involve Whisper decoding, but one ignores skip_special_tokens and the other fails on empty transcription after align_special_tokens." }, { "left": "issue:43541", "right": "issue:45084", "accept": false, - "reason": "Torch dynamo grouped_mm tracing failure is unrelated to template compilation errors." + "reason": "Mixtral grouped_mm tracing errors and non-template-node compilation errors are not the same underlying issue." }, { "left": "issue:44451", "right": "issue:44987", "accept": false, - "reason": "Both are model-loading regressions, but they concern different repositories and likely different root causes." + "reason": "Different models and different loading failures; the shared symptom of 'cannot load' is too broad." }, { "left": "issue:42915", "right": "issue:44530", "accept": false, - "reason": "Qwen3Moe FineGrainedFP8Config and Qwen3.5 PagedAttentionCache failures are different code paths." + "reason": "Both are Qwen-family regressions, but FP8 config failure and invalid group type in PagedAttentionCache are separate code paths." }, { "left": "issue:44464", "right": "issue:45987", "accept": false, - "reason": "Chunked generation inconsistency and StaticCache seq_length return type are not the same bug." + "reason": "Chunked generation inconsistency with compiled forward is unrelated to StaticCache.get_seq_length returning the wrong type." }, { "left": "issue:43644", "right": "issue:43646", "accept": false, - "reason": "Both are Transformers v5 regressions, but one is buffer initialization and the other is custom model initialization." + "reason": "Both are v5 regressions around model initialization, but non-persistent buffer junk and custom-model init breakage are different concrete defects." }, { "left": "issue:43452", "right": "issue:43482", - "accept": true, - "reason": "These look like the same GGUF/from_pretrained loader regression, with 43482 being a specific model instance of the broader 43452 failure path." + "accept": false, + "reason": "GGUF-related loading failures, but one is a generic gguf_file/from_pretrained issue and the other is a specific Qwen2.5-GGUF v5 compatibility break." + }, + { + "left": "issue:44849", + "right": "issue:45325", + "accept": false, + "reason": "Qwen3.5 output_hidden_states and Qwen2.5-VL rope index scaling are unrelated model-specific bugs." }, { "left": "issue:36032", "right": "issue:43645", "accept": false, - "reason": "Tokenizer special-token method conflict and Jupyter custom-model initialization are unrelated." + "reason": "Tokenizer special-token method conflicts and notebook custom-model initialization failures do not share the same underlying defect." }, { "left": "issue:43317", "right": "issue:43482", "accept": false, - "reason": "GPU/CPU offload loading failure and GGUF loading failure are different mechanisms." - }, - { - "left": "issue:44849", - "right": "issue:45325", - "accept": false, - "reason": "Different model families and different bugs: hidden-states output vs RoPE scaling for still images." + "reason": "gpu+cpu offload dequantized-model loading and Qwen2.5-GGUF loading are different failure modes." }, { "left": "issue:44987", "right": "issue:44991", "accept": false, - "reason": "Both are load failures on newer transformers, but they affect different models and the titles do not indicate the same root cause." + "reason": "Both are Transformers>=5 loading regressions, but they affect different artifacts and appear to stem from different loader paths." }, { "left": "issue:45084", "right": "issue:45507", "accept": false, - "reason": "Template compilation and GraniteMoEHybrid invalid method calls are different issues." + "reason": "Template compilation and GraniteMoEHybrid invalid-method calls are unrelated." }, { "left": "issue:43883", "right": "issue:44991", "accept": false, - "reason": "Missing tied-weights metadata and tokenizer loading failure are unrelated." + "reason": "Molmo weight-key attribute handling is unrelated to tokenizer loading for est-roberta." }, { "left": "issue:43262", "right": "issue:43377", "accept": false, - "reason": "Audio chat-template sampling-rate defaulting and MIMI batched-vs-single padding-mask behavior are different bugs." + "reason": "Audio chat-template sampling-rate defaults and MIMI batched-vs-single output differences are separate bugs." }, { "left": "issue:43089", "right": "issue:45987", "accept": false, - "reason": "Generation performance overhead and cache seq_length return-type mismatch are unrelated." + "reason": "Generation performance overhead and StaticCache length type contract are not the same issue." }, { "left": "issue:43645", "right": "issue:44534", "accept": false, - "reason": "Custom model initialization in notebooks and non-persistent buffer junk are different regressions." - }, - { - "left": "issue:43646", - "right": "issue:44534", - "accept": false, - "reason": "Custom model initialization and non-persistent buffer filling are not the same underlying bug." + "reason": "Both are v5 regressions, but Jupyter custom-model initialization and non-persistent buffer junk are distinct bugs." } ] }, - "evaluator_result": { - "accept": true, - "feedback": "The verdicts are conservative and grounded in the packet. The canonical choice of issue:43452 is well supported by its broader title, and the single accepted soft pair (43452/43482) is a plausible shared GGUF/from_pretrained regression. The remaining pair rejections stay within what can be inferred from titles alone and do not overclaim duplicate evidence." - }, - "evaluator_used": true, + "evaluator_result": null, + "evaluator_used": false, "retried": false, - "accepted_nontrivial_soft_edge": true, + "accepted_nontrivial_soft_edge": false, "error_kind": null, "error_message": null }, @@ -2893,7 +2841,7 @@ "component_index": 1, "component_count": 17, "review_unit_index": 9, - "review_unit_count": 50, + "review_unit_count": 49, "status": "reviewed", "reason": null, "source": "llm", @@ -2901,7 +2849,6 @@ "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-426", "nodes": [ - "issue:33666", "issue:36032", "issue:40444", "issue:40990", @@ -2919,6 +2866,7 @@ "issue:43582", "issue:43611", "issue:43632", + "issue:43646", "issue:43742", "issue:43792", "issue:43819", @@ -2952,10 +2900,11 @@ ], "soft_pairs": [ "issue:43404|issue:44466", - "issue:43322|issue:43540", + "issue:43646|issue:44534", "issue:44403|issue:45081", - "issue:43994|issue:44610", + "issue:43322|issue:43540", "issue:42915|issue:45084", + "issue:43994|issue:44610", "issue:43377|issue:43819", "issue:44220|issue:44683", "issue:44265|issue:44610", @@ -2963,256 +2912,255 @@ "issue:44987|issue:45081", "issue:42915|issue:44291", "issue:45216|issue:45357", + "issue:44062|issue:44987", "issue:43632|issue:44756", "issue:36032|issue:43742", - "issue:44062|issue:44987", "issue:42915|issue:44589", "issue:44479|issue:45325", "issue:43404|issue:45127", "issue:43296|issue:43531", + "issue:36032|issue:41720", "issue:40444|issue:45325", "issue:43531|issue:44877", - "issue:36032|issue:41720", "issue:43611|issue:44534", "issue:43425|issue:44403", "issue:40990|issue:44811", - "issue:43632|issue:44534", "issue:43792|issue:44589", + "issue:43632|issue:44534", "issue:44079|issue:45663", - "issue:44683|issue:45020", "issue:41553|issue:43582", - "issue:42175|issue:45542", - "issue:33666|issue:40444" + "issue:44683|issue:45020", + "issue:42175|issue:45542" ], - "prepared_review_unit_hash": "fc88df72f136227b77a933bb4d569581e4f6e97014ca64103b05919e566d7180", + "prepared_review_unit_hash": "cc463fac0119b0f097d1a1e2a574524cc09941fd3e5fdd49b2f8f667f970447d", "component_budget": { "node_count": 426, "item_count": 426, - "soft_pair_count": 1572, - "serialized_chars": 814130, - "estimated_input_tokens": 203533, - "estimated_eval_tokens": 407322 + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 32, - "serialized_chars": 37179, - "estimated_input_tokens": 9295, - "estimated_eval_tokens": 18846 + "serialized_chars": 37206, + "estimated_input_tokens": 9302, + "estimated_eval_tokens": 18860 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a grab-bag of distinct issue reports with broad subsystem overlap, especially around Qwen/VL and v5 regressions. I\u2019d keep only one true duplicate-style pair: the Qwen3.5 save_pretrained regression reports. The Qwen VL temporal-position/video pair should be treated conservatively as adjacent similarity, not a confirmed duplicate.", - "confidence": 0.86, - "canonical_issue_reason": "issue:45216 is the best canonical issue because it describes the broader Qwen3.5 save_pretrained regression, while issue:45357 is a narrower symptom report about incorrect visual encoder keys.", + "summary": "This cluster is mostly a grab-bag of unrelated Transformers bug reports, many of them v5 regression/loading issues. Only the Qwen3.5 save_pretrained reports look like a likely duplicate pair; the rest are too different in model, failure mode, or code path to merge.", + "confidence": 0.68, + "canonical_issue_reason": "issue:43646 is the broadest representative of the v5 regression/theme here: a general loading/initialization breakage that fits the cluster better than the more model-specific reports.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45216 is the strongest global issue anchor in this set: it is concrete, clearly regression-focused, and has the cleanest nearby follow-up in the cluster.", + "best_issue_reason": "issue:43646 is the best single triage representative because it is broad, clear, and central to the repeated Transformers v5 breakage pattern in this set.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43404", "right": "issue:44466", "accept": false, - "reason": "Same general tied-weights theme, but different concrete failures: Mistral3 lm_head tying vs Qwen3.5 save_pretrained serialization on device." + "reason": "Related tied-weights symptoms, but one is model init tying and the other is save_pretrained serialization; different code paths." }, { - "left": "issue:43322", - "right": "issue:43540", + "left": "issue:43646", + "right": "issue:44534", "accept": false, - "reason": "Both are multimodal failures, but one is a Llava Next loading segfault and the other is a Qwen3OmniMoe video-processing ValueError." + "reason": "Both are v5 regressions, but custom model initialization and non-persistent buffer junk are distinct bugs." }, { "left": "issue:44403", "right": "issue:45081", "accept": false, - "reason": "Unrelated bugs: generic loading noise vs a specific Mistral tokenizer regex patch crash." + "reason": "Unrelated failures: noisy loading message vs Mistral tokenizer regex patch crash." }, { - "left": "issue:43994", - "right": "issue:44610", + "left": "issue:43322", + "right": "issue:43540", "accept": false, - "reason": "Different models and symptoms: SigLIP2 bad outputs vs OmDet-Turbo processor/model input-size mismatch." + "reason": "Different multimodal failures on different models and inputs; not the same underlying bug." }, { "left": "issue:42915", "right": "issue:45084", "accept": false, - "reason": "One is a Qwen3Moe FP8 config failure; the other is a generic compile/template-node error. No shared code-path is evident." + "reason": "FP8 config failure and template compilation error are unrelated." + }, + { + "left": "issue:43994", + "right": "issue:44610", + "accept": false, + "reason": "Both vision-related, but different models and root causes." }, { "left": "issue:43377", "right": "issue:43819", "accept": false, - "reason": "Both mention mismatches between two execution paths, but they are different models and different mechanisms (padding mask vs STE)." + "reason": "Both audio-related, but different models and distinct defects." }, { "left": "issue:44220", "right": "issue:44683", "accept": false, - "reason": "Both involve torch internals, but one is fbank feature extraction and the other is compiled flex_attention on newer torch." + "reason": "Feature extraction bug vs torch-version flex_attention compile failure; unrelated." }, { "left": "issue:44265", "right": "issue:44610", "accept": false, - "reason": "torch.export with torch_compilable_check and an OmDet-Turbo processor image-size bug are unrelated." + "reason": "torch.export/check logic and processor image-size mismatch are different bugs." }, { "left": "issue:44451", "right": "issue:45081", "accept": false, - "reason": "Different targets and failures: ScandiBERT loading vs Mistral tokenizer backend_tokenizer crash." + "reason": "Different loading problems: ScandiBERT load failure vs Mistral regex crash." }, { "left": "issue:44987", "right": "issue:45081", "accept": false, - "reason": "A broad transformers>=5.1.0 remote_code regression is not the same as the specific Mistral regex patch tokenizer crash." + "reason": "Remote-code loading regression and tokenizer regex crash do not share the same code path." }, { "left": "issue:42915", "right": "issue:44291", "accept": false, - "reason": "Both are v5-related loading issues, but one is FP8 config handling and the other is init_empty_weights/_is_hf_initialized argument handling." + "reason": "Both are v5-related, but the failures are distinct: FP8 config vs unexpected _is_hf_initialized arg." }, { "left": "issue:45216", "right": "issue:45357", "accept": true, - "reason": "Same underlying Qwen3.5 save_pretrained regression: both report incorrect checkpoint contents, with 45357 narrowing it to visual encoder keys." + "reason": "Same Qwen3.5 save_pretrained regression; the later issue narrows the same underlying wrong-checkpoint/keys problem to visual encoder keys." + }, + { + "left": "issue:44062", + "right": "issue:44987", + "accept": false, + "reason": "AddedToken special-argument bug and remote_code loading failure are unrelated." }, { "left": "issue:43632", "right": "issue:44756", "accept": false, - "reason": "Transformers v5 initialization flag breakage is unrelated to disabling mmap on Strix Halo for OOM avoidance." + "reason": "Different problems: _is_hf_initialized flag handling vs mmap/OOM behavior." }, { "left": "issue:36032", "right": "issue:43742", "accept": false, - "reason": "Both are loading errors, but one is a tokenizer attribute conflict and the other is a key error for a specific model." - }, - { - "left": "issue:44062", - "right": "issue:44987", - "accept": false, - "reason": "Tokenizer AddedToken argument duplication is unrelated to a broad remote_code loading regression." + "reason": "Tokenizer special-token conflict and MobileLLM key error are different bugs." }, { "left": "issue:42915", "right": "issue:44589", "accept": false, - "reason": "Different bugs: FP8 config failure versus missing Float8 storage type." + "reason": "Different FP8/storage issues with no shared failing path." }, { "left": "issue:44479", "right": "issue:45325", "accept": false, - "reason": "Related Qwen/VL area, but not enough evidence for the same bug: 44479 is a video-input regression across several Qwen VL variants, while 45325 is a still-image temporal position_id scaling issue in get_rope_index." + "reason": "Both touch Qwen vision models, but video-input regression and still-image rope indexing are different defects." }, { "left": "issue:43404", "right": "issue:45127", "accept": false, - "reason": "Both touch tied weights, but one is Mistral3 lm_head tying and the other is LoRA merge collapse with extended vocabulary." + "reason": "Both involve tied weights, but one is missing tying in the model and the other is LoRA-merge collapse after vocab extension." }, { "left": "issue:43296", "right": "issue:43531", "accept": false, - "reason": "PaddleOCR-VL loading and Qwen3-MoE sliding_window behavior are unrelated model-specific issues." + "reason": "Different model families and different failures." + }, + { + "left": "issue:36032", + "right": "issue:41720", + "accept": false, + "reason": "Tokenizer load conflict and CUDA assert from auto device mapping are unrelated." }, { "left": "issue:40444", "right": "issue:45325", "accept": false, - "reason": "Both involve Qwen VL, but one is an IterableDataset multi-image finetuning failure while the other is a temporal position-id scaling regression." + "reason": "Multiple-image dataset failure and Qwen2.5-VL rope scaling bug are different issues." }, { "left": "issue:43531", "right": "issue:44877", "accept": false, - "reason": "Sliding_window in Qwen3-MoE and strict config loading for granite_speech are separate problems." - }, - { - "left": "issue:36032", - "right": "issue:41720", - "accept": false, - "reason": "Tokenizer loading conflict and Qwen3 auto device mapping cudaErrorAssert are unrelated." + "reason": "Sliding-window bug and granite_speech config loading are unrelated." }, { "left": "issue:43611", "right": "issue:44534", "accept": false, - "reason": "Base-model-prefix loading breakage and non-persistent-buffer junk initialization are different v5 regressions." + "reason": "Base-model-prefix loading regression and junk-filled non-persistent buffers are different v5 bugs." }, { "left": "issue:43425", "right": "issue:44403", "accept": false, - "reason": "Torch 2.10 incompatibility is not the same as generic transformer loading noise." + "reason": "Torch compatibility issue and loading-noise issue are unrelated." }, { "left": "issue:40990", "right": "issue:44811", "accept": false, - "reason": "Perplexity on gpt-oss-20b and Whisper processor batch_decode skipping special tokens are unrelated." - }, - { - "left": "issue:43632", - "right": "issue:44534", - "accept": false, - "reason": "Both are v5-related, but one is the _is_hf_initialized flag and the other is non-persistent buffer initialization." + "reason": "Perplexity regression and Whisper batch_decode skip_special_tokens bug are unrelated." }, { "left": "issue:43792", "right": "issue:44589", "accept": false, - "reason": "Whisper runtime failure and missing Float8 storage type are unrelated." + "reason": "Whisper runtime failure and Float8 storage lookup error are different code paths." }, { - "left": "issue:44079", - "right": "issue:45663", + "left": "issue:43632", + "right": "issue:44534", "accept": false, - "reason": "ModelOutput key assignment and Gemma4 shared_kv_states/FSDP2 KeyError are distinct bugs." + "reason": "Distinct v5 regressions: _is_hf_initialized handling vs buffer serialization." }, { - "left": "issue:44683", - "right": "issue:45020", + "left": "issue:44079", + "right": "issue:45663", "accept": false, - "reason": "Compiled flex_attention on torch>=2.9 and remote_code loading regressions are separate issues." + "reason": "ModelOutput key assignment and Gemma4 FSDP2 attention KeyError are unrelated." }, { "left": "issue:41553", "right": "issue:43582", "accept": false, - "reason": "Bad AutoTokenizer error for Voxtral and Apple Silicon caching_allocator_warmup TypeError are unrelated." + "reason": "Bad AutoTokenizer error message and AppleSilicon warmup TypeError are unrelated." }, { - "left": "issue:42175", - "right": "issue:45542", + "left": "issue:44683", + "right": "issue:45020", "accept": false, - "reason": "Both mention TensorFlow/backend packaging, but one is missing TF in the torch extra and the other is an undefined tf backend error from tensorboard-only installs; too different to merge." + "reason": "flex_attention compile failure and remote_code loading regression are unrelated." }, { - "left": "issue:33666", - "right": "issue:40444", + "left": "issue:42175", + "right": "issue:45542", "accept": false, - "reason": "Both are Qwen VL training issues, but one is multi-GPU training and the other is iterable-dataset multi-image finetuning." + "reason": "Both concern TensorFlow packaging/backend availability, but they describe different failure paths and are not clearly the same bug." } ] }, "evaluator_result": { - "accept": false, - "feedback": "The writeup is mostly grounded, but one soft-edge verdict is too strong: issue:44479 vs issue:45325 is treated as a true duplicate, yet the packet only supports broad thematic similarity (Qwen/VL, temporal-position logic) rather than clear same-bug evidence. One is a video-input regression across several Qwen VL variants; the other is a still-image temporal position_id scaling bug in get_rope_index. That looks adjacent, not clearly the same defect. The summary should stay at near-duplicate/conservative similarity unless there is stronger direct evidence." + "accept": true, + "feedback": "The summary is grounded in the packet: most items are clearly different bugs by model, failure mode, or code path, and the single accepted Qwen3.5 save_pretrained pair looks plausibly mergeable as one regression. The rejected soft-edge verdicts are generally conservative rather than overclaiming duplicates." }, "evaluator_used": true, - "retried": true, + "retried": false, "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null @@ -3222,7 +3170,7 @@ "component_index": 1, "component_count": 17, "review_unit_index": 10, - "review_unit_count": 50, + "review_unit_count": 49, "status": "reviewed", "reason": null, "source": "llm", @@ -3230,6 +3178,7 @@ "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-426", "nodes": [ + "issue:33666", "issue:35707", "issue:40444", "issue:41720", @@ -3274,251 +3223,253 @@ "issue:45479", "issue:45491", "issue:45538", - "issue:45800", "issue:45812", "issue:45834", "issue:45923" ], "soft_pairs": [ "issue:44403|issue:44991", + "issue:33666|issue:40444", "issue:43421|issue:44568", "issue:35707|issue:40444", "issue:43818|issue:45357", - "issue:43317|issue:43856", "issue:44462|issue:45812", "issue:43381|issue:45923", "issue:44291|issue:44589", "issue:43482|issue:45538", + "issue:43317|issue:43856", "issue:44829|issue:45479", "issue:43482|issue:43525", "issue:44079|issue:44534", - "issue:43377|issue:45491", "issue:43531|issue:44291", + "issue:43377|issue:45491", "issue:41720|issue:45081", "issue:40444|issue:45381", "issue:43749|issue:44079", "issue:44265|issue:45072", "issue:43644|issue:44079", "issue:42175|issue:43296", - "issue:43632|issue:45072", "issue:45325|issue:45381", + "issue:43632|issue:45072", "issue:43792|issue:43957", "issue:43611|issue:43646", "issue:43994|issue:44898", "issue:45325|issue:45834", "issue:43454|issue:44466", - "issue:42915|issue:43482", - "issue:43296|issue:45800" + "issue:42915|issue:43482" ], - "prepared_review_unit_hash": "a249aa08dc3b992c1ef7cbbd0a1cf051309a3a78e3fc4110b92e93168d1901c3", + "prepared_review_unit_hash": "1bbb2b861515981d502d518182f5c5f71915a50198214d3202a05ffaa12b5439", "component_budget": { "node_count": 426, "item_count": 426, - "soft_pair_count": 1572, - "serialized_chars": 814130, - "estimated_input_tokens": 203533, - "estimated_eval_tokens": 407322 + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 29, - "serialized_chars": 36646, - "estimated_input_tokens": 9162, - "estimated_eval_tokens": 18580 + "serialized_chars": 36625, + "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 grab-bag of unrelated closed issues, though many share a broad Transformers v5 regression theme around loading, initialization, tokenization, and multimodal model behavior. None of the soft-edge pairs are strong enough to treat as true duplicates.", - "confidence": 0.81, - "canonical_issue_reason": "issue:43632 is the best central representative because it describes a foundational Transformers v5 initialization regression (_is_hf_initialized) that sits near the middle of several other v5 loading/custom-model breakages in this cluster.", + "summary": "This cluster is mostly a set of unrelated issue reports. The only plausible duplicate-like pair is the Transformers v5 model-initialization/loading regression pair (43611/43646); the rest are distinct bugs across different models, features, and failure modes.", + "confidence": 0.68, + "canonical_issue_reason": "Issue 43646 is the broadest and most central of the near-duplicate loading regressions: it describes a Transformers v5 custom model initialization break, which likely subsumes the more specific base_model_prefix loading symptom in 43611.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43632 is the most suitable global issue to anchor the cluster: it is broad, concrete, and closer to a root-cause-style regression than the more model-specific reports.", + "best_issue_reason": "43646 is the best representative issue in this set because it captures the underlying v5 model initialization regression at a higher level than the narrower 43611 report.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44403", "right": "issue:44991", "accept": false, - "reason": "Both involve loading/tokenizer behavior, but one is about noisy model loading and the other about AutoTokenizer ignoring tokenizer.json for a specific repo. Different failures." + "reason": "Both involve tokenizer/model loading, but one is generic loading noise and the other is a specific tokenizer failure for EMBEDDIA/est-roberta; different bugs." + }, + { + "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 IterableDataset/multi-image finetuning; not the same failure mode." }, { "left": "issue:43421", "right": "issue:44568", "accept": false, - "reason": "One is about updating the backend post-processor after runtime special-token changes; the other is about add_special_tokens not inserting BOS/EOS. Related area, different bug." + "reason": "Both touch special-token handling, but one is a runtime post-processor update issue and the other is add_special_tokens not inserting BOS/EOS; distinct code paths." }, { "left": "issue:35707", "right": "issue:40444", "accept": false, - "reason": "Progressive generation with inputs_embeds/past_key_values is unrelated to finetuning Qwen2.5-VL on an IterableDataset with multiple images." + "reason": "Progressive generation with inputs_embeds/past_key_values is unrelated to Qwen2.5-VL multi-image finetuning." }, { "left": "issue:43818", "right": "issue:45357", "accept": false, - "reason": "Video-LLaVA temporal-attention/weight-sharing bug and Qwen3.5 visual-encoder key serialization bug are different model families and code paths." - }, - { - "left": "issue:43317", - "right": "issue:43856", - "accept": false, - "reason": "GPU+CPU offload load failure is not the same as Qwen3 MoE training memory inefficiency." + "reason": "Different models and different defects: missing temporal attention/weight sharing vs incorrect save_pretrained visual encoder keys." }, { "left": "issue:44462", "right": "issue:45812", "accept": false, - "reason": "Both are tokenizer-related, but one is repository tokenizer resolution and the other is wrong token IDs for Granite models; not the same concrete defect." + "reason": "Both are tokenizer-related, but one ignores tokenizer.json while the other produces wrong IDs for Granite; not the same bug." }, { "left": "issue:43381", "right": "issue:45923", "accept": false, - "reason": "Gradient checkpointing in eval mode is distinct from a missing supports_gradient_checkpointing flag on a trust_remote_code model." + "reason": "One is a usage restriction for gradient checkpointing in eval mode; the other is a missing supports_gradient_checkpointing flag on a remote-code model." }, { "left": "issue:44291", "right": "issue:44589", "accept": false, - "reason": "Unexpected _is_hf_initialized argument during init_empty_weights is a different loading-path failure from the Float8 storage object lookup error." + "reason": "Both are loading regressions in v5, but one is an init_empty_weights argument mismatch and the other is a Float8 storage lookup error; different concrete failures." }, { "left": "issue:43482", "right": "issue:45538", "accept": false, - "reason": "GGUF loading failure and CLIPTokenizer model_max_length are unrelated tokenizer/model-loading issues." + "reason": "Completely different areas: Qwen2.5-GGUF loading under v5 vs CLIPTokenizer model_max_length behavior." + }, + { + "left": "issue:43317", + "right": "issue:43856", + "accept": false, + "reason": "Device-map/offload loading failure is unrelated to Qwen3 MoE memory usage during training." }, { "left": "issue:44829", "right": "issue:45479", "accept": false, - "reason": "FlashAttention-3 causing degenerate training is different from a single-label-classification zero-loss problem." + "reason": "Both describe degenerate training behavior, but one is tied to flash_attention_3 and the other to problem_type=num_labels=1 classification loss handling." }, { "left": "issue:43482", "right": "issue:43525", "accept": false, - "reason": "GGUF loading failure is unrelated to Llama4Config missing pad_token_id." + "reason": "Different models and symptoms: GGUF loading failure vs missing pad_token_id on Llama4Config." }, { "left": "issue:44079", "right": "issue:44534", "accept": false, - "reason": "ModelOutput key assignment with None and non-persistent buffer junk are different regressions." + "reason": "ModelOutput key assignment and non-persistent buffer initialization are separate internal issues." }, { - "left": "issue:43377", - "right": "issue:45491", + "left": "issue:43531", + "right": "issue:44291", "accept": false, - "reason": "MIMI batching/padding-mask mismatch and Gemma3 mixed-length NaNs are both padding-adjacent but not the same bug." + "reason": "Qwen3-MoE sliding_window behavior is unrelated to init_empty_weights/_is_hf_initialized constructor loading errors." }, { - "left": "issue:43531", - "right": "issue:44291", + "left": "issue:43377", + "right": "issue:45491", "accept": false, - "reason": "Qwen3-MoE sliding_window behavior and init_empty_weights/_is_hf_initialized loading errors are separate problems." + "reason": "Both involve batching/padding, but one is a MIMI encoder padding-mask bug and the other is a Gemma3 sliding-window/all-padding-window NaN issue." }, { "left": "issue:41720", "right": "issue:45081", "accept": false, - "reason": "Qwen3 auto device mapping cudaErrorAssert and Mistral tokenizer regex patch crash are unrelated." + "reason": "Qwen3 auto device-mapping cudaErrorAssert and Mistral tokenizer regex patch crashes are unrelated." }, { "left": "issue:40444", "right": "issue:45381", "accept": false, - "reason": "Multiple-image finetuning failure and Qwen2.5-VL video position-id bug may be adjacent, but they are not clearly the same concrete failure." + "reason": "Both are Qwen2.5-VL issues, but one is multi-image finetuning and the other is video vision_position_ids; different scenarios and likely different fixes." }, { "left": "issue:43749", "right": "issue:44079", "accept": false, - "reason": "FSDP CPU RAM efficient loading and ModelOutput key assignment are different subsystems and symptoms." + "reason": "FSDP CPU RAM efficient loading and ModelOutput key assignment are unrelated." }, { "left": "issue:44265", "right": "issue:45072", "accept": false, - "reason": "torch.export/exportability with torch_compilable_check is unrelated to dtype mismatches in bfloat16 inference." + "reason": "torch.export/export-time checks and bfloat16 dtype mismatches in inference are distinct problems." }, { "left": "issue:43644", "right": "issue:44079", "accept": false, - "reason": "Non-persistent buffer serialization junk and ModelOutput key assignment are both v5 regressions but not the same defect." + "reason": "Non-persistent buffer junk fill and ModelOutput key assignment are separate regressions." }, { "left": "issue:42175", "right": "issue:43296", "accept": false, - "reason": "Missing TensorFlow in a pip extra and a PaddleOCR-VL load failure in vLLM are unrelated." + "reason": "Packaging/backends issue vs a specific PaddleOCR-VL model loading failure; no shared bug." }, { - "left": "issue:43632", - "right": "issue:45072", + "left": "issue:45325", + "right": "issue:45381", "accept": false, - "reason": "_is_hf_initialized flag regression is not the same as dtype mismatches in bfloat16 inference." + "reason": "Both are Qwen2.5-VL position-id issues, but one is still-image temporal scaling and the other is video vision_position_ids; too different to merge confidently." }, { - "left": "issue:45325", - "right": "issue:45381", + "left": "issue:43632", + "right": "issue:45072", "accept": false, - "reason": "Both concern Qwen2.5-VL position IDs, but one is still-image temporal scaling and the other is video-input vision_position_ids; close, but not clearly the same bug." + "reason": "_is_hf_initialized breakage and bfloat16 dtype mismatch are unrelated Transformers v5 regressions." }, { "left": "issue:43792", "right": "issue:43957", "accept": false, - "reason": "Whisper-large-v2 runtime failure and torch.device('meta') model-loading breakage are different issues." + "reason": "Whisper model runtime failure and torch.device('meta') loading regressions are not the same bug." }, { "left": "issue:43611", "right": "issue:43646", - "accept": false, - "reason": "Base_model_prefix loading regression and custom model initialization regression are related to v5 breakage but not the same concrete root cause." + "accept": true, + "reason": "These look like the same Transformers v5 model-loading/init regression: one reports breakage for models with base_model_prefix, the other for custom model initialization. A single fix to initialization/loading could plausibly address both." }, { "left": "issue:43994", "right": "issue:44898", "accept": false, - "reason": "Siglip2 nonsensical outputs and Perceiver non-default-resolution failure are unrelated model-specific bugs." + "reason": "Different vision-model correctness issues: SigLIP2 nonsensical outputs vs Perceiver non-default resolution failure." }, { "left": "issue:45325", "right": "issue:45834", "accept": false, - "reason": "Qwen2.5-VL position-id scaling and Kosmos2.5 long-OCR index errors are different multimodal failures." + "reason": "Both are vision/sequence-length related, but one is a Qwen2.5-VL position-id scaling bug and the other is a Kosmos2.5 long-OCR index error." }, { "left": "issue:43454", "right": "issue:44466", "accept": false, - "reason": "AyaVision lm_head tying bug and inconsistent lm_head.weight serialization are both weight-tying adjacent, but they are different problems." + "reason": "Both mention lm_head weight handling, but one is a tied-weights bug causing bad generation and the other is device-dependent serialization inconsistency." }, { "left": "issue:42915", "right": "issue:43482", "accept": false, - "reason": "Qwen3Moe FineGrainedFP8Config failure is unrelated to Qwen2.5-GGUF loading in v5." - }, - { - "left": "issue:43296", - "right": "issue:45800", - "accept": false, - "reason": "PaddleOCR-VL load failure in vLLM and torch/transformers incompatibility are separate compatibility issues." + "reason": "Qwen3Moe FineGrainedFP8Config failure and Qwen2.5-GGUF loading failure are unrelated." } ] }, - "evaluator_result": null, - "evaluator_used": false, + "evaluator_result": { + "accept": true, + "feedback": "The result is grounded in the packet and the soft-edge judgments are conservative overall. The single accepted pair, 43611/43646, is plausibly the same Transformers v5 loading/initialization regression, while the other pairs are rejected with appropriately narrow reasoning rather than overstating duplicate evidence." + }, + "evaluator_used": true, "retried": false, - "accepted_nontrivial_soft_edge": false, + "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null }, @@ -3527,7 +3478,7 @@ "component_index": 1, "component_count": 17, "review_unit_index": 11, - "review_unit_count": 50, + "review_unit_count": 49, "status": "reviewed", "reason": null, "source": "llm", @@ -3564,7 +3515,6 @@ "issue:44186", "issue:44265", "issue:44291", - "issue:44466", "issue:44479", "issue:44488", "issue:44534", @@ -3575,23 +3525,24 @@ "issue:44991", "issue:45072", "issue:45325", - "issue:45357", "issue:45584", "issue:45701", "issue:45759", + "issue:45800", "issue:45823", "issue:45920", "issue:45923", "issue:46050" ], "soft_pairs": [ + "issue:43296|issue:45800", "issue:44479|issue:45823", "issue:43701|issue:45325", "issue:44291|issue:44877", "issue:43881|issue:44488", "issue:43606|issue:44265", - "issue:43873|issue:45701", "issue:43262|issue:43746", + "issue:43873|issue:45701", "issue:40990|issue:45584", "issue:34634|issue:35707", "issue:44568|issue:45920", @@ -3602,10 +3553,10 @@ "issue:43454|issue:43550", "issue:39692|issue:40444", "issue:43377|issue:44610", - "issue:43317|issue:43873", "issue:43454|issue:45072", - "issue:43377|issue:44898", + "issue:43317|issue:43873", "issue:43572|issue:45759", + "issue:43377|issue:44898", "issue:43257|issue:45923", "issue:40990|issue:42947", "issue:43322|issue:46050", @@ -3614,230 +3565,232 @@ "issue:41720|issue:43322", "issue:43881|issue:44991", "issue:43296|issue:44877", - "issue:43749|issue:44534", - "issue:44466|issue:45357" + "issue:43749|issue:44534" ], - "prepared_review_unit_hash": "dd6ce5272b9aa53e116146e11ba9c7c7e25b1a59e20e59cd672f912863332b63", + "prepared_review_unit_hash": "fc2a39c1128886c918f362a1d4971d109f3717d049f4a29ceaee7f1ea28e3ca5", "component_budget": { "node_count": 426, "item_count": 426, - "soft_pair_count": 1572, - "serialized_chars": 814130, - "estimated_input_tokens": 203533, - "estimated_eval_tokens": 407322 + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 }, "budget": { - "node_count": 48, - "item_count": 48, + "node_count": 47, + "item_count": 47, "soft_pair_count": 31, - "serialized_chars": 37218, - "estimated_input_tokens": 9305, - "estimated_eval_tokens": 18866 + "serialized_chars": 36541, + "estimated_input_tokens": 9136, + "estimated_eval_tokens": 18528 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The cluster is mostly a set of unrelated transformer issues across loading, quantization/offloading, tokenization, multimodal processors, and training behavior. The soft-similarity edges are too broad to treat as true duplicates, so I reject all of them.", - "confidence": 0.54, - "canonical_issue_reason": "issue:43873 is the broadest representative of the offloading/quantization failure reports in this set and has the most neighboring similarity links.", + "summary": "Most items are distinct model-specific regressions, not duplicates. The only plausible duplicate family is offloading/quantization loading behavior, centered on 43873; the rest span unrelated tokenizer, multimodal, config, export, and compile bugs.", + "confidence": 0.81, + "canonical_issue_reason": "43873 is the broadest and most central issue in the only overlapping area here: offloading/quantization loading failures. It has the strongest discussion/reference signal and can anchor this cluster.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43873 is the best single issue to anchor the cluster because it captures the most general underlying theme: offloading behaving incorrectly under quantization.", + "best_issue_reason": "43873 is the best representative issue because it is open, broad, and closest to the only near-duplicate pair in the set (43317).", "best_pr_reason": null, "soft_edge_verdicts": [ + { + "left": "issue:43296", + "right": "issue:45800", + "accept": false, + "reason": "Different failures: a model/processor loading problem in vLLM vs a general torch/transformers version incompatibility." + }, { "left": "issue:44479", "right": "issue:45823", "accept": false, - "reason": "Different models and different failure modes: video-input regression vs device-mismatch during forward." + "reason": "Both involve multimodal models, but the concrete bugs differ: video-input regression vs device-mismatch during forward." }, { "left": "issue:43701", "right": "issue:45325", "accept": false, - "reason": "Checkpoint key mismatch is unrelated to Qwen2.5-VL rope/position id scaling." + "reason": "Unrelated code paths: checkpoint key mismatch vs Qwen2.5-VL RoPE/position-id scaling." }, { "left": "issue:44291", "right": "issue:44877", "accept": false, - "reason": "`init_empty_weights` argument handling and strict config loading are separate code paths." + "reason": "Different config/load errors: unexpected init arg vs strict config blocking granite_speech loading." }, { "left": "issue:43881", "right": "issue:44488", "accept": false, - "reason": "Both are load failures, but they concern different models and no shared concrete bug is evident." + "reason": "Separate model-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` failing on `torch_compilable_check`." + "reason": "CPU-offload device mismatch and torch.export failure are unrelated bugs." }, { - "left": "issue:43873", - "right": "issue:45701", + "left": "issue:43262", + "right": "issue:43746", "accept": false, - "reason": "Quantization/offloading behavior and tokenization regressions are unrelated." + "reason": "Audio chat-template sampling-rate default and GraniteSpeech PEFT checkpoint loading are different problems." }, { - "left": "issue:43262", - "right": "issue:43746", + "left": "issue:43873", + "right": "issue:45701", "accept": false, - "reason": "Audio chat-template sampling-rate defaults and PEFT checkpoint loading are different problems." + "reason": "Offloading/quantization behavior is unrelated to a tokenizer-version regression." }, { "left": "issue:40990", "right": "issue:45584", "accept": false, - "reason": "Perplexity anomaly and Whisper empty-transcription generation failure do not share the same bug." + "reason": "Different symptoms and subsystems: perplexity regression vs Whisper empty-transcription generation failure." }, { "left": "issue:34634", "right": "issue:35707", "accept": false, - "reason": "Bark voice preset handling is unrelated to progressive generation with `inputs_embeds`/`past_key_values`." + "reason": "Bark voice preset handling and progressive generation with inputs_embeds/past_key_values are unrelated." }, { "left": "issue:44568", "right": "issue:45920", "accept": false, - "reason": "These are both tokenizer regressions, but one is special-token insertion and the other is broader ID mismapping across models." + "reason": "Both are tokenizer regressions, but one is a single tokenizer special-token bug and the other is broader wrong-token-id behavior across multiple models." }, { "left": "issue:35707", "right": "issue:41720", "accept": false, - "reason": "Generation cache handling and Qwen3 device-map asserts are unrelated." + "reason": "Progressive generation bug and Qwen3 device-map CUDA assert are different failures." }, { "left": "issue:44186", "right": "issue:44265", "accept": false, - "reason": "Tokenizer padding/truncation crashes and export-time checks are different issues." + "reason": "Tokenizer padding/truncation crash and torch.export/torch_compilable_check are unrelated." }, { "left": "issue:43606", "right": "issue:43819", "accept": false, - "reason": "Device mismatch during offload and DAC latent consistency are unrelated." + "reason": "Device-mismatch on offload vs DAC latent/STE mismatch are different code paths." }, { "left": "issue:41720", "right": "issue:45823", "accept": false, - "reason": "Both involve device placement, but the concrete models and failures differ." + "reason": "Both mention device_map issues, but they affect different models and likely different underlying bugs." }, { "left": "issue:43454", "right": "issue:43550", "accept": false, - "reason": "Weight tying corruption and SDPA/torch.compile failure are separate bugs." + "reason": "Weight-tying bug and torch.compile/SDPA failure are unrelated." }, { "left": "issue:39692", "right": "issue:40444", "accept": false, - "reason": "SigLIP2 doc/example problems and Qwen2.5-VL multi-image iterable dataset failure are unrelated." + "reason": "Documentation/model-processor mismatch and IterableDataset multi-image fine-tuning failure are distinct." }, { "left": "issue:43377", "right": "issue:44610", "accept": false, - "reason": "Missing padding-mask support in MIMI and OmDet input-size mismatch are unrelated." - }, - { - "left": "issue:43317", - "right": "issue:43873", - "accept": false, - "reason": "These are both offload/quantization themed, but one is a specific dequantized-model load failure and the other is a broader offloading complaint." + "reason": "MIMI padding-mask inconsistency and OmDet-Turbo image-size mismatch are unrelated." }, { "left": "issue:43454", "right": "issue:45072", "accept": false, - "reason": "Tied-weight serialization and bfloat16 dtype mismatches are not the same concrete defect." + "reason": "Different model bugs: tied weights vs dtype mismatch in inference." }, { - "left": "issue:43377", - "right": "issue:44898", - "accept": false, - "reason": "MIMI batching/padding-mask behavior and Perceiver positional interpolation are unrelated." + "left": "issue:43317", + "right": "issue:43873", + "accept": true, + "reason": "Same offloading/quantization loading family: both describe device_map/offload failures when loading quantized or dequantized models, with overlapping accelerate loading behavior." }, { "left": "issue:43572", "right": "issue:45759", "accept": false, - "reason": "Missing `pad_token_idx` and `text_config` unwrapping are different config-loading issues." + "reason": "A missing config field and an AutoModel text_config unwrapping bug are different API regressions." + }, + { + "left": "issue:43377", + "right": "issue:44898", + "accept": false, + "reason": "Padding-mask batch inconsistency and Perceiver resolution/interpolation failure are unrelated." }, { "left": "issue:43257", "right": "issue:45923", "accept": false, - "reason": "MOE weight conversion under accelerate/deepspeed is unrelated to a missing gradient-checkpointing flag." + "reason": "Accelerate+deepspeed MOE conversion issue and missing gradient-checkpointing flag are separate problems." }, { "left": "issue:40990", "right": "issue:42947", "accept": false, - "reason": "High perplexity and ineffective gradient checkpointing are unrelated training issues." + "reason": "High perplexity and ineffective gradient checkpointing are different training/inference bugs." }, { "left": "issue:43322", "right": "issue:46050", "accept": false, - "reason": "Llava loading segfaults and sinq save/load quantization breakage are different failures." + "reason": "Llava loading segfault and sinq save/load quantization breakage are unrelated." }, { "left": "issue:43526", "right": "issue:43883", "accept": false, - "reason": "Label reduction in an image processor and missing tied-weights metadata are unrelated." + "reason": "BeitImageProcessorFast label reduction and Molmo tied-weights attribute error are unrelated." }, { "left": "issue:43454", "right": "issue:43720", "accept": false, - "reason": "Weight tying/serialization and BitNet packed-weight unpacking are separate loader bugs." + "reason": "AyaVision weight tying and BitNet packed-weight unpacking are different loading/model-init bugs." }, { "left": "issue:41720", "right": "issue:43322", "accept": false, - "reason": "Qwen3 device-map asserts and Llava Next segfaults are not the same bug." + "reason": "Both are loading-time failures, but they affect different model families and likely different root causes." }, { "left": "issue:43881", "right": "issue:44991", "accept": false, - "reason": "A specific model load failure and tokenizer loading failure do not match." + "reason": "Model loading failure and tokenizer loading failure are not the same bug." }, { "left": "issue:43296", "right": "issue:44877", "accept": false, - "reason": "PaddleOCR-VL loading in vLLM and strict config rejection for granite_speech are unrelated." + "reason": "vLLM model load failure and strict config rejection are different issues." }, { "left": "issue:43749", "right": "issue:44534", "accept": false, - "reason": "FSDP CPU-efficient loading and non-persistent buffer initialization are different code paths." - }, - { - "left": "issue:44466", - "right": "issue:45357", - "accept": false, - "reason": "General `lm_head.weight` serialization inconsistency and Qwen3.5 visual-encoder key saving are only loosely related." + "reason": "FSDP CPU-efficient loading breakage and non-persistent buffer initialization junk are not the same underlying defect." } ] }, - "evaluator_result": null, - "evaluator_used": false, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet and stays conservative overall. The single accepted pair (43317, 43873) is a plausible overlap in offloading/quantization loading behavior, and the rejected pairs are mostly distinguished at a high level by title/topic. No obvious overstatement stands out." + }, + "evaluator_used": true, "retried": false, - "accepted_nontrivial_soft_edge": false, + "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null }, @@ -3846,7 +3799,7 @@ "component_index": 1, "component_count": 17, "review_unit_index": 12, - "review_unit_count": 50, + "review_unit_count": 49, "status": "reviewed", "reason": null, "source": "llm", @@ -3879,6 +3832,7 @@ "issue:44186", "issue:44387", "issue:44423", + "issue:44466", "issue:44479", "issue:44610", "issue:44671", @@ -3903,18 +3857,19 @@ "issue:45823" ], "soft_pairs": [ + "issue:44466|issue:45357", "issue:41720|issue:44977", "issue:42617|issue:42915", "issue:40990|issue:45701", "issue:44008|issue:45823", - "issue:44387|issue:45704", "issue:43404|issue:45357", + "issue:44387|issue:45704", "issue:44186|issue:45072", "issue:36331|issue:42617", "issue:43577|issue:43708", "issue:44877|issue:44991", - "issue:43873|issue:44756", "issue:36331|issue:42915", + "issue:43873|issue:44756", "issue:43450|issue:45701", "issue:43550|issue:43720", "issue:44423|issue:45406", @@ -3922,8 +3877,8 @@ "issue:41720|issue:42617", "issue:44479|issue:44918", "issue:44610|issue:44898", - "issue:44186|issue:44610", "issue:43550|issue:43606", + "issue:44186|issue:44610", "issue:42915|issue:43317", "issue:43295|issue:45701", "issue:41720|issue:45161", @@ -3934,214 +3889,220 @@ "issue:43531|issue:43572", "issue:43316|issue:44913" ], - "prepared_review_unit_hash": "10bc8816e98fdc2fa0afda72504ef3be809f931ea8c58011c4fecb45e99290c4", + "prepared_review_unit_hash": "291d53af0f91b8a200c5d1c4597cc1ad7d6524e66b7879a851c96e49923f7de9", "component_budget": { "node_count": 426, "item_count": 426, - "soft_pair_count": 1572, - "serialized_chars": 814130, - "estimated_input_tokens": 203533, - "estimated_eval_tokens": 407322 + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 }, "budget": { - "node_count": 47, - "item_count": 47, - "soft_pair_count": 30, - "serialized_chars": 36283, - "estimated_input_tokens": 9071, - "estimated_eval_tokens": 18398 + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 37223, + "estimated_input_tokens": 9306, + "estimated_eval_tokens": 18868 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This is not a coherent duplicate cluster; it mixes unrelated tokenizer, config, multimodal, training, device-mapping, and quantization/offloading regressions. The only mild thematic neighborhood is memory/offloading, but none of the soft pairs look like the same concrete bug.", - "confidence": 0.95, - "canonical_issue_reason": "No single true canonical duplicate exists here; issue 43873 is only the closest umbrella because it is a broad open offloading/quantization report with the most inbound references, but the rest of the items span different code paths and model families.", + "summary": "These items form a loose grab-bag of Transformers bugs/regressions, but the soft pairs mostly point to different concrete code paths. I do not see any safe duplicate merges among the candidate pairs.", + "confidence": 0.64, + "canonical_issue_reason": "issue:43873 is the best umbrella representative: it is a concrete, open regression about offloading/quantization/device placement, has substantial discussion, and matches the strongest recurring theme across the set without being overly narrow.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 43873 is the best available representative for triage simply because it is broad and active, but it should not be treated as a real duplicate target for most of the cluster.", + "best_issue_reason": "Among the issues, issue:43873 is the most suitable anchor for triage because it is specific enough to be actionable while still broad enough to represent the offloading/quantization/device-mapping subgroup in this cluster.", "best_pr_reason": null, "soft_edge_verdicts": [ + { + "left": "issue:44466", + "right": "issue:45357", + "accept": false, + "reason": "Different bugs: tied lm_head serialization vs visual encoder key persistence in Qwen3.5." + }, { "left": "issue:41720", "right": "issue:44977", "accept": false, - "reason": "Same broad Qwen3.x area, but one is an auto-device-map CUDA assert and the other is a flash-attention generation failure; different bugs and code paths." + "reason": "Same model family, but different failures and code paths: cuda assert on auto device map vs flash-attention generation regression." }, { "left": "issue:42617", "right": "issue:42915", "accept": false, - "reason": "Completely different problems: a 3d_parallel.py execution failure versus a Qwen3Moe FineGrainedFP8Config issue." + "reason": "Unrelated failures: 3d_parallel.py execution vs Qwen3Moe FP8 config loading." }, { "left": "issue:40990", "right": "issue:45701", "accept": false, - "reason": "High perplexity on a model/eval run is unrelated to a version-dependent tokenization change." + "reason": "Different scope: perplexity on one model/dataset vs broad tokenizer-version behavior change." }, { "left": "issue:44008", "right": "issue:45823", "accept": false, - "reason": "Different models and failure modes: Gemma3n forward variable collision versus Gemma4 PLE device mismatch." + "reason": "Different Gemma generations and different symptoms: audio mask AttributeError vs PLE device mismatch." }, { - "left": "issue:44387", - "right": "issue:45704", + "left": "issue:43404", + "right": "issue:45357", "accept": false, - "reason": "Int4 quantization OOM is not the same as T5's apex.FusedRMSNorm memory leak." + "reason": "Different models and root causes: Mistral3 weight tying vs Qwen3.5 visual encoder save/load keys." }, { - "left": "issue:43404", - "right": "issue:45357", + "left": "issue:44387", + "right": "issue:45704", "accept": false, - "reason": "Mistral3 lm_head tying issue and Qwen3.5 save_pretrained visual-key regression are unrelated." + "reason": "Both mention memory, but one is int4 quantization OOM while the other is a T5/apex RMSNorm leak." }, { "left": "issue:44186", "right": "issue:45072", "accept": false, - "reason": "Tokenizer crash on NER/batched padding is unrelated to dtype mismatches in inference wrappers." + "reason": "Tokenizer crash vs dtype mismatch; different subsystems and failure modes." }, { "left": "issue:36331", "right": "issue:42617", "accept": false, - "reason": "CustomTrainer.compute_loss kwarg regression and 3d_parallel.py failure do not share the same underlying bug." + "reason": "Custom trainer API breakage is unrelated to the 3d_parallel.py runtime failure." }, { "left": "issue:43577", "right": "issue:43708", "accept": false, - "reason": "Blip2 dtype persistence and Trainer resume/max_steps computation are different subsystems." + "reason": "Different trainer/model issues: dtype remaining float32 vs checkpoint resume max_steps miscalculation." }, { "left": "issue:44877", "right": "issue:44991", "accept": false, - "reason": "Strict config loading for granite_speech and tokenizer loading for est-roberta are different issues." + "reason": "Different tokenizer/config problems: strict config rejection vs tokenizer loading failure." }, { - "left": "issue:43873", - "right": "issue:44756", + "left": "issue:36331", + "right": "issue:42915", "accept": false, - "reason": "Both mention offloading/memory, but one is quantization offload behavior and the other is mmap on Strix Halo; not the same concrete bug." + "reason": "Trainer loss signature change is unrelated to Qwen3Moe FP8 config failure." }, { - "left": "issue:36331", - "right": "issue:42915", + "left": "issue:43873", + "right": "issue:44756", "accept": false, - "reason": "Trainer loss-signature regression and Qwen3Moe FP8 config failure are unrelated." + "reason": "Both involve memory/offload concerns, but one is quantization offloading behavior and the other is mmap/OOM on Strix Halo." }, { "left": "issue:43450", "right": "issue:45701", "accept": false, - "reason": "Video processor batch-shape bug is unrelated to tokenizer behavior changes." + "reason": "Video batching shape bug is unrelated to a tokenization version regression." }, { "left": "issue:43550", "right": "issue:43720", "accept": false, - "reason": "Bamba torch.compile/SDPA failure and BitNet packed-weight loading failure are different model/runtime problems." + "reason": "Different code paths: torch.compile+SDPA vs accelerate loading of packed BitNet weights." }, { "left": "issue:44423", "right": "issue:45406", "accept": false, - "reason": "Continuous-batching multimodal crash and Gemma4Processor missing _tokenizer are separate processor bugs." + "reason": "Serve crashes from different causes: multimodal tensor handling vs Gemma4Processor missing _tokenizer." }, { "left": "issue:45632", "right": "issue:45684", "accept": false, - "reason": "Cache-path collision for trust_remote_code and permission propagation in save_pretrained are unrelated." + "reason": "Cache-path collision and file permission propagation are distinct bugs." }, { "left": "issue:41720", "right": "issue:42617", "accept": false, - "reason": "Both are model-running failures, but one is Qwen3 device-mapping CUDA assert and the other is a 3d_parallel script issue." + "reason": "Different failures: Qwen3 auto device mapping assert vs 3d_parallel.py runtime error." }, { "left": "issue:44479", "right": "issue:44918", "accept": false, - "reason": "Qwen video-input regression and TRL SFT embedding unpacking are different code paths and symptoms." + "reason": "Both are Qwen-related, but one is a video-input regression and the other is an SFT embedding-unpack failure." }, { "left": "issue:44610", "right": "issue:44898", "accept": false, - "reason": "OmDet-Turbo input-size mismatch and Perceiver non-default-resolution failure are different model-specific resizing bugs." + "reason": "Both concern image sizes, but the concrete model expectations and bug mechanisms differ." }, { - "left": "issue:44186", - "right": "issue:44610", + "left": "issue:43550", + "right": "issue:43606", "accept": false, - "reason": "LayoutLMv2 tokenizer crashes and OmDet processor output-size mismatch are unrelated." + "reason": "Torch.compile/SDPA issue is unrelated to bark CPU-offload device mismatch." }, { - "left": "issue:43550", - "right": "issue:43606", + "left": "issue:44186", + "right": "issue:44610", "accept": false, - "reason": "Bamba torch.compile/SDPA and bark-small CPU-offload device mismatch are unrelated model/runtime failures." + "reason": "Tokenizer crash and processor output-size mismatch are different bugs." }, { "left": "issue:42915", "right": "issue:43317", "accept": false, - "reason": "Qwen3Moe FineGrainedFP8Config and dequantized model loading with device_map=auto are different quantization/offload issues." + "reason": "FP8 config failure is distinct from dequantized model loading with gpu+cpu offload." }, { "left": "issue:43295", "right": "issue:45701", "accept": false, - "reason": "Processor.tokenizer regression and version-dependent tokenization changes are not the same underlying bug." + "reason": "Processor.tokenizer/image handling regression is not the same as general tokenization drift." }, { "left": "issue:41720", "right": "issue:45161", "accept": false, - "reason": "Qwen3 auto device mapping crash and GPT-OSS TP failure are separate parallelism/device-placement issues." + "reason": "Different model families and failures: Qwen3 auto device-map assert vs GPT-OSS tensor-parallel issue." }, { "left": "issue:44671", "right": "issue:44743", "accept": false, - "reason": "CamemBERT masked-LM prediction regression is unrelated to Qwen3.5 recurrent-state reset logic." + "reason": "CamemBERT MLM regression and Qwen3.5 recurrent-state reset are unrelated." }, { "left": "issue:43526", "right": "issue:43881", "accept": false, - "reason": "BEiT fast processor label reduction bug and glm-4v-9b loading failure are unrelated." + "reason": "ImageProcessor label reduction bug is unrelated to glm-4v loading failure." }, { "left": "issue:36010", "right": "issue:42617", "accept": false, - "reason": "ImportError for GenerationMixin and a 3d_parallel.py runtime failure are not the same bug." + "reason": "Importing GenerationMixin and running 3d_parallel.py are unrelated issues." }, { "left": "issue:45081", "right": "issue:45812", "accept": false, - "reason": "Both involve tokenizers, but Mistral regex patching and Granite token ID regression are distinct model-specific issues." + "reason": "Both are tokenizer-related, but one is a Mistral regex patch crash and the other is wrong Granite token IDs; different root causes." }, { "left": "issue:43531", "right": "issue:43572", "accept": false, - "reason": "Qwen3-MoE sliding_window behavior and StableLmConfig pad_token_idx are unrelated config/model bugs." + "reason": "Qwen3-MoE sliding-window behavior and StableLmConfig pad_token_idx are different regressions." }, { "left": "issue:43316", "right": "issue:44913", "accept": false, - "reason": "Gemma3TextConfig API mismatch and GPTNeoX rotary_pct reload regression affect different configs and code paths." + "reason": "Gemma3TextConfig API mismatch and GPTNeoXConfig rotary_pct reload behavior are distinct config bugs." } ] }, @@ -4157,7 +4118,7 @@ "component_index": 1, "component_count": 17, "review_unit_index": 13, - "review_unit_count": 50, + "review_unit_count": 49, "status": "reviewed", "reason": null, "source": "llm", @@ -4215,23 +4176,23 @@ ], "soft_pairs": [ "issue:43540|issue:43782", - "issue:42915|issue:43322", "issue:43421|issue:45812", + "issue:42915|issue:43322", "issue:44186|issue:44361", "issue:29159|issue:36032", "issue:43388|issue:45059", "issue:45406|issue:45464", - "issue:30064|issue:34634", "issue:33357|issue:35707", + "issue:30064|issue:34634", "issue:41720|issue:43746", "issue:43606|issue:43720", "issue:43550|issue:43653", "issue:43421|issue:45701", "issue:43873|issue:45704", - "issue:43606|issue:43653", "issue:44492|issue:44509", - "issue:42490|issue:43329", + "issue:43606|issue:43653", "issue:34634|issue:41720", + "issue:42490|issue:43329", "issue:43531|issue:45759", "issue:34567|issue:35707", "issue:44514|issue:44545", @@ -4244,217 +4205,217 @@ "issue:44077|issue:44625", "issue:45759|issue:45859" ], - "prepared_review_unit_hash": "681df3b8e6cbfe4602006121e6fb13f548c728f67995be3e1054c4ff924bc76c", + "prepared_review_unit_hash": "91fa3fe5332025ec4ec6b8b8aef858db88fdc8236599b96219e6a202adf6b460", "component_budget": { "node_count": 426, "item_count": 426, - "soft_pair_count": 1572, - "serialized_chars": 814130, - "estimated_input_tokens": 203533, - "estimated_eval_tokens": 407322 + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 }, "budget": { "node_count": 47, "item_count": 47, "soft_pair_count": 29, - "serialized_chars": 35960, - "estimated_input_tokens": 8990, - "estimated_eval_tokens": 18236 + "serialized_chars": 35968, + "estimated_input_tokens": 8992, + "estimated_eval_tokens": 18240 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The cluster is mostly a grab bag of unrelated bug reports; only one soft pair looks like a true duplicate. Most other pairs share a broad subsystem (tokenizers, multimodal, offloading, loading) but do not describe the same concrete failure.", - "confidence": 0.86, - "canonical_issue_reason": "issue:44545 is the clearest duplicate target in the set: it matches issue:44514 exactly by title and topic, and has inbound references that make it a better consolidation point.", - "canonical_pr_reason": null, - "best_issue_reason": "issue:44545 is the best issue to keep as the canonical representative of the only real duplicate pair; it is specific, well-scoped, and directly matches issue:44514.", - "best_pr_reason": null, + "summary": "This cluster is mostly a false-positive similarity set spanning unrelated models, processors, and failure modes. The only clear duplicate is the Qwen2_5_VLProcessor.apply_chat_template crash pair (44514/44545), which appears to be the same issue/same crash signature with the same padding=False condition.", + "confidence": 0.91, + "canonical_issue_reason": "Issue 44514 is the best canonical candidate because 44545 is the same Qwen2_5_VLProcessor.apply_chat_template crash report with the same padding=False condition; the rest of the cluster is heterogeneous.", + "canonical_pr_reason": "No pull requests are present in this cluster.", + "best_issue_reason": "44514 is the cleanest representative of the only true duplicate pair; it matches 44545 on the same underlying apply_chat_template crash signature.", + "best_pr_reason": "No pull requests are present in this cluster.", "soft_edge_verdicts": [ { "left": "issue:43540", "right": "issue:43782", "accept": false, - "reason": "Different models and different failure modes: Qwen3OmniMoe video processing vs Qwen3VL weight_only loading error." + "reason": "Different failure surfaces: Qwen3OmniMoe video processing ValueError vs Qwen3VL from_pretrained weight_only load error." }, { - "left": "issue:42915", - "right": "issue:43322", + "left": "issue:43421", + "right": "issue:45812", "accept": false, - "reason": "Both are model-loading failures, but the root causes and models are unrelated (Qwen3Moe FP8 config vs Llava Next segfault)." + "reason": "Both concern tokenization, but one is runtime post-processor updates and the other is a Granite model token-ID regression." }, { - "left": "issue:43421", - "right": "issue:45812", + "left": "issue:42915", + "right": "issue:43322", "accept": false, - "reason": "Both involve tokenization, but one is runtime post-processor syncing and the other is a Granite-specific wrong-token-IDs regression." + "reason": "Unrelated models and bugs: FineGrainedFP8Config failure for Qwen3Moe vs Llava Next segmentation fault." }, { "left": "issue:44186", "right": "issue:44361", "accept": false, - "reason": "Different tokenizer classes and distinct crashes; this is not the same underlying bug." + "reason": "Different tokenizers and symptoms: LayoutLMv2Tokenizer batching/NER crash vs MLukeTokenizer AttributeError." }, { "left": "issue:29159", "right": "issue:36032", "accept": false, - "reason": "One is a general slow-vs-fast tokenizer inconsistency; the other is a T5Tokenizer load-time method-name conflict." + "reason": "Tokenizer inconsistency report vs a T5Tokenizer name conflict on load; not the same bug." }, { "left": "issue:43388", "right": "issue:45059", "accept": false, - "reason": "Completely different subsystems: metric gathering vs multimodal PCS input handling." + "reason": "Metric gathering drops labels in the last batch, while SAM3 text/bbox behavior is a separate multimodal input issue." }, { "left": "issue:45406", "right": "issue:45464", "accept": false, - "reason": "Both affect serving, but the crashes come from different code paths and symptoms." + "reason": "Gemma4Processor missing _tokenizer is unrelated to Qwen3.5-0.8B chat/completions streaming failures." }, { - "left": "issue:30064", - "right": "issue:34634", + "left": "issue:33357", + "right": "issue:35707", "accept": false, - "reason": "Unrelated issues: segmentation-map processing vs BarkProcessor voice preset handling." + "reason": "MacOS CLIP bus error and progressive generation with inputs_embeds/past_key_values are different code paths." }, { - "left": "issue:33357", - "right": "issue:35707", + "left": "issue:30064", + "right": "issue:34634", "accept": false, - "reason": "Different problems: MacOS bus error on CLIP load vs progressive generation with inputs_embeds/past_key_values." + "reason": "Void segmentation-map processing and BarkProcessor voice_preset are unrelated processor bugs." }, { "left": "issue:41720", "right": "issue:43746", "accept": false, - "reason": "Different model families and different failures; auto device mapping cudaErrorAssert is not the same as PEFT checkpoint loading." + "reason": "Qwen3 auto device mapping cuda assert is different from GraniteSpeech PEFT local-checkpoint loading." }, { "left": "issue:43606", "right": "issue:43720", "accept": false, - "reason": "Both touch loading/offload machinery, but the concrete bug is different (CPU offload mismatch vs packed-weight unpacking)." + "reason": "Both mention loading/offload, but CPU offload device mismatch and BitNet packed-weight unpacking are different concrete failures." }, { "left": "issue:43550", "right": "issue:43653", "accept": false, - "reason": "Unrelated bugs: torch.compile/SDPA behavior vs tokenizer special-token registration." + "reason": "torch.compile/SDPA bug is unrelated to BigBirdTokenizer special-token registration." }, { "left": "issue:43421", "right": "issue:45701", "accept": false, - "reason": "Broad tokenization complaints, but one is a post-processor update bug and the other is a version-to-version tokenization change report." + "reason": "Runtime special-token/post-processor syncing is not the same as a broad version-dependent tokenization change report." }, { "left": "issue:43873", "right": "issue:45704", "accept": false, - "reason": "Different features and root causes: quantized offloading vs an apex RMSNorm memory leak." - }, - { - "left": "issue:43606", - "right": "issue:43653", - "accept": false, - "reason": "Different model/component scope and different symptoms; not the same bug." + "reason": "Quantization/offloading behavior is unrelated to the T5 apex FusedRMSNorm memory leak." }, { "left": "issue:44492", "right": "issue:44509", "accept": false, - "reason": "A typo in cache strategies is unrelated to outdated docs mentioning removed pipeline tasks." + "reason": "A typo in cache strategies is unrelated to outdated docs about removed pipeline tasks." }, { - "left": "issue:42490", - "right": "issue:43329", + "left": "issue:43606", + "right": "issue:43653", "accept": false, - "reason": "Local-path import warning and multimodal token-counting video-branch bug are unrelated." + "reason": "CPU offload device mismatch and tokenizer special-token registration are different bugs." }, { "left": "issue:34634", "right": "issue:41720", "accept": false, - "reason": "Different model and runtime issues; BarkProcessor voice preset has no overlap with Qwen3 device mapping errors." + "reason": "BarkProcessor voice_preset and Qwen3 auto device mapping are unrelated." + }, + { + "left": "issue:42490", + "right": "issue:43329", + "accept": false, + "reason": "Local-path import warning is unrelated to the undefined video-branch helpers in multimodal token counting." }, { "left": "issue:43531", "right": "issue:45759", "accept": false, - "reason": "Both are Qwen3-family issues, but one is sliding_window behavior and the other is text_config unwrapping in from_config." + "reason": "Qwen3-MoE sliding_window behavior and AutoModelForCausalLM.from_config text_config unwrapping are different issues." }, { "left": "issue:34567", "right": "issue:35707", "accept": false, - "reason": "TrainerState token counting and generation with inputs_embeds/past_key_values are different bugs." + "reason": "TrainerState token counter not updating is unrelated to progressive generation with inputs_embeds/past_key_values." }, { "left": "issue:44514", "right": "issue:44545", "accept": true, - "reason": "Exact same reported bug and title; these are clear duplicates and should be consolidated." + "reason": "Same Qwen2_5_VLProcessor.apply_chat_template crash signature with the same padding=False condition; these are duplicate reports of the same bug." }, { "left": "issue:43421", "right": "issue:43927", "accept": false, - "reason": "Both involve tokenizer/config state, but one is runtime post-processor updates and the other is save/load losing custom token IDs." + "reason": "Runtime post-processor updates and DiaConfig save/load token-ID loss are different tokenization/config issues." }, { "left": "issue:44625", "right": "issue:45479", "accept": false, - "reason": "Different issues: config propagation of num_labels vs a classification loss degeneracy across models." + "reason": "num_labels propagation in Qwen3.5 is not the same as the single-label-classification zero-loss issue across models." }, { "left": "issue:44451", "right": "issue:44991", "accept": false, - "reason": "Both are tokenizer-loading failures, but they concern different models and likely different conversion problems." + "reason": "Both are loading failures, but for different models and different root causes." }, { "left": "issue:44898", "right": "issue:45072", "accept": false, - "reason": "Different subsystems entirely: image classification resolution handling vs dtype mismatch during bfloat16 inference." + "reason": "Perceiver non-default resolution failure is unrelated to bfloat16 dtype mismatches in SwitchTransformers/TimmWrapperModel." }, { "left": "issue:43531", "right": "issue:44589", "accept": false, - "reason": "Sliding-window model behavior is unrelated to the Float8 storage lookup TypeError." + "reason": "Qwen3-MoE sliding_window bug is unrelated to the Float8 storage object lookup error." }, { "left": "issue:43782", "right": "issue:45381", "accept": false, - "reason": "Both are Qwen multimodal issues, but one is a load-time weight_only error and the other is video vision_position_ids being wrong." + "reason": "Qwen3VL weight_only load error and Qwen2.5-VL video position-id bug are different multimodal issues." }, { "left": "issue:44077", "right": "issue:44625", "accept": false, - "reason": "No underlying overlap: a patchtsmixer post_init validation issue is unrelated to Qwen3.5 label propagation." + "reason": "patchtsmixer post_init validation and Qwen3.5 num_labels propagation do not share a code path." }, { "left": "issue:45759", "right": "issue:45859", "accept": false, - "reason": "Same broader model family, but different concrete failure points: from_config text_config unwrapping vs rotary embedding CPU-offload compatibility." + "reason": "Composite config unwrapping and CPU-offload rotary-embedding incompatibility are unrelated." } ] }, "evaluator_result": { - "accept": true, - "feedback": "The summary is well grounded in the packet, and the one accepted duplicate pair (issue:44514 / issue:44545) is supported by the near-exact title match plus inbound references. The other soft-edge verdicts stay conservative and do not overclaim duplication from the limited evidence shown." + "accept": false, + "feedback": "The overall clustering judgment is plausible, but the writeup is slightly overstated in a way that matters for precision: 44514 and 44545 do not have the exact same title text, only the same underlying bug wording. Please soften that claim to something like \"same issue/same crash signature\" rather than \"exact same title.\" The rest of the soft-edge rejections look conservative, and the no-PR note is grounded." }, "evaluator_used": true, - "retried": false, + "retried": true, "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null @@ -4464,7 +4425,7 @@ "component_index": 1, "component_count": 17, "review_unit_index": 14, - "review_unit_count": 50, + "review_unit_count": 49, "status": "reviewed", "reason": null, "source": "llm", @@ -4526,8 +4487,8 @@ "issue:44368|issue:45357", "issue:44062|issue:44451", "issue:45588|issue:45759", - "issue:43421|issue:43653", "issue:43746|issue:46050", + "issue:43421|issue:43653", "issue:39692|issue:44610", "issue:45704|issue:45923", "issue:44918|issue:44936", @@ -4553,214 +4514,214 @@ "issue:43618|issue:43697", "issue:45183|issue:45479" ], - "prepared_review_unit_hash": "37f2b68afbd7cfd7be020957ee785a1ff600e427cdeeb8188ec7248a1badfad9", + "prepared_review_unit_hash": "183da652f83f324bb62c8ac9d4497d6e00f742faacdbb963fb9dda5320e6795d", "component_budget": { "node_count": 426, "item_count": 426, - "soft_pair_count": 1572, - "serialized_chars": 814130, - "estimated_input_tokens": 203533, - "estimated_eval_tokens": 407322 + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 30, - "serialized_chars": 37212, - "estimated_input_tokens": 9303, - "estimated_eval_tokens": 18862 + "serialized_chars": 37220, + "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 mostly a set of unrelated bug reports and regressions across tokenizers, multimodal processors, training, offload, and model-loading paths. The soft-similarity edges are theme-level only; none look like the same underlying defect or a mergeable duplicate PR.", - "confidence": 0.97, - "canonical_issue_reason": null, + "summary": "These are mostly superficial text matches, not duplicate pairs. The issues span unrelated models, subsystems, and failure modes (tokenizers, quantization, offloading, vision processors, training, offline hub behavior), so I would not merge any soft pair as the same underlying bug/change.", + "confidence": 0.95, + "canonical_issue_reason": "issue:43873 is the strongest standalone anchor by activity and breadth, but the set is too heterogeneous to represent a true duplicate cluster.", "canonical_pr_reason": null, - "best_issue_reason": null, + "best_issue_reason": "issue:43873 has the richest discussion/inbound activity and a concrete reproducible report, so it is the best representative issue here even though it is not a duplicate of the others.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44987", "right": "issue:45042", "accept": false, - "reason": "Different failures: one is a model-loading regression for a specific repo, the other is PIL/image-processor requiring torchvision in v5.4.0." + "reason": "Different bugs: model loading regression vs image processor requiring torchvision." }, { "left": "issue:44368", "right": "issue:45357", "accept": false, - "reason": "One is a LoRA training warning about tie_word_embeddings; the other is a save_pretrained bug for Qwen3.5 visual encoder keys." + "reason": "Unrelated Qwen3.5 warning about tie_word_embeddings vs wrong visual encoder keys on save." }, { "left": "issue:44062", "right": "issue:44451", "accept": false, - "reason": "AddedToken keyword collision in tokenization code vs failure to load a specific model repo; different causes and code paths." + "reason": "AddedToken keyword collision is a tokenizer API bug; ScandiBERT load failure is a separate model-loading issue." }, { "left": "issue:45588", "right": "issue:45759", "accept": false, - "reason": "Flash-attention sink-less crash and composite Qwen multimodal config unwrapping are unrelated bugs." + "reason": "Flash-attention crash on s_aux=None is unrelated to multimodal config unwrapping for Qwen 3.5/3.6." }, { - "left": "issue:43421", - "right": "issue:43653", + "left": "issue:43746", + "right": "issue:46050", "accept": false, - "reason": "Both involve special tokens, but one is runtime post-processor refresh in TokenizersBackend and the other is BigBirdTokenizer mask-token registration/decoding." + "reason": "PEFT adapter checkpoint loading and sinq quantization save/load are different code paths." }, { - "left": "issue:43746", - "right": "issue:46050", + "left": "issue:43421", + "right": "issue:43653", "accept": false, - "reason": "Local PEFT checkpoint loading for GraniteSpeech vs quantized model save/load breakage after sinq; distinct loading flows." + "reason": "Runtime post-processor syncing is not the same as BigBirdTokenizer special-token registration." }, { "left": "issue:39692", "right": "issue:44610", "accept": false, - "reason": "Both are doc/processor mismatches, but they affect different models and different expected image sizes/quantization behavior." + "reason": "Documentation/example errors for SigLIP2 are unrelated to OmDet-Turbo processor size mismatch." }, { "left": "issue:45704", "right": "issue:45923", "accept": false, - "reason": "Apex RMSNorm memory leak in T5 is unrelated to a missing supports_gradient_checkpointing flag on a trust_remote_code model." + "reason": "T5 apex RMSNorm leak is unrelated to a missing supports_gradient_checkpointing flag." }, { "left": "issue:44918", "right": "issue:44936", "accept": false, - "reason": "TRL SFT embedding unpacking failure vs trainer.evaluate() after train; different trainer lifecycle issues." + "reason": "TRL SFT unpacking failure is a different bug from trainer.evaluate() failing after train()." }, { "left": "issue:45663", "right": "issue:45923", "accept": false, - "reason": "FSDP2 shared_kv_states KeyError in Gemma-4 is not the same defect as the missing gradient-checkpointing flag." + "reason": "FSDP2 shared_kv_states rebuild bug is not the same as a missing gradient-checkpointing capability flag." }, { "left": "issue:42886", "right": "issue:44843", "accept": false, - "reason": "Both mention offline mode, but one is tokenizer cache loading and the other is an unconditional model_info() call in _patch_mistral_regex." + "reason": "Both involve offline hub behavior, but one is cache loading and the other is an unconditional model_info() call in a regex patch." }, { "left": "issue:43381", "right": "issue:43844", "accept": false, - "reason": "Gradient checkpointing blocked in eval mode vs abnormal gradient growth under ZeRO-3; different training bugs." + "reason": "Eval-mode gradient checkpointing restriction and ZeRO-3 gradient growth are distinct problems." }, { "left": "issue:43421", "right": "issue:44186", "accept": false, - "reason": "Runtime special-token post-processor updates and LayoutLMv2Tokenizer NER padding/truncation crashes are separate tokenizer problems." + "reason": "Special-token post-processor syncing does not match LayoutLMv2 NER/batched padding crashes." }, { "left": "issue:43531", "right": "issue:43645", "accept": false, - "reason": "Qwen3-MoE sliding_window behavior and Jupyter custom model initialization in Transformers 5.0 are unrelated." + "reason": "Qwen3-MoE sliding-window behavior is unrelated to custom model initialization in notebooks." }, { "left": "issue:45704", "right": "issue:45859", "accept": false, - "reason": "T5 RMSNorm memory leak is unrelated to Qwen3.5-MoE rotary embedding CPU-offload incompatibility." + "reason": "RMSNorm memory leak in T5 is unrelated to CPU offload compatibility in Qwen3.5 MoE rotary embedding." }, { "left": "issue:43994", "right": "issue:44671", "accept": false, - "reason": "SigLIP2 bad AutoModel/pipeline outputs and CamemBERT masked-LM prediction regression are different model-specific issues." + "reason": "Different models and symptoms: SigLIP2 pipeline output issue vs CamemBERT MLM regression." }, { "left": "issue:41720", "right": "issue:45704", "accept": false, - "reason": "Qwen3 auto device-map cuda assert and T5 apex RMSNorm leak are different subsystems and symptoms." + "reason": "CUDA assert with auto device mapping on Qwen3 is unrelated to T5 RMSNorm leak." }, { "left": "issue:43596", "right": "issue:44220", "accept": false, - "reason": "Deepspeed zero3/BertModel init crash and _torch_extract_fbank_features failure are unrelated runtime paths." + "reason": "Zero3/BertModel init failure and fbank feature extraction issue are unrelated." }, { "left": "issue:43746", "right": "issue:43873", "accept": false, - "reason": "Checkpoint loading for GraniteSpeech PEFT adapters vs quantization offloading behavior; not the same bug." + "reason": "Local checkpoint loading with PEFT adapters is a different bug from quantization/offloading behavior." }, { "left": "issue:45081", "right": "issue:45356", "accept": false, - "reason": "Both mention Mistral regex/fix warnings, but one is a tokenizer crash and the other is a Kimi codec regression; not the same concrete defect." + "reason": "Shared tokenizer-regex terminology, but one is an AttributeError in _patch_mistral_regex and the other is a Kimi tokenizer codec regression." }, { "left": "issue:39692", "right": "issue:43873", "accept": false, - "reason": "SigLIP2 processor/model mismatch is unrelated to offloading not working with quantization." + "reason": "SigLIP2 doc/example problems are unrelated to offloading with quantization." }, { "left": "issue:40444", "right": "issue:43746", "accept": false, - "reason": "Qwen2.5-VL multi-image IterableDataset failure and GraniteSpeech local checkpoint loading are separate multimodal/training bugs." + "reason": "Multi-image IterableDataset finetuning failure is unrelated to PEFT local checkpoint loading." }, { "left": "issue:45356", "right": "issue:45907", "accept": false, - "reason": "Kimi tokenizer regression and list_repo_files not forwarding revision are unrelated." + "reason": "Tokenizer codec/regression bug is unrelated to list_repo_files needing to forward revision." }, { "left": "issue:30064", "right": "issue:36010", "accept": false, - "reason": "Void segmentation map processing and GenerationMixin import error are entirely different issues." + "reason": "Void segmentation-map processing and GenerationMixin import failure are completely different issues." }, { "left": "issue:44220", "right": "issue:44661", "accept": false, - "reason": "Audio fbank extraction failure vs add-new-model-like logic with TOKENIZER_MAPPING_NAMES; different code paths." + "reason": "Audio fbank extraction and TOKENIZER_MAPPING_NAMES handling are unrelated." }, { "left": "issue:43720", "right": "issue:43819", "accept": false, - "reason": "BitNet packed-weight unpacking during accelerate loading and DAC.from_latents STE mismatch are unrelated model bugs." + "reason": "BitNet packed-weight loading and DAC.from_latents missing-STE behavior are different models and code paths." }, { "left": "issue:43653", "right": "issue:43720", "accept": false, - "reason": "BigBirdTokenizer special-token handling and BitNet packed-weight loading are unrelated." + "reason": "Special-token decode bug in BigBirdTokenizer is unrelated to BitNet accelerate-loading behavior." }, { "left": "issue:43540", "right": "issue:44008", "accept": false, - "reason": "Qwen3OmniMoe video-input validation and Gemma 3n audio_mel_mask name collision are different multimodal forward-pass bugs." + "reason": "Video-processing ValueError in Qwen3OmniMoe is unrelated to Gemma 3n audio_mel_mask name collision." }, { "left": "issue:43618", "right": "issue:43697", "accept": false, - "reason": "CLIPOutput attentions regression and RTDetrV2 differing outputs in v5 are separate model-specific issues." + "reason": "CLIP attentions assignment regression and RTDetrV2 output mismatch are separate model-specific bugs." }, { "left": "issue:45183", "right": "issue:45479", "accept": false, - "reason": "Wrong input_ids/inputs_embeds validation messages and degenerate zero-loss classification behavior are unrelated bugs." + "reason": "Input validation error messaging is unrelated to zero-loss classification with num_labels=1." } ] }, @@ -4776,7 +4737,7 @@ "component_index": 1, "component_count": 17, "review_unit_index": 15, - "review_unit_count": 50, + "review_unit_count": 49, "status": "reviewed", "reason": null, "source": "llm", @@ -4835,223 +4796,223 @@ ], "soft_pairs": [ "issue:43550|issue:45072", - "issue:43388|issue:46032", "issue:43295|issue:45834", + "issue:43388|issue:46032", "issue:43550|issue:43819", "issue:43716|issue:44008", "issue:44062|issue:44488", "issue:43322|issue:45685", "issue:43335|issue:43756", "issue:44186|issue:44898", - "issue:43950|issue:45987", "issue:43653|issue:43819", + "issue:43950|issue:45987", "issue:43531|issue:43643", "issue:44843|issue:45081", "issue:43643|issue:44877", "issue:43540|issue:45381", "issue:30064|issue:36331", "issue:44448|issue:44671", - "issue:43577|issue:43818", "issue:43421|issue:44811", "issue:43782|issue:44977", "issue:41720|issue:44991", + "issue:43577|issue:43818", "issue:44821|issue:45084", "issue:39692|issue:45834", "issue:45584|issue:45834", "issue:44186|issue:44442", - "issue:43644|issue:43749", - "issue:44561|issue:45020" + "issue:44561|issue:45020", + "issue:43644|issue:43749" ], - "prepared_review_unit_hash": "d5d62991e697afb83fe7d97009ad5f4ae5e431bd16484a2f96ff9bc92ccf2da6", + "prepared_review_unit_hash": "3bceb5a717713e7f592f2822df3a21298b7eb768bad6588e8f8ebd2e08cdd7b6", "component_budget": { "node_count": 426, "item_count": 426, - "soft_pair_count": 1572, - "serialized_chars": 814130, - "estimated_input_tokens": 203533, - "estimated_eval_tokens": 407322 + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 27, - "serialized_chars": 36047, - "estimated_input_tokens": 9012, - "estimated_eval_tokens": 18280 + "serialized_chars": 36053, + "estimated_input_tokens": 9014, + "estimated_eval_tokens": 18284 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "No soft pair looks like a true duplicate; these are mostly unrelated bug reports that share broad subsystem vocabulary (tokenizers, processors, remote_code, dtype, loading). No PRs are present. I\u2019d use issue 45020 as the broadest issue anchor, but only because the set lacks a clearer canonical duplicate hub.", - "confidence": 0.78, - "canonical_issue_reason": "Issue 45020 is the broadest, most central regression report in the set (\u201cRecent transformers versions break models using `remote_code`\u201d) and has enough scope to serve as the least-bad anchor when no duplicate cluster is obvious.", + "summary": "This cluster is mostly heterogeneous: many pairs share broad Transformers/tokenizer/vision/remote_code wording, but they describe different failure modes, models, and code paths. I do not see any safe duplicate/merge candidates among the soft edges; the closest thematic grouping is around recent v5 / remote_code / tokenizer-loading regressions.", + "confidence": 0.84, + "canonical_issue_reason": "issue:45020 is the broadest umbrella among the set, covering recent Transformers regressions affecting remote_code-based models; it best matches the recurring theme behind several nearby tokenizer/config-loading reports.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 45020 is the strongest global representative: broad scope, regression framing, and multiple references make it a better anchor than the very specific single-model bug reports.", + "best_issue_reason": "issue:45020 is the strongest representative issue for this cluster because it is general, clearly stated, and could subsume adjacent remote_code/tokenizer loading regressions better than the narrower bug reports.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43550", "right": "issue:45072", "accept": false, - "reason": "Different models and different failure modes: torch.compile+SDPA vs bfloat16 dtype mismatch." + "reason": "Different models and bugs: torch.compile/SDPA behavior for Bamba vs dtype mismatch in SwitchTransformers/TimmWrapperModel." }, { - "left": "issue:43388", - "right": "issue:46032", + "left": "issue:43295", + "right": "issue:45834", "accept": false, - "reason": "Different code paths: metric gathering/last-batch label handling vs Mamba2 cache behavior with seq_len>1." + "reason": "Unrelated regressions: processor.tokenizer/images regression vs Kosmos2.5 long OCR index error." }, { - "left": "issue:43295", - "right": "issue:45834", + "left": "issue:43388", + "right": "issue:46032", "accept": false, - "reason": "Both involve processors/tokenizers, but one is a processor.tokenizer regression and the other is a long OCR-input index error in Kosmos2.5." + "reason": "Different subsystems: metric label truncation in gather_for_metrics vs Mamba2 use_cache correctness." }, { "left": "issue:43550", "right": "issue:43819", "accept": false, - "reason": "Unrelated bugs: Bamba compile/SDPA failure vs DAC.from_latents/STE mismatch." + "reason": "Both are runtime failures, but the code paths are unrelated: compile/SDPA vs DAC.from_latents/forward mismatch." }, { "left": "issue:43716", "right": "issue:44008", "accept": false, - "reason": "Different models and distinct bugs: image preprocessor dtype mismatch vs Gemma3n variable-name collision in forward()." + "reason": "Separate model bugs: image preprocessor/model dtype mismatch vs Gemma3n variable name collision in forward()." }, { "left": "issue:44062", "right": "issue:44488", "accept": false, - "reason": "Both are tokenizer-loading issues, but one is an AddedToken keyword collision and the other is a model-specific load failure." + "reason": "Tokenizer-related, but not the same bug: AddedToken special-kwarg duplication vs loading a specific tokenizer model." }, { "left": "issue:43322", "right": "issue:45685", "accept": false, - "reason": "Completely different failures: Llava Next loading segfault vs an MPS histogram kernel limitation." + "reason": "Completely different failure modes: Llava Next load segfault vs MPS int histogram unsupported in moe code." }, { "left": "issue:43335", "right": "issue:43756", "accept": false, - "reason": "Different architecture/config bugs: sparse layer creation vs RoPE layer dropping in Smollm3." + "reason": "Both are config/model bugs, but one is sparse-layer creation and the other is RoPE-layer dropping; different underlying changes." }, { "left": "issue:44186", "right": "issue:44898", "accept": false, - "reason": "Different model families and failure mechanisms; only a broad preprocessing label of tokenization/image handling overlaps." + "reason": "Different components and symptoms: LayoutLMv2 tokenizer NER/padding crash vs Perceiver image classification with non-default resolution." }, { - "left": "issue:43950", - "right": "issue:45987", + "left": "issue:43653", + "right": "issue:43819", "accept": false, - "reason": "Separate loading/runtime contract bugs: non-persistent buffer corruption vs StaticCache sequence-length return type." + "reason": "Not the same issue: BigBirdTokenizer special-token registration vs DAC latent/forward mismatch." }, { - "left": "issue:43653", - "right": "issue:43819", + "left": "issue:43950", + "right": "issue:45987", "accept": false, - "reason": "Tokenizer special-token registration bug vs DAC latent reconstruction mismatch; no shared underlying bug." + "reason": "Both mention contract/regression style problems, but one is non-persistent buffer corruption and the other is cache sequence-length return type." }, { "left": "issue:43531", "right": "issue:43643", "accept": false, - "reason": "Different issues in different layers: Qwen3-MoE sliding_window behavior vs AutoConfig trust_remote_code missing fields." + "reason": "Different regression targets: Qwen3-MoE sliding_window behavior vs trust_remote_code config field loss." }, { "left": "issue:44843", "right": "issue:45081", "accept": false, - "reason": "Same helper name, but distinct problems: offline model_info call vs backend_tokenizer attribute crash." + "reason": "Same helper name, but distinct bugs: unconditional model_info() network access vs backend_tokenizer attribute crash." }, { "left": "issue:43643", "right": "issue:44877", "accept": false, - "reason": "Both touch config loading, but one is missing fields under trust_remote_code and the other is strict config rejection for granite_speech." + "reason": "Both involve config loading, but one is missing fields with trust_remote_code and the other is strict config rejection for granite_speech." }, { "left": "issue:43540", "right": "issue:45381", "accept": false, - "reason": "Different multimodal models and different errors: video input ValueError vs wrong vision_position_ids in qwen2.5-vl." + "reason": "Different model families and failures: Qwen3OmniMoe video input ValueError vs qwen2.5-vl vision_position_ids issue." }, { "left": "issue:30064", "right": "issue:36331", "accept": false, - "reason": "Image segmentation-map processing bug vs CustomTrainer compute_loss signature regression; unrelated." + "reason": "Unrelated domains: image processor segmentation-map handling vs CustomTrainer.compute_loss signature error." }, { "left": "issue:44448", "right": "issue:44671", "accept": false, - "reason": "Different model outputs/regressions: Pegasus generation drift vs CamemBERT masked-LM prediction error." - }, - { - "left": "issue:43577", - "right": "issue:43818", - "accept": false, - "reason": "Different models and bugs: BLIP2 dtype handling vs Video-LLaVA temporal-attention/weight-sharing issue." + "reason": "Both are output regressions in v5, but different models and behaviors: Pegasus generation vs CamemBERT masked LM predictions." }, { "left": "issue:43421", "right": "issue:44811", "accept": false, - "reason": "Related to special tokens/decoding, but one is runtime post-processor refresh and the other is Whisper batch_decode skip_special_tokens behavior." + "reason": "Tokenizer/post-processing related, but different bugs: runtime special-token updates vs Whisper batch_decode skip_special_tokens handling." }, { "left": "issue:43782", "right": "issue:44977", "accept": false, - "reason": "Different Qwen variants and distinct failures: weight_only load error vs flash-attention generation bug." + "reason": "Different Qwen variants and failures: weight_only loading error vs flash-attention generation correctness." }, { "left": "issue:41720", "right": "issue:44991", "accept": false, - "reason": "No shared underlying bug: CUDA assert on device mapping vs tokenizer loading failure for a specific model." + "reason": "Completely unrelated: Qwen3 auto device mapping CUDA assert vs tokenizer loading failure for a RoBERTa model." + }, + { + "left": "issue:43577", + "right": "issue:43818", + "accept": false, + "reason": "Both are multimodal model reports, but one is dtype persistence in BLIP2 and the other is Llava Next architecture/weights problem." }, { "left": "issue:44821", "right": "issue:45084", "accept": false, - "reason": "Different code paths: AutoImageProcessor URL loading vs template compilation error." + "reason": "Different failures: loading AutoImageProcessor from URL vs template compilation error." }, { "left": "issue:39692", "right": "issue:45834", "accept": false, - "reason": "Documentation/example errors vs a runtime OCR index error; not the same bug." + "reason": "Documentation/example errors for SigLIP2 are unrelated to Kosmos2.5 OCR index handling." }, { "left": "issue:45584", "right": "issue:45834", "accept": false, - "reason": "Whisper empty-transcription generation bug is unrelated to Kosmos2.5 long OCR input indexing." + "reason": "Whisper empty-transcription generation bug and Kosmos2.5 OCR index error are unrelated code paths." }, { "left": "issue:44186", "right": "issue:44442", "accept": false, - "reason": "Tokenizer crash on NER/batching vs AutoTokenizer failing to load a specific tokenizer class." + "reason": "Both are tokenizer problems, but for different tokenizers and crashes; not a single underlying bug." }, { - "left": "issue:43644", - "right": "issue:43749", + "left": "issue:44561", + "right": "issue:45020", "accept": false, - "reason": "Both are loading-related regressions, but one corrupts non-persistent buffers and the other breaks FSDP CPU RAM-efficient loading." + "reason": "Related only at a high level (v5 remote_code breakage), but these are not the same concrete bug or fix." }, { - "left": "issue:44561", - "right": "issue:45020", + "left": "issue:43644", + "right": "issue:43749", "accept": false, - "reason": "Both mention remote_code, but one is a specific missing-function regression and the other is a broad class of recent remote_code failures." + "reason": "Both are transformer loading regressions, but one corrupts non-persistent buffers while the other breaks FSDP CPU RAM efficient loading." } ] }, @@ -5067,7 +5028,7 @@ "component_index": 1, "component_count": 17, "review_unit_index": 16, - "review_unit_count": 50, + "review_unit_count": 49, "status": "reviewed", "reason": null, "source": "llm", @@ -5127,11 +5088,11 @@ "issue:43606|issue:45072", "issue:44479|issue:45834", "issue:43825|issue:43994", - "issue:43577|issue:43582", "issue:34634|issue:36010", + "issue:43577|issue:43582", "issue:44805|issue:44936", - "issue:44671|issue:45701", "issue:44485|issue:44509", + "issue:44671|issue:45701", "issue:43825|issue:43827", "issue:28218|issue:29159", "issue:44610|issue:45072", @@ -5139,215 +5100,215 @@ "issue:43782|issue:45325", "issue:42175|issue:45081", "issue:43749|issue:44291", - "issue:43582|issue:44857", "issue:45081|issue:45406", + "issue:43582|issue:44857", "issue:43650|issue:43906", "issue:43262|issue:43819", - "issue:44060|issue:44368", "issue:43526|issue:43697", + "issue:44060|issue:44368", "issue:34634|issue:36331", "issue:43653|issue:44912", + "issue:43742|issue:45020", "issue:45084|issue:45538", "issue:43526|issue:44554", - "issue:43742|issue:45020", "issue:43329|issue:45600" ], - "prepared_review_unit_hash": "1d094772d2e3c090a0fbde006fa41bb31311a7a86f1f8b8a378aa2b0f84cd0fe", + "prepared_review_unit_hash": "9c4125658a2713a99f4f9b48866c41c27b296475a5938e2a2f4a0534b6994bfc", "component_budget": { "node_count": 426, "item_count": 426, - "soft_pair_count": 1572, - "serialized_chars": 814130, - "estimated_input_tokens": 203533, - "estimated_eval_tokens": 407322 + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 }, "budget": { "node_count": 47, "item_count": 47, "soft_pair_count": 27, - "serialized_chars": 35511, - "estimated_input_tokens": 8878, - "estimated_eval_tokens": 18012 + "serialized_chars": 35505, + "estimated_input_tokens": 8877, + "estimated_eval_tokens": 18010 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These items are mostly unrelated issues spanning tokenization, multimodal processors, loading, training, and docs. I only see one plausible duplicate pair in the tokenizer-added-token spacing reports; the rest should stay separate.", - "confidence": 0.83, - "canonical_issue_reason": "No single issue cleanly represents the full set because the artifacts span many unrelated bugs. The tokenizer spacing/inconsistency reports are the only plausible duplicate subcluster.", + "summary": "This cluster is mostly a grab bag of unrelated Transformers bug reports across tokenizers, model loading, processors, and v5 regressions. The only clear duplicate-like pair is the tokenizer spacing/inconsistency reports; a couple of others are thematically related but not the same concrete bug.", + "confidence": 0.71, + "canonical_issue_reason": "issue:45020 is the broadest and most representative report in the set: it covers recent-version breakage for models using `remote_code`, which matches the cluster\u2019s general pattern of version-regression/loading failures and has the strongest overall signal among the issues.", "canonical_pr_reason": null, - "best_issue_reason": "If a representative issue must be chosen, issue 28218 is the most concrete user-facing report of the tokenizer-added-token spacing bug.", + "best_issue_reason": "issue:45020 is the best single issue to anchor the cluster because it is broad, well-discussed, and captures the most general regression theme; the rest are narrower point bugs.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43606", "right": "issue:45072", "accept": false, - "reason": "Both are inference dtype/mismatch failures, but they affect different models and code paths; not the same bug." + "reason": "Both are model-loading/runtime mismatches, but they affect different subsystems and different concrete failure modes (CPU offload device mismatch vs dtype mismatch)." }, { "left": "issue:44479", "right": "issue:45834", "accept": false, - "reason": "Different model families and different failures: a broad video-input regression vs an OCR length/index error." + "reason": "Both involve multimodal/video-style inputs, but the model families and bugs differ: a v5 video-input regression versus an OCR long-input index error." }, { "left": "issue:43825", "right": "issue:43994", "accept": false, - "reason": "One is a wrong pipeline error message about removed translation tasks; the other is a model output regression. Same words, different bug." + "reason": "One is a wrong error message about removed translation pipelines; the other is incorrect outputs from AutoModel/pipeline on SigLIP. Same broad area, not the same bug." }, { - "left": "issue:43577", - "right": "issue:43582", + "left": "issue:34634", + "right": "issue:36010", "accept": false, - "reason": "Blip2 dtype propagation and an Apple Silicon allocator TypeError are unrelated." + "reason": "Completely different problems: BarkProcessor voice preset handling versus a GenerationMixin import error." }, { - "left": "issue:34634", - "right": "issue:36010", + "left": "issue:43577", + "right": "issue:43582", "accept": false, - "reason": "BarkProcessor voice preset behavior and a GenerationMixin import error are unrelated issues." + "reason": "Different code paths and symptoms: Blip2 dtype propagation versus an Apple Silicon allocator warmup TypeError." }, { "left": "issue:44805", "right": "issue:44936", "accept": false, - "reason": "Both mention failures at runtime, but the underlying contexts differ; this is not one concrete bug." + "reason": "These are unrelated failures: a tensor mask shape mismatch versus trainer.evaluate() failing after training." }, { - "left": "issue:44671", - "right": "issue:45701", + "left": "issue:44485", + "right": "issue:44509", "accept": false, - "reason": "CamemBERT masked-LM regression is not the same underlying issue as a broad tokenizer-version change report." + "reason": "One is a GLM-5 RoPE implementation discussion; the other is a docs cleanup issue about removed pipeline tasks." }, { - "left": "issue:44485", - "right": "issue:44509", + "left": "issue:44671", + "right": "issue:45701", "accept": false, - "reason": "GLM-5 RoPE implementation discussion and v5 pipeline doc leftovers are unrelated." + "reason": "One is a CamemBERT masked-LM regression; the other is a generic tokenization-version change report. Not the same underlying bug." }, { "left": "issue:43825", "right": "issue:43827", "accept": false, - "reason": "One is a runtime error message, the other is docs still referencing removed pipelines; related topic, not the same bug." + "reason": "Both are about pipeline removals in v5, but one is runtime/error-message behavior and the other is documentation references; too broad to treat as one bug." }, { "left": "issue:28218", "right": "issue:29159", "accept": true, - "reason": "Both describe the same tokenizer added-token spacing/inconsistency problem; 29159 is the broader slow-vs-fast framing of the same symptom." + "reason": "Both point to the same tokenizer added-token inconsistency, especially around spacing/slow-vs-fast behavior, and look plausibly fixable by one tokenizer patch." }, { "left": "issue:44610", "right": "issue:45072", "accept": false, - "reason": "Different model-specific bugs: image size mismatch versus dtype mismatch." + "reason": "Different issues: processor image-size mismatch versus dtype mismatch in inference." }, { "left": "issue:43720", "right": "issue:45072", "accept": false, - "reason": "Packed-weight loading failure in BitNet is unrelated to the dtype mismatch reports." + "reason": "Packed-weight loading in BitNet and dtype mismatches in SwitchTransformers/TimmWrapper are separate bugs." }, { "left": "issue:43782", "right": "issue:45325", "accept": false, - "reason": "A weight_only loading error and a RoPE position-id scaling bug are different code paths." + "reason": "Different Qwen-VL failures: from_pretrained weight_only loading versus RoPE/position-id scaling." }, { "left": "issue:42175", "right": "issue:45081", "accept": false, - "reason": "Backend package selection and a tokenizer regex crash are unrelated." + "reason": "Unrelated: packaging/backend dependency selection versus a tokenizer regex patch crash." }, { "left": "issue:43749", "right": "issue:44291", "accept": false, - "reason": "Both are loading-related, but one is FSDP CPU RAM efficient loading and the other is init_empty_weights argument handling; not the same bug." + "reason": "Both involve loading paths, but one is FSDP CPU RAM-efficient loading and the other is init_empty_weights argument handling; not the same bug." }, { - "left": "issue:43582", - "right": "issue:44857", + "left": "issue:45081", + "right": "issue:45406", "accept": false, - "reason": "Apple Silicon warmup TypeError and CUDA float16 loss crash are unrelated." + "reason": "Both are AttributeErrors during loading, but they arise from different objects and different processor/tokenizer code paths." }, { - "left": "issue:45081", - "right": "issue:45406", + "left": "issue:43582", + "right": "issue:44857", "accept": false, - "reason": "Both mention missing tokenizer/processor attributes, but they occur in different components and are not the same concrete failure." + "reason": "Different runtime environments and failures: Apple Silicon allocator warmup TypeError versus CUDA float16 AMP loss crash." }, { "left": "issue:43650", "right": "issue:43906", "accept": false, - "reason": "The titles do not describe the same bug; one is nonspecific noise and the other is a reproduction of a different issue." + "reason": "The first is a vague data request; the second is a reproduction issue for another bug. No duplicate signal." }, { "left": "issue:43262", "right": "issue:43819", "accept": false, - "reason": "Audio chat-template sampling-rate handling and DAC latent reconstruction are unrelated." + "reason": "Audio chat-template sampling-rate defaults and DAC.from_latents/STE mismatch are unrelated." }, { - "left": "issue:44060", - "right": "issue:44368", + "left": "issue:43526", + "right": "issue:43697", "accept": false, - "reason": "Both concern warning behavior, but they target different models and different tied-weight logic." + "reason": "Different model families and failure modes: Beit label reduction bug versus RTDetrV2 output differences in v5." }, { - "left": "issue:43526", - "right": "issue:43697", + "left": "issue:44060", + "right": "issue:44368", "accept": false, - "reason": "BeitImageProcessorFast label reduction and RTDetrV2 output drift are unrelated." + "reason": "Both mention tied-weight warnings in Qwen models, but the titles suggest different manifestations and model variants; not enough to call the same concrete bug." }, { "left": "issue:34634", "right": "issue:36331", "accept": false, - "reason": "BarkProcessor voice_preset and Trainer.compute_loss signature mismatch are unrelated." + "reason": "Bark processor voice_preset and CustomTrainer.compute_loss signature changes are unrelated." }, { "left": "issue:43653", "right": "issue:44912", "accept": false, - "reason": "Tokenizer special-token registration and MXFP4 quantization loading are unrelated." + "reason": "Tokenizer special-token registration and MXFP4 quantization fallback are completely different issues." + }, + { + "left": "issue:43742", + "right": "issue:45020", + "accept": false, + "reason": "A specific key error while loading one model is not the same concrete issue as the broader remote_code regression report." }, { "left": "issue:45084", "right": "issue:45538", "accept": false, - "reason": "Template compilation failure and CLIPTokenizer max length are unrelated." + "reason": "Template compilation failure and an extreme `model_max_length` setting are unrelated." }, { "left": "issue:43526", "right": "issue:44554", "accept": false, - "reason": "Image label reduction and MPS attention correctness are unrelated." - }, - { - "left": "issue:43742", - "right": "issue:45020", - "accept": false, - "reason": "A key error loading one model is not the same as the broad remote_code breakage report." + "reason": "Beit label reduction and an MPS attention correctness issue are separate bugs." }, { "left": "issue:43329", "right": "issue:45600", "accept": false, - "reason": "A multimodal token-counting video branch bug and stale auto_mappings config references are unrelated." + "reason": "One is a multimodal-token-counting video-branch bug; the other is stale auto_mappings config references. Different code paths." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is well grounded in the packet: it correctly treats the set as mostly unrelated and conservatively isolates only the tokenizer added-token spacing / slow-vs-fast inconsistency pair as a plausible duplicate subcluster. The soft-edge verdicts stay on the cautious side and do not overmerge distinct model-specific or component-specific failures." + "feedback": "The review is conservative and grounded in the packet. The cluster summary correctly frames most items as unrelated or only loosely thematically similar, and the one accepted pair (issue:28218 / issue:29159) is plausibly the same tokenizer bug from the titles alone. The canonical issue choice of issue:45020 is also reasonable as the broadest representative of the regression/loading-failure theme. I did not see an overstatement of duplicate evidence in the soft-edge verdicts." }, "evaluator_used": true, "retried": false, @@ -5360,7 +5321,7 @@ "component_index": 1, "component_count": 17, "review_unit_index": 17, - "review_unit_count": 50, + "review_unit_count": 49, "status": "reviewed", "reason": null, "source": "llm", @@ -5419,26 +5380,26 @@ ], "soft_pairs": [ "issue:44589|issue:45042", - "issue:43638|issue:44936", "issue:43746|issue:44811", "issue:43452|issue:44991", + "issue:43638|issue:44936", "issue:44387|issue:44617", "issue:45459|issue:45803", "issue:36010|issue:43645", - "issue:43550|issue:44912", "issue:43819|issue:43927", + "issue:43550|issue:44912", "issue:43854|issue:44977", "issue:43257|issue:43329", "issue:39692|issue:44912", "issue:43720|issue:44610", "issue:38175|issue:44610", - "issue:43454|issue:43577", "issue:36010|issue:36331", "issue:43531|issue:45042", + "issue:43454|issue:43577", "issue:43257|issue:43696", "issue:43262|issue:44811", - "issue:43329|issue:44479", "issue:42490|issue:43257", + "issue:43329|issue:44479", "issue:43643|issue:45759", "issue:43749|issue:44756", "issue:43329|issue:43450", @@ -5448,20 +5409,20 @@ "issue:43257|issue:43931", "issue:44849|issue:45701" ], - "prepared_review_unit_hash": "a7350aab048b85191ff87c2820f241f62f83f77a043af8d4ea088b33acb607a3", + "prepared_review_unit_hash": "84dba4b135a6b0e21decee58db714d8cf58a6e7292e4cc5680e664d20ecfd4c8", "component_budget": { "node_count": 426, "item_count": 426, - "soft_pair_count": 1572, - "serialized_chars": 814130, - "estimated_input_tokens": 203533, - "estimated_eval_tokens": 407322 + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 29, - "serialized_chars": 36616, + "serialized_chars": 36615, "estimated_input_tokens": 9154, "estimated_eval_tokens": 18564 }, @@ -5470,186 +5431,186 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is not a true duplicate set: it mixes many unrelated Transformers issues across training, loading, tokenization, multimodal/video, audio, quantization, and offload paths. The soft edges mostly reflect broad subsystem similarity, not the same underlying bug.", - "confidence": 0.97, - "canonical_issue_reason": "No canonical issue\u2014the items are distinct bug reports with different root causes and code paths.", + "summary": "This cluster is highly heterogeneous: most items are unrelated model-loading, tokenizer, multimodal, training, or memory bugs. I found no convincing duplicate core; the closest theme is broad tokenizer/regression reports, but even those are different code paths and failure modes.", + "confidence": 0.33, + "canonical_issue_reason": "Issue 45701 is the broadest representative item here: it describes a version-dependent tokenization regression with multiple downstream reports, so it best captures the widest cross-cutting symptom in an otherwise mixed cluster.", "canonical_pr_reason": null, - "best_issue_reason": "No single issue is a good representative for the cluster because it is too heterogeneous.", + "best_issue_reason": "45701 is the best single issue candidate because it is broad, externally corroborated, and more representative of the cluster than the many highly specific model- or feature-specific bugs.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44589", "right": "issue:45042", "accept": false, - "reason": "Different failures: Float8 storage import error vs PIL image processor torchvision dependency." - }, - { - "left": "issue:43638", - "right": "issue:44936", - "accept": false, - "reason": "Unrelated trainer bugs: Deepspeed/ZeRO3 index error vs evaluate-after-train behavior." + "reason": "Float8 storage lookup failure and image processor torchvision dependency are different subsystems and different failure modes." }, { "left": "issue:43746", "right": "issue:44811", "accept": false, - "reason": "Different code paths: local checkpoint loading with PEFT vs Whisper batch_decode skip_special_tokens." + "reason": "Local checkpoint/PEFT loading bug and Whisper batch_decode special-token handling are unrelated code paths." }, { "left": "issue:43452", "right": "issue:44991", "accept": false, - "reason": "Both are loading/tokenizer problems, but the model targets and failure modes are different." + "reason": "Both involve loading, but one is gguf_file handling for AutoTokenizer/AutoModel and the other is a specific tokenizer regression for one model." + }, + { + "left": "issue:43638", + "right": "issue:44936", + "accept": false, + "reason": "Deepspeed ZeRO3 index error and trainer.evaluate() failing after train are distinct trainer bugs." }, { "left": "issue:44387", "right": "issue:44617", "accept": false, - "reason": "Both mention OOM, but one is an int4 quantization memory regression and the other is model-specific video OOM." + "reason": "Both mention OOM, but the model, trigger, and underlying cause are different." }, { "left": "issue:45459", "right": "issue:45803", "accept": false, - "reason": "Both involve bad exception handling, but in different functions with different symptoms and fixes." + "reason": "Both concern exception handling, but they affect different functions and different tokenizer error paths." }, { "left": "issue:36010", "right": "issue:43645", "accept": false, - "reason": "Different regressions: GenerationMixin import breakage vs Jupyter custom-model initialization." + "reason": "GenerationMixin import failure and Jupyter custom-model initialization regression are unrelated." }, { - "left": "issue:43550", - "right": "issue:44912", + "left": "issue:43819", + "right": "issue:43927", "accept": false, - "reason": "Unrelated issues: torch.compile/SDPA failure vs MXFP4 quantization loading fallback." + "reason": "DAC from_latents STE mismatch and DiaConfig token-ID loss on save/load are different model/config issues." }, { - "left": "issue:43819", - "right": "issue:43927", + "left": "issue:43550", + "right": "issue:44912", "accept": false, - "reason": "Different models and bugs: DAC forward/from_latents mismatch vs DiaConfig token ID persistence." + "reason": "torch.compile+SDPA failure and MXFP4 quantization fallback are separate inference/loading problems." }, { "left": "issue:43854", "right": "issue:44977", "accept": false, - "reason": "Both are model-specific inference failures, but GLM loading and Qwen3.5 flash-attention are separate problems." + "reason": "Both are Qwen-family reports, but one is a loading/test issue and the other is flash-attention generation behavior." }, { "left": "issue:43257", "right": "issue:43329", "accept": false, - "reason": "Different bugs: MoE weight conversion during accelerate/deepspeed loading vs undefined video token-counting variables." + "reason": "Qwen3 MoE weight conversion under accelerate/deepspeed and multimodal video token counting are different bugs." }, { "left": "issue:39692", "right": "issue:44912", "accept": false, - "reason": "Unrelated: SigLIP2 docs/example issues vs Qwen/gpt-oss quantization behavior." + "reason": "Documentation example errors and quantization fallback/OOM behavior are unrelated." }, { "left": "issue:43720", "right": "issue:44610", "accept": false, - "reason": "Different code paths: accelerate weight unpacking vs OmDet-Turbo processor image-size mismatch." + "reason": "Packed-weight loading in accelerate and processor image-size mismatch are not the same defect." }, { "left": "issue:38175", "right": "issue:44610", "accept": false, - "reason": "Different symptoms: zero probabilities in SigLIP2 vs incorrect processor output shape." - }, - { - "left": "issue:43454", - "right": "issue:43577", - "accept": false, - "reason": "Separate model bugs: lm_head weight tying vs dtype propagation for BLIP2/QFormer." + "reason": "Unexpected zero probabilities and input-resolution mismatch are different model issues." }, { "left": "issue:36010", "right": "issue:36331", "accept": false, - "reason": "Different areas: import error for GenerationMixin vs CustomTrainer compute_loss signature change." + "reason": "Importing GenerationMixin and custom compute_loss kwarg compatibility are unrelated." }, { "left": "issue:43531", "right": "issue:45042", "accept": false, - "reason": "No shared underlying bug; Qwen3-MoE sliding-window logic is unrelated to PIL/torchvision imports." + "reason": "Qwen3-MoE sliding_window behavior and PIL backend torchvision dependency are unrelated." + }, + { + "left": "issue:43454", + "right": "issue:43577", + "accept": false, + "reason": "Weight-tying failure in AyaVision and dtype propagation failure in Blip2 are different initialization bugs." }, { "left": "issue:43257", "right": "issue:43696", "accept": false, - "reason": "Both are Qwen/generation related only at a high level; MoE loading and GPT-oss OOM are unrelated." + "reason": "Qwen3 MoE loading/conversion and GPT-oss OOM are unrelated symptoms and fixes." }, { "left": "issue:43262", "right": "issue:44811", "accept": false, - "reason": "Both touch audio processors, but one is chat-template sampling rate and the other is batch_decode special-token handling." + "reason": "Audio chat-template sampling-rate default and Whisper batch_decode special-token handling are different processor behaviors." }, { - "left": "issue:43329", - "right": "issue:44479", + "left": "issue:42490", + "right": "issue:43257", "accept": false, - "reason": "Both involve video multimodal paths, but one is an undefined-variable bug in token counting and the other is a broader video-input regression." + "reason": "Local-path trailing-slash warning and Qwen3 MoE loading are unrelated." }, { - "left": "issue:42490", - "right": "issue:43257", + "left": "issue:43329", + "right": "issue:44479", "accept": false, - "reason": "Unrelated: trailing-slash local model import warning vs Qwen3 MoE conversion/loading." + "reason": "Both are video/VLM related, but one is an undefined-variable bug in token counting and the other is a broader model-specific regression; not enough evidence they are the same fix." }, { "left": "issue:43643", "right": "issue:45759", "accept": false, - "reason": "Different config-loading problems: missing fields with trust_remote_code vs not unwrapping text_config in from_config." + "reason": "Missing config fields with trust_remote_code and unwrapping text_config for composite multimodal configs are different config-loading bugs." }, { "left": "issue:43749", "right": "issue:44756", "accept": false, - "reason": "Both concern loading/memory, but FSDP CPU-efficient loading and Strix Halo mmap OOM are different issues." + "reason": "FSDP CPU RAM efficient loading and disabling mmap on Strix Halo are different memory-management problems." }, { "left": "issue:43329", "right": "issue:43450", "accept": false, - "reason": "Both touch video processing, but token counting with undefined names is not the same as batched video shape handling." + "reason": "Video token-count helper bug and batched video processor shape bug are different layers of the stack." }, { "left": "issue:34567", "right": "issue:40990", "accept": false, - "reason": "Completely different bugs: trainer token-count tracking vs GPT-oss perplexity." + "reason": "Trainer token-count tracking and GPT-oss perplexity are unrelated." }, { "left": "issue:43606", "right": "issue:45859", "accept": false, - "reason": "Both mention CPU offload, but the affected models and failure mechanisms differ." + "reason": "CPU offload device mismatch and rotary embedding CPU-offload incompatibility are related only by offload context, not the same code-path bug." }, { "left": "issue:44568", "right": "issue:45356", "accept": false, - "reason": "Both are tokenizer regressions, but they affect different models and have different mechanics." + "reason": "Tokenizer special-token handling for mdeberta and Kimi codec handling are different tokenizer regressions." }, { "left": "issue:43257", "right": "issue:43931", "accept": false, - "reason": "Different loading failures: Qwen3 MoE conversion under accelerate/deepspeed vs Qwen3-VL weight-shape mismatch." + "reason": "Qwen3 MoE accelerate/deepspeed conversion failure and Qwen3-VL weight-shape mismatch are different loading failures." }, { "left": "issue:44849", "right": "issue:45701", "accept": false, - "reason": "Same general tokenization area, but output_hidden_states in Qwen3.5 is not the same as a broad tokenization-version change." + "reason": "Qwen3.5 hidden-states bug and broad tokenization-version regression are different APIs and failure modes." } ] }, @@ -5665,7 +5626,7 @@ "component_index": 1, "component_count": 17, "review_unit_index": 18, - "review_unit_count": 50, + "review_unit_count": 49, "status": "reviewed", "reason": null, "source": "llm", @@ -5680,6 +5641,7 @@ "issue:33357", "issue:34567", "issue:34689", + "issue:38175", "issue:41720", "issue:42175", "issue:42915", @@ -5694,6 +5656,7 @@ "issue:43742", "issue:43749", "issue:43756", + "issue:43782", "issue:43824", "issue:43906", "issue:43927", @@ -5703,8 +5666,8 @@ "issue:44488", "issue:44560", "issue:44568", - "issue:44737", "issue:44821", + "issue:44898", "issue:44912", "issue:44938", "issue:44987", @@ -5717,21 +5680,20 @@ "issue:45405", "issue:45488", "issue:45538", - "issue:45588", "issue:45907", "issue:45987" ], "soft_pairs": [ "issue:44568|issue:45488", "issue:29159|issue:34567", - "issue:42915|issue:43742", "issue:44291|issue:45042", + "issue:42915|issue:43742", "issue:42915|issue:44821", "issue:44987|issue:45538", - "issue:43452|issue:43582", - "issue:42915|issue:45538", - "issue:41720|issue:43540", "issue:43697|issue:45357", + "issue:41720|issue:43540", + "issue:42915|issue:45538", + "issue:43452|issue:43582", "issue:43927|issue:44912", "issue:44479|issue:45381", "issue:33357|issue:34567", @@ -5742,245 +5704,252 @@ "issue:43257|issue:43756", "issue:44987|issue:45907", "issue:44062|issue:44991", - "issue:43577|issue:45072", "issue:24643|issue:29159", "issue:45137|issue:45405", + "issue:43577|issue:45072", "issue:28282|issue:34689", "issue:43742|issue:44291", "issue:42175|issue:44488", - "issue:42915|issue:43749", "issue:42175|issue:44938", + "issue:42915|issue:43749", "issue:43906|issue:45381", - "issue:43404|issue:45987", "issue:28282|issue:29127", - "issue:44737|issue:45588" + "issue:43404|issue:45987", + "issue:38175|issue:44898", + "issue:43582|issue:43782" ], - "prepared_review_unit_hash": "6661babc874433a1a082b9ad04b5c0ae0bd9ae6d9a7b5d5d83f7b5532faf32ff", + "prepared_review_unit_hash": "0ac98a271c4a3f700608e1a5113758e90d14895c0fe2aec50f95700577aac904", "component_budget": { "node_count": 426, "item_count": 426, - "soft_pair_count": 1572, - "serialized_chars": 814130, - "estimated_input_tokens": 203533, - "estimated_eval_tokens": 407322 + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 }, "budget": { - "node_count": 47, - "item_count": 47, - "soft_pair_count": 32, - "serialized_chars": 36811, - "estimated_input_tokens": 9203, - "estimated_eval_tokens": 18662 + "node_count": 48, + "item_count": 48, + "soft_pair_count": 33, + "serialized_chars": 37643, + "estimated_input_tokens": 9411, + "estimated_eval_tokens": 19078 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Most items are unrelated to each other; the only clear duplicate pair is the Qwen video-input regression around v5.3.0. I\u2019d keep issue 44479 as the canonical report and treat 45381 as its narrower duplicate.", - "confidence": 0.72, - "canonical_issue_reason": "44479 is the broad umbrella report for the shared Qwen video-input regression across several model variants, so it best captures the underlying bug family.", + "summary": "These are mostly unrelated issues across different subsystems. Only the Qwen2.5-VL / Qwen3 VL video-input regression pair looks like the same underlying bug; the rest are false-positive similarity matches.", + "confidence": 0.86, + "canonical_issue_reason": "Issue 44479 is the best umbrella representative here: it describes a concrete 5.3.0 video-input regression across multiple Qwen VL variants, and it cleanly covers the one true duplicate-like pair in this set.", "canonical_pr_reason": null, - "best_issue_reason": "44479 is the most representative and actionable issue in the cluster because it names the regression broadly and aligns with the one clear duplicate pair.", + "best_issue_reason": "Issue 44479 is the strongest canonical candidate because it is broad, version-specific, and well-scoped to the shared Qwen video-input regression path.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44568", "right": "issue:45488", "accept": false, - "reason": "Both are tokenizer regressions, but they involve different tokenizer classes and failure modes; not the same concrete bug." + "reason": "Both are tokenizer-related, but they describe different bugs on different models and code paths." }, { "left": "issue:29159", "right": "issue:34567", "accept": false, - "reason": "Tokenizer slow/fast inconsistency and TrainerState token counting are unrelated subsystems and bugs." + "reason": "One is tokenizer inconsistency; the other is a TrainerState token-count tracking issue." }, { - "left": "issue:42915", - "right": "issue:43742", + "left": "issue:44291", + "right": "issue:45042", "accept": false, - "reason": "Different model-loading failures with different triggers and code paths." + "reason": "Unexpected init-empty-weights argument handling vs image processor torchvision dependency are unrelated." }, { - "left": "issue:44291", - "right": "issue:45042", + "left": "issue:42915", + "right": "issue:43742", "accept": false, - "reason": "One is an init_empty_weights constructor issue; the other is a PIL/torchvision backend requirement regression." + "reason": "Qwen3MoE FP8 failure and MobileLLM key error are different model-loading problems." }, { "left": "issue:42915", "right": "issue:44821", "accept": false, - "reason": "FineGrainedFP8Config loading and AutoImageProcessor-from-URL are unrelated." + "reason": "FineGrainedFP8Config failure is unrelated to loading AutoImageProcessor from URL." }, { "left": "issue:44987", "right": "issue:45538", "accept": false, - "reason": "These are different tokenizer/model_max_length issues with no clear shared code-path." + "reason": "Tokenizer/model-loading failure for one repo does not match the CLIPTokenizer model_max_length bug." }, { - "left": "issue:43452", - "right": "issue:43582", + "left": "issue:43697", + "right": "issue:45357", "accept": false, - "reason": "One is a gguf/tokenizer/model loading breakage; the other is an Apple Silicon caching allocator TypeError." + "reason": "Different regressions: one changes RTDetr outputs, the other saves incorrect Qwen3.5 visual encoder keys." }, { - "left": "issue:42915", - "right": "issue:45538", + "left": "issue:41720", + "right": "issue:43540", "accept": false, - "reason": "Different features and symptoms; only a broad Transformers version complaint in common." + "reason": "CUDA assert during auto device mapping is a different failure from Qwen3OmniMoe video input parsing." }, { - "left": "issue:41720", - "right": "issue:43540", + "left": "issue:42915", + "right": "issue:45538", "accept": false, - "reason": "Both mention Qwen and runtime failures, but the affected model families and failure paths differ." + "reason": "FP8 config loading failure and CLIPTokenizer max-length behavior are unrelated." }, { - "left": "issue:43697", - "right": "issue:45357", + "left": "issue:43452", + "right": "issue:43582", "accept": false, - "reason": "RTDetr output drift and Qwen3.5 save_pretrained key handling are unrelated." + "reason": "GGUF loading/tokenizer issues and AppleSilicon allocator warmup TypeError are different code paths." }, { "left": "issue:43927", "right": "issue:44912", "accept": false, - "reason": "DiaConfig save/load token IDs and MXFP4 quantization fallback are different bugs." + "reason": "DiaConfig save/load token-ID loss is unrelated to MXFP4 quantization falling back to bf16." }, { "left": "issue:44479", "right": "issue:45381", "accept": true, - "reason": "Both describe the same Qwen video-input regression; 45381 is a narrower symptom report of the broader 44479 issue." + "reason": "Both point to the same 5.3.0 Qwen2.5-VL video-input regression; 45381 is a narrower symptom of the broader bug in 44479." }, { "left": "issue:33357", "right": "issue:34567", "accept": false, - "reason": "MacOS CLIP bus error and token-count tracking are unrelated." + "reason": "MacOS bus error on a CLIP community model is unrelated to TrainerState token counting." }, { "left": "issue:42175", "right": "issue:43824", "accept": false, - "reason": "Backend packaging/install issue versus missing model import; not the same bug." + "reason": "Backend packaging dependency mismatch and missing Qwen2.5-VL symbol are different issues." }, { "left": "issue:43540", "right": "issue:44479", "accept": false, - "reason": "Both are video-input related, but the model family and reported failure differ enough that they do not clearly share one concrete bug." + "reason": "Different Qwen video bugs: one is a ValueError processing video inputs, the other is a broader v5.3.0 regression report." }, { "left": "issue:43317", "right": "issue:45538", "accept": false, - "reason": "Device-map/offload loading failure and CLIP tokenizer max-length bug are unrelated." + "reason": "Device-map/offload dequantization failure is unrelated to CLIPTokenizer max-length behavior." }, { "left": "issue:43540", "right": "issue:44560", "accept": false, - "reason": "Different Qwen video-processing failures with different model variants and symptoms." + "reason": "Both involve video inputs, but they are different models and symptoms." }, { "left": "issue:43257", "right": "issue:43756", "accept": false, - "reason": "Qwen3 MoE weight conversion and Smollm3 RoPE-layer handling are separate model-specific issues." + "reason": "Qwen3 MoE weight-conversion failure is unrelated to Smollm3 RoPE-layer dropping." }, { "left": "issue:44987", "right": "issue:45907", "accept": false, - "reason": "Model loading regression and a repo listing API forwarding bug are unrelated." + "reason": "Model loading regression and list_repo_files revision forwarding are unrelated." }, { "left": "issue:44062", "right": "issue:44991", "accept": false, - "reason": "The latter may surface a tokenizer-loading problem, but the titles do not establish the same underlying AddedToken bug." - }, - { - "left": "issue:43577", - "right": "issue:45072", - "accept": false, - "reason": "Both concern dtype behavior, but on different models and code paths; not mergeable as one PR/bug." + "reason": "AddedToken special-flag duplication and tokenizer loading failure for EMBEDDIA/est-roberta are not the same bug." }, { "left": "issue:24643", "right": "issue:29159", "accept": false, - "reason": "DeepSpeed training weight shape error and tokenizer inconsistency are unrelated." + "reason": "DeepSpeed training weight shape error is unrelated to tokenizer fast/slow inconsistency." }, { "left": "issue:45137", "right": "issue:45405", "accept": false, - "reason": "ZeRO3 deque error and PEFT version pinning are unrelated." + "reason": "DeepSpeed ZeRO3 deque error and PEFT minimum-version bump are unrelated." + }, + { + "left": "issue:43577", + "right": "issue:45072", + "accept": false, + "reason": "Both mention dtype, but they affect different model families and different loading/inference paths." }, { "left": "issue:28282", "right": "issue:34689", "accept": false, - "reason": "Missing PyTorch dependency import error and Llama 3.2 vision loading regression are different issues." + "reason": "Missing PyTorch import error is not the same as Llama 3.2 Vision model-loading breakage." }, { "left": "issue:43742", "right": "issue:44291", "accept": false, - "reason": "MobileLLM key error and init_empty_weights argument mismatch are unrelated loader failures." + "reason": "MobileLLM key error and init_empty_weights argument incompatibility are different failures." }, { "left": "issue:42175", "right": "issue:44488", "accept": false, - "reason": "Package extras/backend composition issue versus a model-loading failure for a specific checkpoint." + "reason": "Packaging/backend selection issue and model loading failure for cjvt/sleng-bert are unrelated." }, { - "left": "issue:42915", - "right": "issue:43749", + "left": "issue:42175", + "right": "issue:44938", "accept": false, - "reason": "Qwen3Moe FP8 config failure and FSDP CPU RAM efficient loading are different features and failures." + "reason": "Packaging backend contents and Python 3.14 load failure are different problems." }, { - "left": "issue:42175", - "right": "issue:44938", + "left": "issue:42915", + "right": "issue:43749", "accept": false, - "reason": "Packaging/backends issue and Python 3.14 load failure are not the same underlying bug." + "reason": "Qwen3MoE FP8 failure is unrelated to FSDP CPU RAM efficient loading." }, { "left": "issue:43906", "right": "issue:45381", "accept": false, - "reason": "The issue title suggests a reproduction of another bug, but there is no evidence it is the same Qwen video regression." + "reason": "An isolated reproduction of another issue is not the same as the Qwen2.5-VL video-position-id bug." + }, + { + "left": "issue:28282", + "right": "issue:29127", + "accept": false, + "reason": "AutoModel missing PyTorch and LayoutLMv3 box-validation messaging are unrelated." }, { "left": "issue:43404", "right": "issue:45987", "accept": false, - "reason": "Mistral3 lm_head tying and StaticCache sequence length contract are unrelated." + "reason": "Mistral3 lm_head tying bug and StaticCache sequence-length contract bug are different code paths." }, { - "left": "issue:28282", - "right": "issue:29127", + "left": "issue:38175", + "right": "issue:44898", "accept": false, - "reason": "ImportError for missing PyTorch and LayoutLMv3 box-clarity error are unrelated." + "reason": "Siglip2 zero-probability outputs and Perceiver resize failures are different image-model bugs." }, { - "left": "issue:44737", - "right": "issue:45588", + "left": "issue:43582", + "right": "issue:43782", "accept": false, - "reason": "XLNet CPU placement bug and flash_attention s_aux=None crash are unrelated." + "reason": "AppleSilicon allocator TypeError and Qwen3VL weight_only load error are unrelated." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is grounded in the packet: 44479 is the broad umbrella Qwen video-input regression, and 45381 is a narrower report that matches the same bug family closely enough to be treated as the clear duplicate pair. The other soft-edge verdicts stay conservative and do not overclaim identity from title-level evidence alone." + "feedback": "The summary is grounded in the packet: most pairs are clearly cross-subsystem mismatches, and the single accepted Qwen video-input pair is the only plausible duplicate-like match. The soft-edge verdicts are conservative overall." }, "evaluator_used": true, "retried": false, @@ -5993,7 +5962,7 @@ "component_index": 1, "component_count": 17, "review_unit_index": 19, - "review_unit_count": 50, + "review_unit_count": 49, "status": "reviewed", "reason": null, "source": "llm", @@ -6005,20 +5974,19 @@ "issue:29127", "issue:34567", "issue:36331", - "issue:38175", "issue:41762", "issue:43452", "issue:43493", "issue:43526", "issue:43531", "issue:43540", - "issue:43582", "issue:43643", + "issue:43688", "issue:43697", "issue:43720", "issue:43742", + "issue:43749", "issue:43756", - "issue:43782", "issue:43994", "issue:44062", "issue:44291", @@ -6026,6 +5994,7 @@ "issue:44514", "issue:44589", "issue:44625", + "issue:44737", "issue:44792", "issue:44821", "issue:44877", @@ -6043,6 +6012,7 @@ "issue:45381", "issue:45405", "issue:45479", + "issue:45588", "issue:45600", "issue:45823", "issue:45920", @@ -6050,11 +6020,10 @@ "issue:46170" ], "soft_pairs": [ - "issue:38175|issue:44898", - "issue:43582|issue:43782", + "issue:44737|issue:45588", "issue:43540|issue:45823", - "issue:43493|issue:43994", "issue:43526|issue:45479", + "issue:43493|issue:43994", "issue:44291|issue:44821", "issue:43756|issue:45325", "issue:34567|issue:36331", @@ -6068,186 +6037,193 @@ "issue:43531|issue:44821", "issue:43643|issue:45375", "issue:45405|issue:45600", - "issue:43720|issue:44898", "issue:43697|issue:45216", "issue:44625|issue:44877", + "issue:43720|issue:44898", "issue:24643|issue:29127", "issue:44514|issue:45381", + "issue:45372|issue:46170", "issue:41762|issue:44936", - "issue:45372|issue:46170" + "issue:43688|issue:43749" ], - "prepared_review_unit_hash": "511e3a086accf23ce4ef2d7f5b891f529b4c0522b2368e410c2005c794d23c7d", + "prepared_review_unit_hash": "2e7b5e427da48904ec8df9e9f1cd6ff18a2294f8ecf6757c7f10be2e50765bb8", "component_budget": { "node_count": 426, "item_count": 426, - "soft_pair_count": 1572, - "serialized_chars": 814130, - "estimated_input_tokens": 203533, - "estimated_eval_tokens": 407322 + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 }, "budget": { - "node_count": 47, - "item_count": 47, + "node_count": 48, + "item_count": 48, "soft_pair_count": 25, - "serialized_chars": 35055, - "estimated_input_tokens": 8764, - "estimated_eval_tokens": 17784 + "serialized_chars": 35649, + "estimated_input_tokens": 8913, + "estimated_eval_tokens": 18082 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a grab-bag of unrelated Transformers bug reports across different models and code paths; I don\u2019t see any true duplicate pair among the soft candidates. No PRs are present.", - "confidence": 0.86, - "canonical_issue_reason": "issue:45920 is the broadest and most central open report in the set, with multiple inbound references and a cross-model tokenizer regression theme.", + "summary": "This set is mostly a grab bag of unrelated bugs across training, config loading, tokenization, multimodal processors, and model-specific regressions. None of the soft edges is strong enough to merge as the same underlying defect, so I would not consolidate any of them.", + "confidence": 0.88, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "issue:45920 is the best representative issue because it is current, high-impact, and spans multiple model families rather than a single niche failure mode.", + "best_issue_reason": null, "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:38175", - "right": "issue:44898", - "accept": false, - "reason": "Both are vision-model failures, but they describe different symptoms and different model families (SigLIP2 outputs vs Perceiver image classification)." - }, - { - "left": "issue:43582", - "right": "issue:43782", + "left": "issue:44737", + "right": "issue:45588", "accept": false, - "reason": "Apple Silicon allocator warmup TypeError and Qwen3VL weight_only load failure are unrelated code paths and model stacks." + "reason": "Different code paths and symptoms: XLNet CPU-bound positional encoding vs a flash-attention AttributeError in sink-less models." }, { "left": "issue:43540", "right": "issue:45823", "accept": false, - "reason": "Video-input processing error in Qwen3OmniMoe is unrelated to the Gemma4 device-mapping forward mismatch." + "reason": "Different models and failures: Qwen3OmniMoe video input processing vs Gemma4 PLE device placement during forward." }, { - "left": "issue:43493", - "right": "issue:43994", + "left": "issue:43526", + "right": "issue:45479", "accept": false, - "reason": "Both mention SigLIP2, but one is a HF-vs-JAX implementation discrepancy and the other is a bad AutoModel/pipeline output report; not clearly the same concrete bug." + "reason": "One is an image processor label-reduction bug; the other is a sequence-classification loss issue with num_labels=1." }, { - "left": "issue:43526", - "right": "issue:45479", + "left": "issue:43493", + "right": "issue:43994", "accept": false, - "reason": "BeitImageProcessorFast reduce_labels bug is unrelated to the sequence-classification zero-loss regression." + "reason": "Both involve SigLIP2, but one is a HF-vs-JAX implementation discrepancy and the other is a specific AutoModel/pipeline symptom; not clearly the same bug." }, { "left": "issue:44291", "right": "issue:44821", "accept": false, - "reason": "An init_empty_weights/_is_hf_initialized loading error is unrelated to AutoImageProcessor-from-URL loading." + "reason": "Unrelated failures: init_empty_weights/_is_hf_initialized handling vs loading AutoImageProcessor from URL." }, { "left": "issue:43756", "right": "issue:45325", "accept": false, - "reason": "Both are RoPE-related, but they target different models and different failures (layer dropping vs temporal position scaling)." + "reason": "Both mention RoPE, but the bugs are model-specific and different: dropped RoPE layers vs wrong temporal position scaling." }, { "left": "issue:34567", "right": "issue:36331", "accept": false, - "reason": "TrainerState token counting not updating is a different issue from CustomTrainer.compute_loss receiving an unexpected argument." + "reason": "TrainerState token counting and CustomTrainer.compute_loss signature regression are different issues." }, { "left": "issue:44913", "right": "issue:46121", "accept": false, - "reason": "Both involve RoPE/config handling, but one is config persistence on reload and the other is a TypeError converting validation keys from JSON lists." + "reason": "Both touch rope config handling, but one is silent default reversion on reload and the other is a TypeError in validation for JSON-loaded lists." }, { "left": "issue:44062", "right": "issue:45081", "accept": false, - "reason": "AddedToken keyword conflict and Mistral regex patch crash are separate tokenizer-loading bugs." + "reason": "Different tokenizer bugs: AddedToken special-kwarg duplication vs backend_tokenizer missing during Mistral regex patching." }, { "left": "issue:44792", "right": "issue:44964", "accept": false, - "reason": "Janus test failure and Phi-4 multimodal loading failure are different multimodal model issues with different symptoms." + "reason": "Different models and failure modes: janus image-generation test failure vs Phi-4 multimodal load failure." }, { "left": "issue:45042", "right": "issue:45081", "accept": false, - "reason": "Image-processor torchvision dependency regression is unrelated to the Mistral tokenizer regex crash." + "reason": "Unrelated subsystems: PIL image processors requiring torchvision vs a Mistral tokenizer regex patch crash." }, { "left": "issue:45357", "right": "issue:45920", "accept": false, - "reason": "Qwen3.5 save_pretrained visual-key serialization and AutoTokenizer wrong token IDs are different regression classes." + "reason": "Different regression classes: Qwen3.5 save_pretrained key handling vs AutoTokenizer returning wrong token IDs across several models." }, { "left": "issue:43452", "right": "issue:44488", "accept": false, - "reason": "GGUF auto-loading breakage and a model-specific sleng-bert load failure do not look like the same underlying defect." + "reason": "Both are loading problems, but one is a gguf_file API break in auto loaders and the other is a specific model load failure; too broad to treat as one bug." }, { "left": "issue:43742", "right": "issue:44589", "accept": false, - "reason": "MobileLLM key error and missing Float8 storage object are distinct loading/runtime failures." + "reason": "Different load-time errors: MobileLLM key error vs missing Float8 storage object." }, { "left": "issue:43531", "right": "issue:44821", "accept": false, - "reason": "Qwen3-MoE sliding_window behavior and AutoImageProcessor URL loading are unrelated." + "reason": "Qwen3-MoE sliding_window behavior is unrelated to AutoImageProcessor URL loading." }, { "left": "issue:43643", "right": "issue:45375", "accept": false, - "reason": "Both are config-field issues, but one is generic trust_remote_code loading and the other is a specific Qwen3.5MoeVisionConfig field dropped by strict config handling." + "reason": "Both concern missing config fields, but one is generic AutoConfig remote-code loading and the other is a specific Qwen3.5 vision config field dropped by @strict." }, { - "left": "issue:43720", - "right": "issue:44898", + "left": "issue:45405", + "right": "issue:45600", "accept": false, - "reason": "BitNet packed-weight loading failure and Perceiver image-classification interpolation failure are unrelated." + "reason": "Unrelated maintenance issues: PEFT minimum-version bump vs stale auto_mappings references." }, { "left": "issue:43697", "right": "issue:45216", "accept": false, - "reason": "RTDetrV2 output drift in v5 and Qwen3.5 save_pretrained checkpoint corruption are different regressions." + "reason": "Different problems: RTDetrV2 output mismatch in v5 vs Qwen3.5 save_pretrained checkpoint corruption." }, { "left": "issue:44625", "right": "issue:44877", "accept": false, - "reason": "num_labels propagation bug and strict config blocking granite_speech loading are different configuration problems." + "reason": "Related only at a very high level; one is num_labels propagation between configs, the other is strict config loading for granite_speech." + }, + { + "left": "issue:43720", + "right": "issue:44898", + "accept": false, + "reason": "Different model families and code paths: BitNet packed-weight loading vs Perceiver interpolation with non-default resolution." }, { "left": "issue:24643", "right": "issue:29127", "accept": false, - "reason": "DeepSpeed weight dimensionality error and LayoutLMv3 box-info error are unrelated." + "reason": "Completely different reports: DeepSpeed training weight-shape runtime error vs LayoutLMv3 error-message clarity." }, { "left": "issue:44514", "right": "issue:45381", "accept": false, - "reason": "Both are Qwen2.5-VL-related, but one crashes in apply_chat_template batching while the other has wrong video vision_position_ids; different code paths." + "reason": "Same Qwen2.5-VL family, but one is a batched chat-template crash and the other is wrong video vision_position_ids; adjacent but not the same bug." + }, + { + "left": "issue:45372", + "right": "issue:46170", + "accept": false, + "reason": "Gemma 4 processor loading via mistral_common and Windows ROCm torch.distributed import issues are unrelated." }, { "left": "issue:41762", "right": "issue:44936", "accept": false, - "reason": "Gemma3 ZeRO-3 load failure and trainer.evaluate() after train() failure are not the same bug." + "reason": "Different trainer/model flows: Gemma3 ZeRO-3 load-time index error vs trainer.evaluate() failing after train()." }, { - "left": "issue:45372", - "right": "issue:46170", + "left": "issue:43688", + "right": "issue:43749", "accept": false, - "reason": "Gemma 4 processor import breakage from mistral_common is unrelated to the Windows ROCm torch.distributed dependency issue." + "reason": "Both involve training/loading infrastructure, but auxiliary-loss normalization and FSDP CPU RAM-efficient loading are distinct bugs." } ] }, @@ -6263,7 +6239,7 @@ "component_index": 1, "component_count": 17, "review_unit_index": 20, - "review_unit_count": 50, + "review_unit_count": 49, "status": "reviewed", "reason": null, "source": "llm", @@ -6321,23 +6297,22 @@ "issue:46050" ], "soft_pairs": [ - "issue:43688|issue:43749", - "issue:43452|issue:43577", "issue:44514|issue:45290", "issue:43452|issue:44843", "issue:44493|issue:46050", - "issue:44466|issue:45005", + "issue:43452|issue:43577", "issue:43819|issue:44811", + "issue:44466|issue:45005", "issue:44936|issue:45823", "issue:30064|issue:33666", "issue:44484|issue:44704", "issue:43295|issue:45356", + "issue:44464|issue:44743", "issue:43653|issue:44361", "issue:44361|issue:44442", - "issue:44464|issue:44743", "issue:28282|issue:32090", - "issue:28282|issue:33357", "issue:41720|issue:43906", + "issue:28282|issue:33357", "issue:38175|issue:43295", "issue:28282|issue:35141", "issue:45491|issue:46002", @@ -6347,239 +6322,236 @@ "issue:44843|issue:44991", "issue:43688|issue:45440", "issue:43646|issue:43749", - "issue:43688|issue:45701", "issue:29159|issue:33357", + "issue:43688|issue:45701", "issue:43531|issue:43742", "issue:43653|issue:43927", "issue:44464|issue:44977", "issue:43550|issue:43927" ], - "prepared_review_unit_hash": "d47338603703a55555b46d65172cc7092d8639c834817ac6fad0561d80fcc22c", + "prepared_review_unit_hash": "0ea78acd5e1a55eab854a1d5a9719e69f92f4cfa96ff4d54d723f37425939f87", "component_budget": { "node_count": 426, "item_count": 426, - "soft_pair_count": 1572, - "serialized_chars": 814130, - "estimated_input_tokens": 203533, - "estimated_eval_tokens": 407322 + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 32, - "serialized_chars": 37318, - "estimated_input_tokens": 9330, - "estimated_eval_tokens": 18916 + "soft_pair_count": 31, + "serialized_chars": 37025, + "estimated_input_tokens": 9257, + "estimated_eval_tokens": 18770 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Mostly unrelated issue set with a few superficial overlaps around tokenizers, tied weights, and v5 regressions; none of the soft pairs look like the same concrete bug/change.", - "confidence": 0.89, - "canonical_issue_reason": "Issue 43646 is the broadest, most representative regression report in the set: a Transformers v5 initialization break that can plausibly subsume several nearby v5-loading/initialization complaints.", + "summary": "This cluster is heterogeneous and does not contain a clear duplicate set. The strongest topical overlap is around tokenizer/load-path regressions, but the pairwise evidence is mostly too weak or points to different concrete failure modes.", + "confidence": 0.32, + "canonical_issue_reason": "issue:43452 is the broadest representative of the recurring from_pretrained/tokenizer-loading regression theme, and it\u2019s the best fit for a mixed cluster with no single clean duplicate family.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 43646 is the best standalone anchor because it is broad, regression-oriented, and central to the cluster\u2019s repeated v5 breakage theme.", + "best_issue_reason": "issue:43452 is the most central and broadly representative issue in the cluster, even though the cluster overall is not cohesive enough to support a tighter canonical duplicate.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:43688", - "right": "issue:43749", - "accept": false, - "reason": "Different bugs: aux-loss normalization vs FSDP CPU RAM-efficient loading." - }, - { - "left": "issue:43452", - "right": "issue:43577", - "accept": false, - "reason": "Different code paths and symptoms: gguf_file loading vs BLIP2 dtype propagation." - }, { "left": "issue:44514", "right": "issue:45290", "accept": false, - "reason": "Both touch apply_chat_template, but one is batched padding handling and the other is assistant tool-call messages with no content." + "reason": "Both are apply_chat_template crashes, but the triggers differ: batched input with padding=False vs assistant messages with tool calls and no content." }, { "left": "issue:43452", "right": "issue:44843", "accept": false, - "reason": "Both are loading-related, but gguf_file regression and offline model_info call are distinct failures." + "reason": "Same loader area, but different concrete bugs: gguf_file support vs unconditional model_info() in offline regex patching." }, { "left": "issue:44493", "right": "issue:46050", "accept": false, - "reason": "Both concern loading/saving, but position-id key warnings and sinq quantization breakage are different issues." + "reason": "Different symptoms and paths: unexpected position-id keys after save/load vs NaNs after sinq quantization round-trip." }, { - "left": "issue:44466", - "right": "issue:45005", + "left": "issue:43452", + "right": "issue:43577", "accept": false, - "reason": "Related to tied weights, but the concrete failures differ: device-dependent serialization vs translation-model tied-weight behavior." + "reason": "Different bugs: gguf_file/from_pretrained loading breakage vs BLIP2 dtype not propagating to qformer." }, { "left": "issue:43819", "right": "issue:44811", "accept": false, - "reason": "Different model behavior: DAC latent reconstruction mismatch vs Whisper batch_decode skip_special_tokens handling." + "reason": "Different subsystems: DAC latent reconstruction mismatch vs Whisper batch_decode skip_special_tokens handling." + }, + { + "left": "issue:44466", + "right": "issue:45005", + "accept": false, + "reason": "Both affect generation behavior, but one is compiled chunked generation and the other is tied-weights serialization." }, { "left": "issue:44936", "right": "issue:45823", "accept": false, - "reason": "Different training/inference problems: trainer lifecycle bug vs Gemma4 device_map forward mismatch." + "reason": "Different concrete problems: trainer lifecycle failure vs Gemma4 device mismatch during forward." }, { "left": "issue:30064", "right": "issue:33666", "accept": false, - "reason": "Both mention training/image models, but void segmentation-map processing and Qwen2-VL multi-GPU training are unrelated." + "reason": "Different tasks and failure modes: void segmentation-map processing vs Qwen2-VL multi-GPU training." }, { "left": "issue:44484", "right": "issue:44704", "accept": false, - "reason": "Different API surfaces: save_pretrained shard sizing vs AutoProcessor cached_file kwargs forwarding." + "reason": "Not the same bug: shard-size default question vs AutoProcessor not forwarding kwargs to cached_file." }, { "left": "issue:43295", "right": "issue:45356", "accept": false, - "reason": "Both are regressions, but one is processor.tokenizer/image-passing breakage and the other is tokenizer codec handling in Kimi-K2.5." + "reason": "Related tokenizer regressions, but different code paths and symptoms: processor.tokenizer/image passing vs Kimi-K2.5 codec/warning handling." + }, + { + "left": "issue:44464", + "right": "issue:44743", + "accept": false, + "reason": "Different generation bugs: compiled-forward chunked output inconsistency vs recurrent-state reset with cache." }, { "left": "issue:43653", "right": "issue:44361", "accept": false, - "reason": "Different tokenizer bugs: BigBird special-token registration vs MLuke tokenizer AttributeError." + "reason": "Different tokenizer-specific bugs: BigBird special-token decode issue vs MLukeTokenizer AttributeError on tasks." }, { "left": "issue:44361", "right": "issue:44442", "accept": false, - "reason": "Both are tokenizer load/task failures, but they affect different tokenizers and failure modes." - }, - { - "left": "issue:44464", - "right": "issue:44743", - "accept": false, - "reason": "Different generation bugs: compiled-forward chunked generation vs recurrent-state reset in modular_qwen3_5." + "reason": "Different failures: MLukeTokenizer task error vs AutoTokenizer loading FastSpeech2ConformerTokenizer." }, { "left": "issue:28282", "right": "issue:32090", "accept": false, - "reason": "ImportError for missing PyTorch is unrelated to Trainer NoneType broadcast errors." + "reason": "Unrelated: missing PyTorch ImportError vs distributed broadcast TypeError on NoneType." }, { - "left": "issue:28282", - "right": "issue:33357", + "left": "issue:41720", + "right": "issue:43906", "accept": false, - "reason": "Environment dependency error vs MacOS bus error on a CLIP model; no shared code defect." + "reason": "43906 is only an isolated reproduction reference; not enough evidence it is the same cudaErrorAssert bug." }, { - "left": "issue:41720", - "right": "issue:43906", + "left": "issue:28282", + "right": "issue:33357", "accept": false, - "reason": "Both involve Qwen3, but one is cudaErrorAssert with auto device mapping and the other is just a reproduction of another issue." + "reason": "Different issues: missing PyTorch dependency vs MacOS bus error on a community CLIP model." }, { "left": "issue:38175", "right": "issue:43295", "accept": false, - "reason": "Different symptom class: zero probabilities in SigLIP2 vs processor/tokenizer regression in v4.57.5." + "reason": "Different model-level bugs: zero probabilities in SigLIP2 vs processor.tokenizer regression." }, { "left": "issue:28282", "right": "issue:35141", "accept": false, - "reason": "Missing PyTorch import error is unrelated to tied embeddings being reinitialized after resizing." + "reason": "Completely different problems: missing PyTorch dependency vs output embedding reinitialization after resize." }, { "left": "issue:45491", "right": "issue:46002", "accept": false, - "reason": "Different numerical bugs: NaNs from sliding-window attention vs NaNs in classification heads after init/checkout." + "reason": "Both mention NaNs, but the mechanisms differ: sliding-window/all-padding Gemma3 embeddings vs classification-head NaNs after checkout/init." }, { "left": "issue:43653", "right": "issue:44568", "accept": false, - "reason": "Both tokenizer-related, but BigBird mask-token registration is not the same as missing BOS/EOS insertion in mdeberta-v3." + "reason": "Both involve special tokens, but they are different tokenizer behaviors and models." }, { "left": "issue:43577", "right": "issue:43697", "accept": false, - "reason": "Both are model output/dtype issues, but BLIP2 dtype loading and RTDetrV2 output divergence are different code paths." + "reason": "Different bugs: BLIP2 dtype propagation vs RTDetrV2 output divergence in v5." }, { "left": "issue:43688", "right": "issue:45704", "accept": false, - "reason": "Aux-loss normalization bug is unrelated to T5's apex.FusedRMSNorm memory leak." + "reason": "Different issues: MoE auxiliary-loss normalization vs T5 apex.FusedRMSNorm memory leak." }, { "left": "issue:44843", "right": "issue:44991", "accept": false, - "reason": "Both are tokenizer loading regressions, but one is offline model_info behavior and the other is a specific tokenizer load failure in v5." + "reason": "Both are tokenizer-loading regressions, but the concrete causes differ." }, { "left": "issue:43688", "right": "issue:45440", "accept": false, - "reason": "Different model families and bugs: MoE aux-loss normalization vs DeepseekV3MoE divergence from remote implementation." + "reason": "Different correctness bugs: MoE auxiliary-loss normalization vs DeepseekV3MoE divergence from remote implementation." }, { "left": "issue:43646", "right": "issue:43749", "accept": false, - "reason": "Broad v5 regression vs FSDP CPU RAM-efficient loading; no concrete shared bug path." + "reason": "Different scopes: custom model initialization breakage in v5 vs FSDP CPU RAM efficient loading regression." }, { - "left": "issue:43688", - "right": "issue:45701", + "left": "issue:29159", + "right": "issue:33357", "accept": false, - "reason": "Aux-loss normalization is unrelated to version-dependent tokenization changes." + "reason": "Different failures: slow/fast tokenizer inconsistency vs MacOS bus error on CLIP model load." }, { - "left": "issue:29159", - "right": "issue:33357", + "left": "issue:43688", + "right": "issue:45701", "accept": false, - "reason": "Tokenizer inconsistency vs MacOS bus error; only a superficial shared theme of model preprocessing." + "reason": "Too broad and not the same bug: specific MoE auxiliary-loss normalization vs generic version-related tokenization changes." }, { "left": "issue:43531", "right": "issue:43742", "accept": false, - "reason": "Different failures: Qwen3-MoE sliding_window behavior vs MobileLLM-125M loading key error." + "reason": "Different models and failures: Qwen3-MoE sliding_window issue vs MobileLLM-125M key error on load." }, { "left": "issue:43653", "right": "issue:43927", "accept": false, - "reason": "Different tokenizer/config bugs: special-token registration vs DiaConfig losing custom token IDs on save/load." + "reason": "Different problems: BigBird special-token registration/decode vs DiaConfig losing custom token IDs after save/load." }, { "left": "issue:44464", "right": "issue:44977", "accept": false, - "reason": "Both affect generation, but compiled-forward chunking and flash-attention generation bugs are different concrete problems." + "reason": "Different generation bugs: compiled-forward chunked generation inconsistency vs Qwen3.5 flash-attention failure." }, { "left": "issue:43550", "right": "issue:43927", "accept": false, - "reason": "torch.compile + SDPA failure in Bamba is unrelated to DiaConfig save/load token ID loss." + "reason": "Different code paths: torch.compile + SDPA failure vs DiaConfig token-ID persistence bug." } ] }, - "evaluator_result": null, - "evaluator_used": false, - "retried": false, + "evaluator_result": { + "accept": false, + "feedback": "The cluster summary is broadly grounded, but one soft-edge verdict is too strong: issue:44514 vs issue:45290 are both apply_chat_template crashes, yet the packet only supports a loose topical similarity, not a clear same-bug duplicate. The rest of the rejects are conservative and well-aligned with the titles, but because the lone accepted pair overstates the evidence, I would reject this evaluation." + }, + "evaluator_used": true, + "retried": true, "accepted_nontrivial_soft_edge": false, "error_kind": null, "error_message": null @@ -6589,7 +6561,7 @@ "component_index": 1, "component_count": 17, "review_unit_index": 21, - "review_unit_count": 50, + "review_unit_count": 49, "status": "reviewed", "reason": null, "source": "llm", @@ -6650,17 +6622,17 @@ "issue:44479|issue:44625", "issue:43550|issue:44361", "issue:41720|issue:44464", - "issue:43577|issue:45588", "issue:43122|issue:43906", + "issue:43577|issue:45588", "issue:43577|issue:44554", "issue:32090|issue:34567", "issue:43504|issue:43720", "issue:43504|issue:43550", "issue:43504|issue:44610", - "issue:32090|issue:33666", "issue:44466|issue:46050", - "issue:33666|issue:36010", + "issue:32090|issue:33666", "issue:43742|issue:44821", + "issue:33666|issue:36010", "issue:45823|issue:46050", "issue:43452|issue:44451", "issue:43720|issue:43927", @@ -6670,27 +6642,27 @@ "issue:44849|issue:45357", "issue:43525|issue:43572", "issue:33666|issue:36331", - "issue:43720|issue:44361", "issue:44625|issue:46002", - "issue:43493|issue:43825", "issue:43756|issue:44913", - "issue:28218|issue:33357", - "issue:44568|issue:45812" + "issue:43720|issue:44361", + "issue:43493|issue:43825", + "issue:44568|issue:45812", + "issue:28218|issue:33357" ], - "prepared_review_unit_hash": "9b789b5e5c1b93fdfdb85fab044e7185e5e3b769f0bfb1490910cb4fd6236184", + "prepared_review_unit_hash": "1eec4a1d6d8ce9cebbdc05b0be572d114944f5cd5fb12c13400e6469ce31024d", "component_budget": { "node_count": 426, "item_count": 426, - "soft_pair_count": 1572, - "serialized_chars": 814130, - "estimated_input_tokens": 203533, - "estimated_eval_tokens": 407322 + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 }, "budget": { "node_count": 47, "item_count": 47, "soft_pair_count": 30, - "serialized_chars": 36215, + "serialized_chars": 36216, "estimated_input_tokens": 9054, "estimated_eval_tokens": 18364 }, @@ -6699,192 +6671,186 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These items are mostly unrelated standalone bug reports. No soft edge pair looks like the same concrete underlying bug, so I would not merge any of them. The strongest single issue to anchor on is the broad tokenizer regression report for Granite models.", - "confidence": 0.41, - "canonical_issue_reason": "issue:45812 is the most canonical-looking report: it describes a broad, explicit v4\u2192v5 regression with clear impact across all Granite models and strong evidence of a tokenizer-path bug.", + "summary": "This cluster is very heterogeneous: the candidate edges mostly connect unrelated regression reports across tokenization, loading, training, device placement, and model-specific behavior. I would not merge any of the soft pairs as duplicates.", + "confidence": 0.93, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "issue:45812 is the best issue candidate because it is broad, concrete, and likely to attract the most useful downstream fixing/triage work compared with narrower model-specific regressions.", + "best_issue_reason": null, "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43643", "right": "issue:44561", "accept": false, - "reason": "Both mention trust_remote_code-related breakage, but one is missing fields in AutoConfig output and the other is a v5 regression from removing is_torch_fx_available. Different failure modes and code paths." + "reason": "Both involve trust_remote_code, but they describe different failures: missing fields in AutoConfig return values vs. removal of is_torch_fx_available breaking remote-code models." }, { "left": "issue:44479", "right": "issue:44625", "accept": false, - "reason": "Video-input regression across several VL models versus num_labels propagation between configs. Different symptoms, components, and fixes." + "reason": "One is a video-input regression for Qwen VL models; the other is num_labels propagation from core config to text config. Different code paths." }, { "left": "issue:43550", "right": "issue:44361", "accept": false, - "reason": "One is a torch.compile/SDPA model failure; the other is an MLukeTokenizer AttributeError. No shared underlying bug." + "reason": "Bamba torch.compile/SDPA failure and MLukeTokenizer AttributeError are unrelated bugs." }, { "left": "issue:41720", "right": "issue:44464", "accept": false, - "reason": "Qwen3 auto device-map cudaErrorAssert and chunked-generation inconsistencies are separate runtime bugs with different triggers." + "reason": "cudaErrorAssert with auto device mapping is not the same as inconsistent chunked generation under compiled forward." }, { - "left": "issue:43577", - "right": "issue:45588", + "left": "issue:43122", + "right": "issue:43906", "accept": false, - "reason": "BF16/FP32 dtype mismatch in Blip2 is unrelated to flash_attention.py crashing on s_aux=None." + "reason": "One reports tokenization changes across versions; the other is just an isolated reproduction reference for a separate upstream issue." }, { - "left": "issue:43122", - "right": "issue:43906", + "left": "issue:43577", + "right": "issue:45588", "accept": false, - "reason": "One reports a tokenization regression across versions; the other is only an isolated reproduction of another issue. There is not enough evidence they are the same concrete bug." + "reason": "Blip2 dtype staying float32 and flash_attention crashing on s_aux=None are different failures." }, { "left": "issue:43577", "right": "issue:44554", "accept": false, - "reason": "Model dtype drift in Blip2 is unrelated to the MPS attention correctness issue with mismatched head dimensions." + "reason": "A dtype propagation bug is unrelated to the MPS attention correctness issue with differing value/query head dims." }, { "left": "issue:32090", "right": "issue:34567", "accept": false, - "reason": "Trainer GPU broadcast NoneType error and num_input_tokens_seen not updating are unrelated trainer issues." + "reason": "Trainer broadcast NoneType error and num_input_tokens_seen not updating are separate Trainer issues." }, { "left": "issue:43504", "right": "issue:43720", "accept": false, - "reason": "Legacy-field loading failure in BEiT and BitNet packed-weight loading failure are different model-loading bugs." + "reason": "Legacy preset loading for Beit and BitNet packed-weight loading during accelerate are different load paths and failure modes." }, { "left": "issue:43504", "right": "issue:43550", "accept": false, - "reason": "BEiT semantic-segmentation load failure and Bamba torch.compile/SDPA failure do not share the same code-path problem." + "reason": "Beit legacy-field loading bug is unrelated to Bamba torch.compile SDPA failure." }, { "left": "issue:43504", "right": "issue:44610", "accept": false, - "reason": "Legacy field loading for BEiT is unrelated to OmDet-Turbo processor producing the wrong image size." - }, - { - "left": "issue:32090", - "right": "issue:33666", - "accept": false, - "reason": "Trainer broadcast NoneType and Qwen2-VL multi-GPU training are different distributed-training issues." + "reason": "Loading a pretrained Beit preset and OmDet processor producing the wrong image size are unrelated." }, { "left": "issue:44466", "right": "issue:46050", "accept": false, - "reason": "Inconsistent lm_head.weight serialization across devices is not the same as sinq quantized model save/load breakage." + "reason": "lm_head.weight serialization depending on device and sinq quantized save/load breakage are both save/load themed but not the same bug." }, { - "left": "issue:33666", - "right": "issue:36010", + "left": "issue:32090", + "right": "issue:33666", "accept": false, - "reason": "Multi-GPU training for Qwen2-VL is unrelated to missing GenerationMixin import." + "reason": "Trainer GPU broadcast TypeError is unrelated to Qwen2-VL multi-GPU training behavior." }, { "left": "issue:43742", "right": "issue:44821", "accept": false, - "reason": "KeyError loading MobileLLM-125M and inability to load AutoImageProcessor from URL are different loader problems." + "reason": "A key error loading MobileLLM and an AutoImageProcessor URL-loading failure are different loader problems." + }, + { + "left": "issue:33666", + "right": "issue:36010", + "accept": false, + "reason": "Qwen2-VL multi-GPU training and a GenerationMixin import error do not share the same underlying cause." }, { "left": "issue:45823", "right": "issue:46050", "accept": false, - "reason": "Gemma4 device-mapping forward mismatch is a separate runtime issue from quantized model persistence failures." + "reason": "Gemma4 device_map auto forward mismatch is unrelated to quantized save/load corruption." }, { "left": "issue:43452", "right": "issue:44451", "accept": false, - "reason": "gguf_file breaking AutoTokenizer/AutoModel loading and inability to load ScandiBERT are not the same bug." + "reason": "gguf_file breaking AutoTokenizer/AutoModel loading and ScandiBERT load failure are distinct model-loading issues." }, { "left": "issue:43720", "right": "issue:43927", "accept": false, - "reason": "BitNet accelerate-loading unpacking failure and DiaConfig save/load token-ID loss are unrelated." + "reason": "BitNet packed-weight unpacking during load and DiaConfig losing custom token IDs on save/load are unrelated." }, { "left": "issue:43606", "right": "issue:43927", "accept": false, - "reason": "CPU offload device mismatch in bark-small is unrelated to DiaConfig losing custom token IDs." + "reason": "CPU offload device mismatch for bark-small is not the same as DiaConfig save/load token ID loss." }, { "left": "issue:43278", "right": "issue:43688", "accept": false, - "reason": "Embedding dtype changing during evaluate and auxiliary-loss normalization are different training/eval correctness issues." + "reason": "Embedding dtype changing on eval and auxiliary loss normalization are separate concerns." }, { "left": "issue:43262", "right": "issue:43295", "accept": false, - "reason": "Audio processor sampling-rate defaulting and processor.tokenizer/image-passing regression are separate processor API regressions." + "reason": "Audio chat-template sampling-rate default and processor.tokenizer/image passing regression are different processor bugs." }, { "left": "issue:44849", "right": "issue:45357", "accept": false, - "reason": "Qwen3.5 hidden-states bug and incorrect visual encoder keys in save_pretrained are distinct issues." + "reason": "Qwen3.5 output_hidden_states bug and incorrect visual encoder keys in save_pretrained are unrelated." }, { "left": "issue:43525", "right": "issue:43572", "accept": false, - "reason": "Both are missing-token-field regressions, but they affect different configs and different missing attributes; not the same concrete bug." + "reason": "Both mention missing pad token config fields, but they affect different model families and different attribute names; not the same concrete bug." }, { - "left": "issue:33666", - "right": "issue:36331", + "left": "issue:44625", + "right": "issue:46002", "accept": false, - "reason": "Qwen2-VL multi-GPU training and compute_loss unexpected num_items_in_batch are different trainer/distributed issues." + "reason": "num_labels propagation failure is not the same as NaNs in classification heads after checkout/init." }, { - "left": "issue:43720", - "right": "issue:44361", + "left": "issue:43756", + "right": "issue:44913", "accept": false, - "reason": "Packed-weight loading in BitNet and MLukeTokenizer AttributeError are unrelated." + "reason": "Smollm3 RoPE-layer dropping and GPTNeoX rotary_pct resetting on reload are different rotary/config issues." }, { - "left": "issue:44625", - "right": "issue:46002", + "left": "issue:43720", + "right": "issue:44361", "accept": false, - "reason": "num_labels propagation between configs and NaNs in classification heads upon init/checkout are different bugs." + "reason": "BitNet load failure and MLukeTokenizer AttributeError are unrelated." }, { "left": "issue:43493", "right": "issue:43825", "accept": false, - "reason": "SigLIP2 implementation discrepancy and pipeline translation-task error messaging are unrelated." + "reason": "SigLIP2 HF/JAX discrepancy and pipeline translation-task error messaging are completely different problems." }, { - "left": "issue:43756", - "right": "issue:44913", + "left": "issue:44568", + "right": "issue:45812", "accept": false, - "reason": "Smollm3 RoPE-layer count mismatch and GPTNeoX rotary_pct reset-on-reload are different configuration problems." + "reason": "These are both tokenizer regressions, but for different models and symptoms: missing BOS/EOS vs wrong token IDs." }, { "left": "issue:28218", "right": "issue:33357", "accept": false, - "reason": "Tokenizer added-space behavior and a MacOS bus error on CLIP model load are not the same underlying bug." - }, - { - "left": "issue:44568", - "right": "issue:45812", - "accept": false, - "reason": "Both are tokenizer regressions in v5, but one is a special-tokens/BOS-EOS issue for a specific model while the other is wrong token IDs across Granite models; different concrete code paths." + "reason": "Tokenizer extra-space behavior and a MacOS bus error on CLIP are not the same bug." } ] }, @@ -6900,7 +6866,7 @@ "component_index": 1, "component_count": 17, "review_unit_index": 22, - "review_unit_count": 50, + "review_unit_count": 49, "status": "reviewed", "reason": null, "source": "llm", @@ -6908,8 +6874,6 @@ "model": "gpt-5.4-mini?service_tier=flex", "cluster_id": "cluster-10000-426", "nodes": [ - "issue:29942", - "issue:30333", "issue:33357", "issue:33666", "issue:39692", @@ -6930,6 +6894,7 @@ "issue:43756", "issue:43824", "issue:43844", + "issue:43873", "issue:43927", "issue:43994", "issue:44112", @@ -6955,28 +6920,29 @@ "issue:45759", "issue:45803", "issue:45823", - "issue:45987" + "issue:45987", + "issue:46050" ], "soft_pairs": [ "issue:44964|issue:45405", "issue:39692|issue:43493", "issue:43540|issue:44977", - "issue:45200|issue:45229", - "issue:43493|issue:43844", "issue:43334|issue:43572", + "issue:43493|issue:43844", + "issue:45200|issue:45229", "issue:44488|issue:44843", "issue:43278|issue:43844", "issue:43257|issue:43299", - "issue:45803|issue:45987", "issue:43756|issue:45083", + "issue:45803|issue:45987", "issue:43334|issue:45759", "issue:39692|issue:43295", - "issue:45229|issue:45823", "issue:43646|issue:44466", "issue:43824|issue:44589", + "issue:45229|issue:45823", "issue:33357|issue:33666", - "issue:43504|issue:43606", "issue:44589|issue:45520", + "issue:43504|issue:43606", "issue:43572|issue:44291", "issue:43927|issue:44361", "issue:44464|issue:45584", @@ -6987,216 +6953,216 @@ "issue:44792|issue:45405", "issue:43994|issue:44493", "issue:44423|issue:44734", - "issue:29942|issue:30333" + "issue:43873|issue:46050" ], - "prepared_review_unit_hash": "85d82fbd3e37354e813d09b2a9ff6daff9e11bb640d531521f837f90ca07a360", + "prepared_review_unit_hash": "c3b704b2219eb3d8926b2705f2b89670ba916edc7b6739f4eecdceb8fd64653b", "component_budget": { "node_count": 426, "item_count": 426, - "soft_pair_count": 1572, - "serialized_chars": 814130, - "estimated_input_tokens": 203533, - "estimated_eval_tokens": 407322 + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 30, - "serialized_chars": 36845, - "estimated_input_tokens": 9212, - "estimated_eval_tokens": 18680 + "serialized_chars": 36850, + "estimated_input_tokens": 9213, + "estimated_eval_tokens": 18682 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is a loose collection of distinct Transformers regression reports, mostly around model loading, multimodal configs, serving, and tokenizer/config edge cases. None of the soft pairs look like true duplicates or the same fixable code-path bug, so I would not merge any of them.", - "confidence": 0.9, - "canonical_issue_reason": "issue:43646 is the broadest and most representative issue here: a Transformers 5.0 initialization regression affecting custom model setup, which matches the dominant theme of version-related model-loading failures in this set.", + "summary": "This cluster is heterogeneous: the soft pairs mostly connect unrelated model-loading, config, quantization, offload, and CI issues rather than true duplicates. I would not merge any of the suggested soft edges; if forced to pick a representative issue, Gemma4 multi-GPU OOM is the strongest anchor.", + "confidence": 0.78, + "canonical_issue_reason": "issue:45229 is the best canonical anchor because it has the clearest concrete runtime failure, the most discussion, and broad user impact compared with the more niche config/docs regressions.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43646 is the best single issue to anchor the cluster because it is general, user-facing, and central to the repeated 5.0/custom-loading regression pattern.", + "best_issue_reason": "issue:45229 is the most suitable overall representative for this cluster; the other items are mostly isolated model-specific or release-specific bugs, not one underlying duplicate.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44964", "right": "issue:45405", "accept": false, - "reason": "One is a Phi-4 multimodal model-loading failure; the other is a PEFT version bump/release problem. Different bugs and code paths." + "reason": "Phi-4 multimodal loading regression vs an unreleased PEFT version bump; different problems." }, { "left": "issue:39692", "right": "issue:43493", "accept": false, - "reason": "Both mention SigLIP2, but one is a broken docs example and the other is an implementation discrepancy vs JAX. Not the same defect." + "reason": "Both are SigLIP2-related, but one is a docs/example mismatch and the other is an implementation discrepancy with JAX." }, { "left": "issue:43540", "right": "issue:44977", "accept": false, - "reason": "Qwen3OmniMoe video-processing error vs Qwen3.5 flash-attention generation failure. Different model families and failures." + "reason": "Qwen3 video-input ValueError vs Qwen3.5 flash-attention generation behavior; not the same bug." }, { - "left": "issue:45200", - "right": "issue:45229", + "left": "issue:43334", + "right": "issue:43572", "accept": false, - "reason": "Defaulting mm_token_type_ids is unrelated to Gemma4 multi-GPU CUDA OOM during inference." + "reason": "Both mention missing pad-token fields after 5.0, but they affect different model configs and code paths." }, { "left": "issue:43493", "right": "issue:43844", "accept": false, - "reason": "SigLIP2 accuracy/discrepancy issue vs a separate ZeRO-3 gradient growth bug. Different symptoms and layers." + "reason": "SigLIP2 implementation mismatch vs ZeRO-3 gradient growth; unrelated failure modes." }, { - "left": "issue:43334", - "right": "issue:43572", + "left": "issue:45200", + "right": "issue:45229", "accept": false, - "reason": "Both are config-attribute regressions, but for different models and different missing fields. Not the same underlying bug." + "reason": "Text-only fine-tuning token defaults vs Gemma4 multi-GPU inference OOM; different issues." }, { "left": "issue:44488", "right": "issue:44843", "accept": false, - "reason": "A specific model load failure is not the same as the HF_HUB_OFFLINE tokenizer patch bug." + "reason": "Model loading failure vs offline hub lookup regression; not the same underlying bug." }, { "left": "issue:43278", "right": "issue:43844", "accept": false, - "reason": "Training/eval dtype drift is unrelated to gradient explosion under HfDeepSpeedConfig + ZeRO-3." + "reason": "Embedding dtype drift vs DeepSpeed gradient anomaly; different code paths." }, { "left": "issue:43257", "right": "issue:43299", "accept": false, - "reason": "Both involve Qwen3 MOE, but one is accelerate/deepspeed weight conversion and the other is dev-version loading regression. Different fixes." + "reason": "Both are Qwen3 MOE-related, but one is accelerate/deepspeed conversion and the other is a dev loading regression for Qwen3VL." }, { - "left": "issue:45803", - "right": "issue:45987", + "left": "issue:43756", + "right": "issue:45083", "accept": false, - "reason": "A bare-except bug in tensor conversion is unrelated to StaticCache returning a tensor instead of an int." + "reason": "Smollm3 RoPE-layer bug vs qwen3_omni_moe feature-length helper behavior; unrelated models." }, { - "left": "issue:43756", - "right": "issue:45083", + "left": "issue:45803", + "right": "issue:45987", "accept": false, - "reason": "RoPE-layer dropping in Smollm3 and a helper-length calculation issue in qwen3_omni_moe are unrelated." + "reason": "Bare except in Fuyu tensor conversion vs StaticCache return-type mismatch; unrelated." }, { "left": "issue:43334", "right": "issue:45759", "accept": false, - "reason": "Both are multimodal config/loading issues, but one is a missing pad_token_id and the other is from_config not unwrapping text_config. Different code paths." + "reason": "Different Qwen multimodal config bugs: missing pad_token_id vs text_config unwrapping in from_config." }, { "left": "issue:39692", "right": "issue:43295", "accept": false, - "reason": "Doc/example problems for SigLIP2 are not the same as the processor.tokenizer regression in custom multimodal code." - }, - { - "left": "issue:45229", - "right": "issue:45823", - "accept": false, - "reason": "CUDA OOM from multi-GPU inference is not the same as a device mismatch during forward with device_map=auto." + "reason": "Docs/example mismatch in SigLIP2 vs processor.tokenizer regression; different bugs." }, { "left": "issue:43646", "right": "issue:44466", "accept": false, - "reason": "Custom model initialization regression and inconsistent lm_head serialization are separate v5 issues." + "reason": "Custom model initialization breakage vs lm_head.weight serialization inconsistency; not the same code-path." }, { "left": "issue:43824", "right": "issue:44589", "accept": false, - "reason": "Importing a missing Qwen2.5 VL class is unrelated to Float8 storage lookup failures." + "reason": "Qwen2.5-VL import failure vs float8 storage lookup error; unrelated." }, { - "left": "issue:33357", - "right": "issue:33666", + "left": "issue:45229", + "right": "issue:45823", "accept": false, - "reason": "MacOS bus error on CLIP is unrelated to Qwen2-VL multi-GPU training." + "reason": "Both are Gemma4, but one is multi-GPU inference OOM and the other is PLE device mismatch in forward." }, { - "left": "issue:43504", - "right": "issue:43606", + "left": "issue:33357", + "right": "issue:33666", "accept": false, - "reason": "Legacy config-field loading for BEiT and device-mismatch under CPU offload in Bark are different bugs." + "reason": "MacOS CLIP bus error vs Qwen2-VL multi-GPU training; unrelated." }, { "left": "issue:44589", "right": "issue:45520", "accept": false, - "reason": "Float8 storage support error and Python 3.13 flash_attn KeyError are separate compatibility issues." + "reason": "Float8 storage object error vs Python 3.13 flash_attn import key error; different subsystem." + }, + { + "left": "issue:43504", + "right": "issue:43606", + "accept": false, + "reason": "Legacy field loading issue in Beit vs CPU-offload device mismatch in Bark; unrelated." }, { "left": "issue:43572", "right": "issue:44291", "accept": false, - "reason": "Both are v5 init/loading regressions, but one is a missing config field and the other is an unexpected constructor argument. Not the same fix." + "reason": "Missing pad_token_idx vs unexpected _is_hf_initialized arg; both are release regressions but not the same bug." }, { "left": "issue:43927", "right": "issue:44361", "accept": false, - "reason": "Custom token ID loss after save/load and MLuke tokenizer task failures are unrelated." + "reason": "DiaConfig token persistence loss vs MLukeTokenizer AttributeError; distinct failures." }, { "left": "issue:44464", "right": "issue:45584", "accept": false, - "reason": "Chunked-generation inconsistency with compiled forward is unrelated to Whisper empty-transcription generation after align_special_tokens." + "reason": "Compiled-forward chunked generation inconsistency vs Whisper empty transcription failure after token alignment." }, { "left": "issue:43606", "right": "issue:44112", "accept": false, - "reason": "A real device mismatch bug is not the same as a stale CI test for GraniteSpeech." + "reason": "Both are CI/device-related, but they concern different models and failure mechanisms." }, { "left": "issue:43504", "right": "issue:43653", "accept": false, - "reason": "BEiT legacy-field loading and BigBirdTokenizer special-token registration are distinct issues." + "reason": "Beit legacy-field loading bug vs BigBirdTokenizer special-token registration issue; unrelated." }, { "left": "issue:43575", "right": "issue:45405", "accept": false, - "reason": "Qwen2-57B tensor-parallel OOM and PEFT version bump are unrelated." + "reason": "Qwen2 tp OOM vs an unreleased PEFT version bump; unrelated." }, { "left": "issue:43278", "right": "issue:43701", "accept": false, - "reason": "Embedding dtype drift and resume_from_checkpoint key mismatch are different bugs." + "reason": "Embedding dtype regression vs resume_from_checkpoint key mismatch; different code paths." }, { "left": "issue:44792", "right": "issue:45405", "accept": false, - "reason": "A Janus image-generation test failure is unrelated to a PEFT release/version issue." + "reason": "Janus image-generation test failure vs PEFT version pin; unrelated." }, { "left": "issue:43994", "right": "issue:44493", "accept": false, - "reason": "SigLIP2 nonsensical outputs and unexpected position-id keys are separate model-behavior regressions." + "reason": "SigLIP2 output discrepancy vs unexpected position-id key warnings; different bugs." }, { "left": "issue:44423", "right": "issue:44734", "accept": false, - "reason": "Both are serving crashes, but they hit different pathways: continuous batching vs KV-cache continuation indexing." + "reason": "Serve crashes in multimodal batching vs KV-cache continuation tensor indexing; both are serve bugs, but not the same issue." }, { - "left": "issue:29942", - "right": "issue:30333", + "left": "issue:43873", + "right": "issue:46050", "accept": false, - "reason": "Flash Attention 2 test failures and MLflow reporting behavior are unrelated." + "reason": "Quantization/offloading behavior vs sinq save/load breakage; related theme but different concrete failures." } ] }, @@ -7212,7 +7178,7 @@ "component_index": 1, "component_count": 17, "review_unit_index": 23, - "review_unit_count": 50, + "review_unit_count": 49, "status": "reviewed", "reason": null, "source": "llm", @@ -7223,6 +7189,8 @@ "issue:15354", "issue:28218", "issue:29159", + "issue:29942", + "issue:30333", "issue:32090", "issue:34567", "issue:37428", @@ -7233,7 +7201,6 @@ "issue:43450", "issue:43493", "issue:43540", - "issue:43550", "issue:43572", "issue:43582", "issue:43696", @@ -7248,7 +7215,6 @@ "issue:44554", "issue:44610", "issue:44849", - "issue:44898", "issue:44964", "issue:44977", "issue:44987", @@ -7261,7 +7227,6 @@ "issue:45325", "issue:45356", "issue:45406", - "issue:45412", "issue:45446", "issue:45520", "issue:45588", @@ -7270,7 +7235,7 @@ "issue:46121" ], "soft_pairs": [ - "issue:43873|issue:46050", + "issue:29942|issue:30333", "issue:45406|issue:46121", "issue:45588|issue:46050", "issue:42490|issue:43696", @@ -7283,226 +7248,218 @@ "issue:44554|issue:45446", "issue:43540|issue:45200", "issue:15354|issue:38175", - "issue:45200|issue:45823", "issue:43232|issue:44849", - "issue:45072|issue:45198", "issue:37428|issue:43957", + "issue:45072|issue:45198", + "issue:45200|issue:45823", "issue:43493|issue:43701", "issue:44315|issue:44849", - "issue:43295|issue:44610", "issue:43572|issue:44987", + "issue:43295|issue:44610", "issue:43295|issue:43450", "issue:44554|issue:45356", "issue:43572|issue:43957", "issue:37428|issue:43824", "issue:45200|issue:45325", "issue:43295|issue:43873", - "issue:44898|issue:45412", - "issue:44466|issue:45127", - "issue:43550|issue:43844" + "issue:44466|issue:45127" ], - "prepared_review_unit_hash": "32bd71834ecd8a2ea4287728edc2d88744545b3307436f5490617d165ed3f37a", + "prepared_review_unit_hash": "97ce52e651462554b0e837d4dc2101d414e731db7eae550a3faccab980acef60", "component_budget": { "node_count": 426, "item_count": 426, - "soft_pair_count": 1572, - "serialized_chars": 814130, - "estimated_input_tokens": 203533, - "estimated_eval_tokens": 407322 + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 }, "budget": { - "node_count": 48, - "item_count": 48, - "soft_pair_count": 30, - "serialized_chars": 36891, - "estimated_input_tokens": 9223, - "estimated_eval_tokens": 18702 + "node_count": 47, + "item_count": 47, + "soft_pair_count": 28, + "serialized_chars": 35685, + "estimated_input_tokens": 8922, + "estimated_eval_tokens": 18100 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is heterogeneous: tokenizer/chat-template, model-loading, attention, training, and quantization issues are all mixed together. I found no soft pair that clearly describes the same underlying bug/change, so all soft edges are rejected. The best single representative issue is the tokenizer/processor regression report in 43295, but only as a loose cluster anchor, not a true duplicate hub.", - "confidence": 0.39, - "canonical_issue_reason": "Issue 43295 is the most representative cluster anchor because it is a concrete regression report with broad processor/tokenizer impact and sits at the center of several soft-similarity links, even though the linked items are not true duplicates.", + "summary": "No strong duplicate cluster here; the items are mostly unrelated bugs/regressions across different features. I\u2019d treat issue 43295 as the best representative issue for triage, but not as a duplicate hub.", + "confidence": 0.96, + "canonical_issue_reason": "Issue 43295 is the most concrete, actionable regression report in the set, with a narrow failure mode and clear reproduction context.", "canonical_pr_reason": null, - "best_issue_reason": "43295 is the clearest, most actionable issue among the set: it describes a specific regression with observable behavior and a plausible fix path, making it the best representative issue for this mixed cluster.", + "best_issue_reason": "Issue 43295 is the strongest overall triage candidate because it describes a specific regression path and is easier to act on than the more vague or subsystem-broad reports.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:43873", - "right": "issue:46050", + "left": "issue:29942", + "right": "issue:30333", "accept": false, - "reason": "Both involve quantization-related failures, but one is about offloading behavior and the other about save/load corruption after sinq quantization; different code paths and symptoms." + "reason": "Flash Attention CI failures and MLFlow job-status reporting are unrelated bugs in different subsystems." }, { "left": "issue:45406", "right": "issue:46121", "accept": false, - "reason": "A Gemma4 processor attribute crash and a rope-parameter JSON list TypeError are unrelated bugs." + "reason": "One is a Gemma4 processor attribute crash; the other is rope-parameter validation with JSON-loaded lists. Different code paths." }, { "left": "issue:45588", "right": "issue:46050", "accept": false, - "reason": "Flash-attention integration crashing on `s_aux=None` is unrelated to quantized model persistence after save_pretrained/from_pretrained." + "reason": "Flash-attention sink handling and sinq save/load corruption are distinct quantization/attention issues." }, { "left": "issue:42490", "right": "issue:43696", "accept": false, - "reason": "Local-path import warning and CUDA OOM are not the same underlying issue." + "reason": "Local-path import warnings and CUDA OOM are not the same underlying problem." }, { "left": "issue:44521", "right": "issue:45290", "accept": false, - "reason": "Both touch chat templating, but one is about multimodal assistant masks and the other about tool-call assistant messages crashing; distinct behaviors." + "reason": "Both involve apply_chat_template, but one is zero assistant masks for multimodal input and the other is a crash on tool-call messages; different failures." }, { "left": "issue:43844", "right": "issue:44964", "accept": false, - "reason": "Gradient blow-up under ZeRO-3 and a model-loading failure for Phi-4 multimodal are unrelated." + "reason": "ZeRO-3 gradient growth and model-loading import failure are unrelated." }, { "left": "issue:28218", "right": "issue:34567", "accept": false, - "reason": "Tokenizer spacing after added tokens is unrelated to `num_input_tokens_seen` not updating." + "reason": "Tokenizer spacing after added tokens is unrelated to TrainerState token-count tracking." }, { "left": "issue:29159", "right": "issue:32090", "accept": false, - "reason": "Fast/slow tokenizer inconsistency and a Trainer `_gpu_broadcast_one` NoneType error are unrelated." + "reason": "Slow/fast tokenizer inconsistency and a Trainer NoneType broadcast error do not share the same bug." }, { "left": "issue:45084", "right": "issue:45520", "accept": false, - "reason": "Template-node compilation failure and `flash_attn` import lookup failure are different subsystems and failure modes." + "reason": "Template-node compilation errors and a Python 3.13 flash_attn import KeyError are different problems." }, { "left": "issue:43582", "right": "issue:44977", "accept": false, - "reason": "Apple Silicon allocator warmup TypeError and Qwen3.5 flash-attention generation issues are unrelated." + "reason": "Apple Silicon allocator TypeError and Qwen3.5 flash-attention generation bugs are not the same issue." }, { "left": "issue:44554", "right": "issue:45446", "accept": false, - "reason": "An MPS attention correctness bug and a flex_attention import version check bug do not describe the same problem." + "reason": "MPS attention correctness and a flex_attention import version check are unrelated." }, { "left": "issue:43540", "right": "issue:45200", "accept": false, - "reason": "Video-input processing in Qwen3OmniMoe and Gemma4's `mm_token_type_ids` defaulting issue are separate multimodal bugs." + "reason": "Video input validation in Qwen3OmniMoe and Gemma4 text-only mm_token_type_ids handling are different bugs." }, { "left": "issue:15354", "right": "issue:38175", "accept": false, - "reason": "TorchScript GeneratorExp support and SigLIP2 zero-probability outputs are unrelated." + "reason": "TorchScript GeneratorExp export failure and SigLIP2 zero-probability behavior are separate issues." }, { - "left": "issue:45200", - "right": "issue:45823", + "left": "issue:43232", + "right": "issue:44849", "accept": false, - "reason": "Gemma4 multimodal token-type defaults and a device mismatch in Gemma4 PLE forward are different code-path issues." + "reason": "Generation kwargs update after sync_gpus and Qwen3.5 hidden-state behavior are different code paths." }, { - "left": "issue:43232", - "right": "issue:44849", + "left": "issue:37428", + "right": "issue:43957", "accept": false, - "reason": "Generation kwargs after `sync_gpus` and Qwen3.5 hidden-state output behavior are unrelated." + "reason": "Missing flash-support import and meta-device model loading breakage are unrelated." }, { "left": "issue:45072", "right": "issue:45198", "accept": false, - "reason": "bfloat16 dtype mismatches in SwitchTransformers/TimmWrapperModel and Wav2Vec2 save/tokenization failure are different bugs." + "reason": "BF16 dtype mismatches in inference and Wav2Vec2 save/tokenization failures are not the same bug." }, { - "left": "issue:37428", - "right": "issue:43957", + "left": "issue:45200", + "right": "issue:45823", "accept": false, - "reason": "A missing flash-attention utility import and meta-device model loading breakage are not the same underlying issue." + "reason": "Both are Gemma4-related, but one is token-type defaults and the other is a device-map forward mismatch; not the same underlying defect." }, { "left": "issue:43493", "right": "issue:43701", "accept": false, - "reason": "SigLIP2 HF/JAX discrepancy and resume-from-checkpoint key mismatch are unrelated." + "reason": "SigLIP2 implementation discrepancy and resume_from_checkpoint key mismatch are unrelated." }, { "left": "issue:44315", "right": "issue:44849", "accept": false, - "reason": "Liger Kernel not applied in `model_init` and a Qwen3.5 hidden-states bug are unrelated." + "reason": "Liger Kernel application in model_init and Qwen3.5 hidden-state behavior do not share a concrete bug." }, { - "left": "issue:43295", - "right": "issue:44610", + "left": "issue:43572", + "right": "issue:44987", "accept": false, - "reason": "Both are regression-style processor issues, but one concerns processor.tokenizer/image handling and the other is an image-size mismatch; not the same bug." + "reason": "StableLm pad_token_idx regression and loading physical-intelligence/fast on newer transformers are different failures." }, { - "left": "issue:43572", - "right": "issue:44987", + "left": "issue:43295", + "right": "issue:44610", "accept": false, - "reason": "Missing `pad_token_idx` in StableLmConfig and a transformers>=5.1.0 loading failure for a specific model are unrelated." + "reason": "Both involve processor behavior, but one is processor.tokenizer/image access regression and the other is a 640px-vs-224px processor/model size mismatch." }, { "left": "issue:43295", "right": "issue:43450", "accept": false, - "reason": "Processor/tokenizer regression and batched video shape issues are different modalities and failure modes." + "reason": "Processor/tokenizer regression and batched video output shape bug are distinct." }, { "left": "issue:44554", "right": "issue:45356", "accept": false, - "reason": "MPS attention correctness and Kimi-K2.5 tokenizer codec regression are unrelated." + "reason": "MPS attention correctness and Kimi tokenizer regression are separate issues." + }, + { + "left": "issue:43572", + "right": "issue:43957", + "accept": false, + "reason": "Pad token config regression and meta-device loading breakage are unrelated." }, { "left": "issue:37428", "right": "issue:43824", "accept": false, - "reason": "Flash-attention import-name failure and Qwen2.5-VL import failure are different import problems." + "reason": "Importing a missing flash-attention helper and missing Qwen2.5-VL class export are different import errors." }, { "left": "issue:45200", "right": "issue:45325", "accept": false, - "reason": "Gemma4 token-type defaults and Qwen2.5-VL rope-index scaling are unrelated multimodal issues." + "reason": "Gemma4 token-type defaults and Qwen2.5-VL rope index scaling are unrelated multimodal bugs." }, { "left": "issue:43295", "right": "issue:43873", "accept": false, - "reason": "Processor/tokenizer regression and quantization offloading behavior are not the same concrete bug." - }, - { - "left": "issue:44898", - "right": "issue:45412", - "accept": false, - "reason": "Perceiver non-default-resolution failure and RT-DETR memory release issues are unrelated." + "reason": "Processor/tokenizer regression and quantization offloading behavior are not the same bug." }, { "left": "issue:44466", "right": "issue:45127", "accept": false, - "reason": "Serialization of tied weights depending on device and LoRA merge collapse with extended vocab are related only loosely by weights, but not the same bug." - }, - { - "left": "issue:43550", - "right": "issue:43844", - "accept": false, - "reason": "torch.compile/SDPA failure and ZeRO-3 gradient growth are unrelated training/runtime issues." + "reason": "lm_head.weight serialization inconsistency and LoRA merge collapse with extended vocab are different weight-handling issues." } ] }, @@ -7518,7 +7475,7 @@ "component_index": 1, "component_count": 17, "review_unit_index": 24, - "review_unit_count": 50, + "review_unit_count": 49, "status": "reviewed", "reason": null, "source": "llm", @@ -7549,19 +7506,19 @@ "issue:43819", "issue:43824", "issue:43827", + "issue:43844", "issue:43856", "issue:43901", "issue:43975", "issue:44112", "issue:44206", "issue:44442", - "issue:44448", "issue:44514", "issue:44521", "issue:44610", - "issue:44743", "issue:44792", "issue:44857", + "issue:44898", "issue:45083", "issue:45127", "issue:45198", @@ -7576,6 +7533,8 @@ "issue:46170" ], "soft_pairs": [ + "issue:44898|issue:45412", + "issue:43550|issue:43844", "issue:43334|issue:43525", "issue:44792|issue:45563", "issue:43824|issue:43975", @@ -7584,8 +7543,8 @@ "issue:43450|issue:43575", "issue:42490|issue:43493", "issue:43278|issue:43856", - "issue:28218|issue:35141", "issue:43975|issue:45520", + "issue:28218|issue:35141", "issue:43653|issue:44112", "issue:38175|issue:43493", "issue:44442|issue:45198", @@ -7599,222 +7558,230 @@ "issue:43819|issue:45083", "issue:44610|issue:44857", "issue:37428|issue:43425", - "issue:43335|issue:43550", "issue:43606|issue:45083", + "issue:43335|issue:43550", "issue:44857|issue:45593", "issue:45290|issue:45381", - "issue:45127|issue:45245", - "issue:44448|issue:44743" + "issue:45127|issue:45245" ], - "prepared_review_unit_hash": "8c958246e63c8b3dce8d188086a983eed299098cceb85eb64207d3e5fbfd6cae", + "prepared_review_unit_hash": "9278750a47ab5b50c09f75f696d438b801a26c18f744c2d5e754f71c0cb12001", "component_budget": { "node_count": 426, "item_count": 426, - "soft_pair_count": 1572, - "serialized_chars": 814130, - "estimated_input_tokens": 203533, - "estimated_eval_tokens": 407322 + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 29, - "serialized_chars": 36540, - "estimated_input_tokens": 9135, - "estimated_eval_tokens": 18526 + "soft_pair_count": 30, + "serialized_chars": 36847, + "estimated_input_tokens": 9212, + "estimated_eval_tokens": 18680 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The set is heterogeneous and the suggested soft pairs are mostly topical lookalikes rather than true duplicates. I would not merge any of the soft edges; they generally describe different bugs, different models, or different code paths.", - "confidence": 0.96, - "canonical_issue_reason": "No single canonical issue stands out: these items span unrelated reports across tokenization, multimodal processors, model configs, import/runtime errors, and training/evaluation behavior.", + "summary": "Mostly unrelated issues clustered by broad subsystem vocabulary. The strongest duplicate-like pair is the SigLIP2 discrepancy report and the zero-probability symptom report, but a few other title-level near-matches exist (notably the two pad_token_id AttributeErrors and the two apply_chat_template reports) without enough evidence here to merge them.", + "confidence": 0.71, + "canonical_issue_reason": "issue:43493 is the best anchor for the SigLIP2 bug family because it describes the broader HF-vs-JAX implementation mismatch; issue:38175 looks like a symptom of that same underlying discrepancy.", "canonical_pr_reason": null, - "best_issue_reason": "issue:44521 is the strongest representative if one issue must be chosen: it is open, has the most discussion/inbound activity among the set, and describes a concrete, user-facing bug.", + "best_issue_reason": "issue:43493 is the strongest global issue candidate because it states the underlying SigLIP2 implementation mismatch more generally than the symptom report in issue:38175.", "best_pr_reason": null, "soft_edge_verdicts": [ + { + "left": "issue:44898", + "right": "issue:45412", + "accept": false, + "reason": "Different models and failure modes: Perceiver resolution handling vs RT-DETR memory cleanup." + }, + { + "left": "issue:43550", + "right": "issue:43844", + "accept": false, + "reason": "Both are training/runtime failures, but the concrete code paths differ: Bamba torch.compile/SDPA vs ZeRO-3 gradient growth." + }, { "left": "issue:43334", "right": "issue:43525", "accept": false, - "reason": "Both are pad_token_id AttributeErrors, but for different configs/models (Qwen3-VL vs Llama4) and likely different config plumbing; not the same underlying bug." + "reason": "Same missing-attribute shape, but different config classes and model families; not enough to treat as one bug." }, { "left": "issue:44792", "right": "issue:45563", "accept": false, - "reason": "Janus test failure vs a stale warning in paged generate are unrelated symptoms and code paths." + "reason": "Janus image-generation test failure vs a stale paged-generate warning; not the same underlying issue." }, { "left": "issue:43824", "right": "issue:43975", "accept": false, - "reason": "ImportError for a missing class and a detokenization regression are different failures with no shared fix path." + "reason": "ImportError for a missing class vs a detokenization regression; no shared concrete code path is evident." }, { "left": "issue:42915", "right": "issue:43824", "accept": false, - "reason": "Qwen3-MoE FP8 failure and Qwen2.5-VL import error are unrelated model-specific issues." + "reason": "Different models and symptoms: FineGrainedFP8Config failure vs Qwen2_5_VL import failure." }, { "left": "issue:43388", "right": "issue:43526", "accept": false, - "reason": "Both involve labels handling, but one is gather_for_metrics truncation in distributed evaluation and the other is BeitImageProcessorFast reduce_labels behavior; different components and bugs." + "reason": "Both involve labels, but one is gather_for_metrics batching and the other is BeitImageProcessorFast label reduction." }, { "left": "issue:43450", "right": "issue:43575", "accept": false, - "reason": "Video processor shape bug and Qwen2-57B-A14B-Instruct TP OOM are not the same code-path problem." + "reason": "Video batch-shape bug and Qwen2-57B OOM are unrelated." }, { "left": "issue:42490", "right": "issue:43493", "accept": false, - "reason": "A local-path trailing-slash import warning and a SigLIP2 implementation discrepancy are unrelated." + "reason": "Local-path import warning vs SigLIP2 implementation discrepancy; no concrete overlap from the titles alone." }, { "left": "issue:43278", "right": "issue:43856", "accept": false, - "reason": "Embedding dtype drift in evaluate and Qwen3 MoE memory usage are different issues in different phases of execution." + "reason": "Embedding dtype drift in eval and MoE training memory usage are different bugs." }, { - "left": "issue:28218", - "right": "issue:35141", + "left": "issue:43975", + "right": "issue:45520", "accept": false, - "reason": "Tokenizer spacing after added tokens and embedding reinitialization after resize are distinct tokenizer/model initialization bugs." + "reason": "Detokenization regression and a Python 3.13 flash_attn import KeyError are unrelated." }, { - "left": "issue:43975", - "right": "issue:45520", + "left": "issue:28218", + "right": "issue:35141", "accept": false, - "reason": "Detokenization mismatch and a Python 3.13 flash_attn import KeyError are unrelated." + "reason": "Tokenizer spacing after added tokens and output embedding reinitialization after resize are distinct issues." }, { "left": "issue:43653", "right": "issue:44112", "accept": false, - "reason": "BigBirdTokenizer special-token registration and a GraniteSpeech stale device-override test are not the same bug." + "reason": "Tokenizer special-token registration bug vs GraniteSpeech CI device-override test failure." }, { "left": "issue:38175", "right": "issue:43493", - "accept": false, - "reason": "Both mention SigLIP2, but one reports zero probabilities and the other a broader HF-vs-JAX discrepancy; not enough to treat as the same concrete defect." + "accept": true, + "reason": "Both are SigLIP2 output-discrepancy reports, and the broader HF-vs-JAX mismatch plausibly explains the zero-probability symptom." }, { "left": "issue:44442", "right": "issue:45198", "accept": false, - "reason": "AutoTokenizer loading failure for FastSpeech2ConformerTokenizer and Wav2Vec2 save/tokenization failures are different tokenizer/model behaviors." + "reason": "FastSpeech2Conformer tokenizer loading and Wav2Vec2 save_pretrained/tokenization failures are different issues." }, { "left": "issue:44514", "right": "issue:44521", "accept": false, - "reason": "Batched apply_chat_template padding=False crash and all-zero assistant masks are both chat-template related but describe different failures and likely different fixes." + "reason": "Both are apply_chat_template reports, but one is a batched padding=False crash and the other is all-zero assistant masks; not enough to merge." }, { "left": "issue:45520", "right": "issue:46170", "accept": false, - "reason": "Both concern import/dependency issues, but on different platforms and with different causes (flash_attn key vs torch.distributed on Windows/ROCm)." + "reason": "Different platform/import failures: flash_attn on Python 3.13 vs Windows ROCm torch.distributed dependency issues." }, { "left": "issue:38175", "right": "issue:45412", "accept": false, - "reason": "SigLIP2 output anomalies and RT-DETR memory release issues are unrelated model bugs." + "reason": "SigLIP2 zero-probability symptom and an RT-DETR memory leak are unrelated." }, { "left": "issue:44206", "right": "issue:45584", "accept": false, - "reason": "A feature-extractor center-argument regression and Whisper empty-transcription failure after token alignment are different code paths." + "reason": "Unsupported center argument in LasrFeatureExtractor is unrelated to Whisper empty-transcription handling." }, { "left": "issue:43827", "right": "issue:43901", "accept": false, - "reason": "Both are docs mismatches around removed/changed pipeline arguments, but they concern different pipelines and different documented APIs, so not one duplicate bug." + "reason": "Both are docs issues, but they reference different APIs and different pages." }, { "left": "issue:43262", "right": "issue:43450", "accept": false, - "reason": "Audio apply_chat_template sample-rate defaults and batched video processor output shapes are unrelated processor bugs." + "reason": "Audio chat-template sampling-rate default and video processor batched-shape bug are unrelated." }, { "left": "issue:43819", "right": "issue:44112", "accept": false, - "reason": "DAC.from_latents mismatch and a GraniteSpeech stale device override test do not share the same failing path." + "reason": "DAC latent reconstruction mismatch vs GraniteSpeech CI device override are unrelated." }, { "left": "issue:43819", "right": "issue:45083", "accept": false, - "reason": "DAC latent reconstruction and qwen3_omni_moe feature-length helper behavior are different subsystems." + "reason": "Different code paths: DAC latent conversion vs qwen3_omni_moe helper length calculation." }, { "left": "issue:44610", "right": "issue:44857", "accept": false, - "reason": "Wrong processor input size and CUDA/float16 AMP loss crash are different issues." + "reason": "Processor output resolution mismatch and LwDetrImageLoss AMP/CUDA crash are different bugs." }, { "left": "issue:37428", "right": "issue:43425", "accept": false, - "reason": "A missing private import and a Torch 2.10 compatibility report are not the same concrete bug." + "reason": "Missing flash-attention import and Torch 2.10 incompatibility are not the same underlying problem." }, { - "left": "issue:43335", - "right": "issue:43550", + "left": "issue:43606", + "right": "issue:45083", "accept": false, - "reason": "A config bug creating sparse layers at zero count and a torch.compile+SDPA failure in Bamba are unrelated." + "reason": "CPU offload device mismatch in bark-small and qwen3_omni_moe length helper behavior are unrelated." }, { - "left": "issue:43606", - "right": "issue:45083", + "left": "issue:43335", + "right": "issue:43550", "accept": false, - "reason": "CPU offload device mismatch and qwen3_omni_moe feature-length helper behavior are different failures." + "reason": "SwitchTransformers sparse-layer construction and Bamba torch.compile/SDPA failure do not share a concrete code path." }, { "left": "issue:44857", "right": "issue:45593", "accept": false, - "reason": "AMP/CUDA crash in LwDetrImageLoss and missing auxiliary losses in D-FINE are unrelated training bugs." + "reason": "AMP/CUDA crash in LwDetrImageLoss vs D-FINE auxiliary-loss behavior are unrelated." }, { "left": "issue:45290", "right": "issue:45381", "accept": false, - "reason": "Tool-call assistant message handling and Qwen2.5-VL video vision_position_ids are different chat/template vs vision-positioning problems." + "reason": "Tool-call assistant message crash and Qwen2.5-VL video vision_position_ids bug are different chat-template/video problems." }, { "left": "issue:45127", "right": "issue:45245", "accept": false, - "reason": "LoRA merge with extended vocabulary causing collapse and a categories-cardinality RuntimeError are unrelated." - }, - { - "left": "issue:44448", - "right": "issue:44743", - "accept": false, - "reason": "Transformers v4/v5 output drift for Pegasus and recurrent state resets in modular_qwen3_5.py are not the same bug." + "reason": "LoRA merge collapse with extended vocab and a categories-limit runtime error are unrelated." } ] }, - "evaluator_result": null, - "evaluator_used": false, - "retried": false, - "accepted_nontrivial_soft_edge": false, + "evaluator_result": { + "accept": false, + "feedback": "The SigLIP2 pair looks plausibly related, but the writeup overstates the evidence by claiming it is the only pair that could be the same underlying issue. From the packet titles alone, there are other plausible near-matches (e.g. the two pad_token_id AttributeErrors and the two apply_chat_template issues), so the summary should be framed more cautiously. The soft-edge rejections are mostly conservative, but the exclusivity claim is too strong for the available evidence." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null }, @@ -7823,7 +7790,7 @@ "component_index": 1, "component_count": 17, "review_unit_index": 25, - "review_unit_count": 50, + "review_unit_count": 49, "status": "reviewed", "reason": null, "source": "llm", @@ -7864,6 +7831,7 @@ "issue:44448", "issue:44493", "issue:44561", + "issue:44743", "issue:44779", "issue:44855", "issue:44857", @@ -7880,276 +7848,274 @@ "issue:45859" ], "soft_pairs": [ + "issue:44448|issue:44743", "issue:29942|issue:32090", - "issue:45200|issue:45588", "issue:37428|issue:44351", + "issue:45200|issue:45588", "issue:44206|issue:45823", - "issue:43278|issue:43825", "issue:43611|issue:43931", + "issue:43278|issue:43825", "issue:43335|issue:43653", "issue:43550|issue:44112", "issue:43819|issue:44493", "issue:44855|issue:45083", "issue:43335|issue:43504", - "issue:45588|issue:45859", "issue:29127|issue:32090", + "issue:45588|issue:45859", "issue:43335|issue:45083", "issue:39692|issue:43526", "issue:45310|issue:45478", "issue:39692|issue:44493", - "issue:43720|issue:44112", "issue:44292|issue:45520", + "issue:43720|issue:44112", "issue:44448|issue:44779", "issue:43335|issue:45072", - "issue:43335|issue:44112", "issue:44360|issue:44561", + "issue:43335|issue:44112", "issue:42490|issue:43854", "issue:44977|issue:45229", "issue:38175|issue:44493", "issue:34689|issue:35141", "issue:43606|issue:44857", - "issue:44857|issue:45072", "issue:44448|issue:44977", + "issue:44857|issue:45072", "issue:40444|issue:43526", "issue:28218|issue:32090", "issue:43701|issue:45083" ], - "prepared_review_unit_hash": "4dc198168e561a58549a088ff24a3b4605f96613a52485c5029ff35c6b44862e", + "prepared_review_unit_hash": "b3261adc3d5b19a34802607b43742c17fc6869dd6ff152e6c6f60ca84aa5387e", "component_budget": { "node_count": 426, "item_count": 426, - "soft_pair_count": 1572, - "serialized_chars": 814130, - "estimated_input_tokens": 203533, - "estimated_eval_tokens": 407322 + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 }, "budget": { - "node_count": 47, - "item_count": 47, - "soft_pair_count": 33, - "serialized_chars": 37003, - "estimated_input_tokens": 9251, - "estimated_eval_tokens": 18758 + "node_count": 48, + "item_count": 48, + "soft_pair_count": 34, + "serialized_chars": 37900, + "estimated_input_tokens": 9475, + "estimated_eval_tokens": 19206 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Most pairs are unrelated issue reports that only share broad subsystem keywords. The only clear duplicate is the Qwen3.5 MoE `from_pretrained` regression pair.", - "confidence": 0.97, - "canonical_issue_reason": "issue:45310 is the earlier and more detailed report of the same Qwen3.5 MoE `from_pretrained` regression as issue:45478.", + "summary": "These are mostly unrelated Transformers bug reports across different models, loaders, and execution modes. None of the provided soft edges look like the same underlying bug, so I would not merge any of them. The only clear duplicate in the set is 45310/45478, which appear to be the same Qwen3.5 MoE from_pretrained regression; 45310 is the better canonical anchor.", + "confidence": 0.89, + "canonical_issue_reason": "issue:45310 is the older, slightly fuller report of the exact duplicate pair with issue:45478 (same Qwen3.5 MoE from_pretrained error).", "canonical_pr_reason": null, - "best_issue_reason": "issue:45310 is the best representative issue: it is the earliest clear report of the duplicate regression and has the fuller discussion.", + "best_issue_reason": "issue:45310 is the strongest canonical issue in the set because it has an exact duplicate and a concrete, reproducible loading failure; it is also the earliest report.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:29942", - "right": "issue:32090", + "left": "issue:44448", + "right": "issue:44743", "accept": false, - "reason": "Different failures: Flash Attention 2 test breakage vs Trainer `_gpu_broadcast_one` rejecting `NoneType`." + "reason": "Different bugs: Pegasus output drift vs Qwen3.5 recurrent-state reset in cached generation." }, { - "left": "issue:45200", - "right": "issue:45588", + "left": "issue:29942", + "right": "issue:32090", "accept": false, - "reason": "Gemma4 text-only `mm_token_type_ids` handling and a flash-attention sink-less model crash are different code paths." + "reason": "Unrelated failures: FlashAttention 2 test regressions vs Trainer `_gpu_broadcast_one` NoneType error." }, { "left": "issue:37428", "right": "issue:44351", "accept": false, - "reason": "Both are import errors, but for different missing symbols in different modules." + "reason": "Both are import errors, but for different missing symbols and different breakage points." }, { - "left": "issue:44206", - "right": "issue:45823", + "left": "issue:45200", + "right": "issue:45588", "accept": false, - "reason": "Unsupported `center` arg in a feature extractor is unrelated to Gemma4 device-map forward mismatch." + "reason": "Same broad Gemma4 area, but one is mm_token_type_ids defaults and the other is a flash-attention `s_aux=None` crash." }, { - "left": "issue:43278", - "right": "issue:43825", + "left": "issue:44206", + "right": "issue:45823", "accept": false, - "reason": "Embedding dtype drift in evaluate is unrelated to a pipeline translation-task error message." + "reason": "Different code paths: unsupported feature-extractor arg vs device-map forward mismatch." }, { "left": "issue:43611", "right": "issue:43931", "accept": false, - "reason": "Base-model-prefix loading regression and Qwen3-VL weight-shape mismatch are different model-loading bugs." + "reason": "Both are model-loading issues, but one is `base_model_prefix` handling and the other is a Qwen3-VL shape mismatch." + }, + { + "left": "issue:43278", + "right": "issue:43825", + "accept": false, + "reason": "Different regressions: dtype change during evaluate vs a pipeline error-message/deprecation issue." }, { "left": "issue:43335", "right": "issue:43653", "accept": false, - "reason": "SwitchTransformers sparse-layer config bug is unrelated to BigBirdTokenizer special-token decoding." + "reason": "Different components and symptoms: SwitchTransformers config bug vs BigBirdTokenizer special-token registration." }, { "left": "issue:43550", "right": "issue:44112", "accept": false, - "reason": "torch.compile SDPA failure in Bamba is not the same as a GraniteSpeech CI device-override test issue." + "reason": "Different failures: Bamba torch.compile/SDPA behavior vs a GraniteSpeech CI device-override test." }, { "left": "issue:43819", "right": "issue:44493", "accept": false, - "reason": "DAC latent/forward mismatch is unrelated to unexpected position-id keys during model loading." + "reason": "Different problems: DAC latent/STE mismatch vs unexpected key handling for position ids." }, { "left": "issue:44855", "right": "issue:45083", "accept": false, - "reason": "Python 3.13 `IndentationError` in DebertaV2 import is unrelated to qwen3_omni_moe feature-length behavior." + "reason": "Different issues: Python 3.13 import parsing error vs qwen3_omni_moe feature-length helper behavior." }, { "left": "issue:43335", "right": "issue:43504", "accept": false, - "reason": "One is a SwitchTransformers config construction bug; the other is a BEiT legacy-field model load failure." + "reason": "Different models and failures: SwitchTransformers sparse-layer creation vs BEiT legacy-field loading." }, { - "left": "issue:45588", - "right": "issue:45859", + "left": "issue:29127", + "right": "issue:32090", "accept": false, - "reason": "Both involve flash attention / MoE models, but one crashes on `s_aux=None` and the other is a CPU-offload rotary embedding issue." + "reason": "Error-message clarity request vs a runtime NoneType broadcast failure; not the same bug." }, { - "left": "issue:29127", - "right": "issue:32090", + "left": "issue:45588", + "right": "issue:45859", "accept": false, - "reason": "LayoutLMv3 error-message clarity and Trainer `_gpu_broadcast_one` `NoneType` support are not the same bug." + "reason": "Both involve flash attention/offload, but they fail in different places with different root causes." }, { "left": "issue:43335", "right": "issue:45083", "accept": false, - "reason": "Sparse-layer construction in SwitchTransformers is unrelated to qwen3_omni_moe feature extraction lengths." + "reason": "SwitchTransformers config creation bug vs qwen3_omni_moe output-length helper behavior." }, { "left": "issue:39692", "right": "issue:43526", "accept": false, - "reason": "SigLIP2 doc example errors are unrelated to BEiT fast processor label reduction." - }, - { - "left": "issue:45310", - "right": "issue:45478", - "accept": true, - "reason": "Same underlying bug: identical Qwen3.5 MoE `from_pretrained` regression, with matching title and symptom." + "reason": "Different areas: SigLIP2 docs/model mismatch vs BeitImageProcessorFast label reduction." }, { "left": "issue:39692", "right": "issue:44493", "accept": false, - "reason": "A SigLIP2 documentation example issue is unrelated to unexpected position-id keys across models." + "reason": "Docs/example errors are unrelated to the position-id key regression." }, { - "left": "issue:43720", - "right": "issue:44112", + "left": "issue:44292", + "right": "issue:45520", "accept": false, - "reason": "Packed-weight loading in BitNet and a stale GraniteSpeech CI override test are different problems." + "reason": "Different failures: Qwen-3-8B-NVFP4 runtime issue vs Python 3.13 flash_attn import KeyError." }, { - "left": "issue:44292", - "right": "issue:45520", + "left": "issue:43720", + "right": "issue:44112", "accept": false, - "reason": "Qwen-3-8B-NVFP4 runtime failure and a Python 3.13 `flash_attn` import key error are not the same bug." + "reason": "BitNet packed-weight loading bug vs stale CI device-override test; different underlying problems." }, { "left": "issue:44448", "right": "issue:44779", "accept": false, - "reason": "Pegasus v4/v5 output differences and Deepseek tokenizer regressions are separate model-specific issues." + "reason": "Pegasus v4/v5 output difference vs Deepseek tokenizer regression; different subsystems." }, { "left": "issue:43335", "right": "issue:45072", "accept": false, - "reason": "SwitchTransformers sparse-layer config and bfloat16 dtype mismatches in SwitchTransformers/TimmWrapperModel are not the same concrete failure." + "reason": "Same model family, but one is config construction and the other is bfloat16 inference dtype mismatch." }, { - "left": "issue:43335", - "right": "issue:44112", + "left": "issue:44360", + "right": "issue:44561", "accept": false, - "reason": "SwitchTransformers config creation bug and GraniteSpeech CI device override failure are unrelated." + "reason": "DSA indexer missing ReLU is unrelated to the `is_torch_fx_available` v5 compatibility break." }, { - "left": "issue:44360", - "right": "issue:44561", + "left": "issue:43335", + "right": "issue:44112", "accept": false, - "reason": "A DSA indexer discussion issue is not the same as removal of `is_torch_fx_available` breaking trust-remote-code models." + "reason": "SwitchTransformers config bug and GraniteSpeech CI device-override failure are not the same issue." }, { "left": "issue:42490", "right": "issue:43854", "accept": false, - "reason": "Local-path trailing-slash import warning and GLM-4.7-Flash unit-test loading failure are different loading issues." + "reason": "Local-path import warning vs model-loading failure in unit tests; no shared root cause." }, { "left": "issue:44977", "right": "issue:45229", "accept": false, - "reason": "Flash-attention generation corruption in Qwen3.5 is unrelated to Gemma4 multi-GPU OOM." + "reason": "Flash-attention generation bug vs CUDA OOM during multi-GPU inference; different failure modes." }, { "left": "issue:38175", "right": "issue:44493", "accept": false, - "reason": "SigLIP2 zero probabilities and position-id key warnings are different symptoms with different likely causes." + "reason": "SigLIP2 zero-probability output bug is unrelated to the position-id key warning/regression." }, { "left": "issue:34689", "right": "issue:35141", "accept": false, - "reason": "Llama 3.2 Vision model loading breakage and embedding reinitialization after resize are unrelated." + "reason": "Different regressions: Llama 3.2 Vision load break vs output embedding reinitialization after resize." }, { "left": "issue:43606", "right": "issue:44857", "accept": false, - "reason": "CPU-offload device mismatch in Bark and float16 AMP crashes in LwDetrImageLoss are different failures." + "reason": "CPU offload device mismatch vs float16 AMP crash in LwDetrImageLoss; unrelated code paths." }, { - "left": "issue:44857", - "right": "issue:45072", + "left": "issue:44448", + "right": "issue:44977", "accept": false, - "reason": "AMP/CUDA crash in LwDetrImageLoss is unrelated to bfloat16 dtype mismatches in SwitchTransformers/TimmWrapperModel." + "reason": "Both are generation regressions, but on different models and different attention stacks." }, { - "left": "issue:44448", - "right": "issue:44977", + "left": "issue:44857", + "right": "issue:45072", "accept": false, - "reason": "Pegasus output drift and Qwen3.5 flash-attention generation failures are unrelated model regressions." + "reason": "AMP loss crash vs dtype mismatch in a different model family; not the same underlying bug." }, { "left": "issue:40444", "right": "issue:43526", "accept": false, - "reason": "Qwen2.5-VL multi-image iterable dataset failure is unrelated to BEiT processor label reduction." + "reason": "Multi-image IterableDataset fine-tuning failure is unrelated to BeitImageProcessorFast `reduce_labels`." }, { "left": "issue:28218", "right": "issue:32090", "accept": false, - "reason": "Tokenizer extra-space behavior and Trainer `_gpu_broadcast_one` type handling are different bugs." + "reason": "Tokenizer added-token spacing bug and Trainer broadcast error are unrelated." }, { "left": "issue:43701", "right": "issue:45083", "accept": false, - "reason": "Checkpoint key mismatch and qwen3_omni_moe feature-length behavior are unrelated." + "reason": "Checkpoint-key mismatch vs qwen3_omni_moe helper behavior; no shared underlying bug." } ] }, - "evaluator_result": { - "accept": true, - "feedback": "The summary is grounded in the packet: 45310/45478 are the only clearly identical pair by title and timing, and the other soft-edge rejections stay conservative by relying on distinct titles/symptoms. No mergeability concerns for PR pairs here." - }, - "evaluator_used": true, + "evaluator_result": null, + "evaluator_used": false, "retried": false, - "accepted_nontrivial_soft_edge": true, + "accepted_nontrivial_soft_edge": false, "error_kind": null, "error_message": null }, @@ -8158,7 +8124,7 @@ "component_index": 1, "component_count": 17, "review_unit_index": 26, - "review_unit_count": 50, + "review_unit_count": 49, "status": "reviewed", "reason": null, "source": "llm", @@ -8222,224 +8188,224 @@ "issue:37428|issue:44987", "issue:44561|issue:44779", "issue:43873|issue:45229", - "issue:44462|issue:44561", "issue:43927|issue:45698", + "issue:44462|issue:44561", "issue:41720|issue:45229", - "issue:43278|issue:43329", "issue:44977|issue:45563", - "issue:10000|issue:15354", "issue:44568|issue:45491", + "issue:43278|issue:43329", + "issue:10000|issue:15354", "issue:42490|issue:43575", "issue:45372|issue:45588", "issue:43412|issue:45588", - "issue:44857|issue:45588", "issue:45198|issue:45803", + "issue:44857|issue:45588", "issue:43976|issue:45440", "issue:45372|issue:45406", "issue:45459|issue:45712", - "issue:45588|issue:45823", "issue:43278|issue:43381", - "issue:43653|issue:44743", + "issue:45588|issue:45823", "issue:44448|issue:44466", + "issue:43653|issue:44743", "issue:45245|issue:45563", "issue:25251|issue:32090" ], - "prepared_review_unit_hash": "6aa311ea6c0c772e55c4cade7fd1b6dc5548d6ab665e1d8aca7fbf57d399309a", + "prepared_review_unit_hash": "7eba210ddc744ebe434eb4989bdfeb06105cef80be654fc8075af137464da2bb", "component_budget": { "node_count": 426, "item_count": 426, - "soft_pair_count": 1572, - "serialized_chars": 814130, - "estimated_input_tokens": 203533, - "estimated_eval_tokens": 407322 + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 }, "budget": { "node_count": 47, "item_count": 47, "soft_pair_count": 28, - "serialized_chars": 35790, - "estimated_input_tokens": 8948, - "estimated_eval_tokens": 18152 + "serialized_chars": 35784, + "estimated_input_tokens": 8946, + "estimated_eval_tokens": 18148 }, "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, loading, attention, device-mapping, serialization, and export bugs. I do not see any true duplicate pair among the soft edges, so all candidate merges are rejected. If forced to anchor the cluster, issue 45588 is the most concrete, reproducible runtime crash and the best triage representative.", - "confidence": 0.97, - "canonical_issue_reason": "Issue 45588 is the cleanest standalone bug report in the set: it names a specific file, a specific failure (`AttributeError` on `s_aux=None`), and a clear runtime path. It is not a duplicate of the others, but it is the best anchor for triage.", + "summary": "This is not a coherent duplicate cluster: the items span unrelated tokenizer, attention, export, loading, memory, and generation bugs. None of the soft pairs look like the same concrete underlying issue closely enough to merge.", + "confidence": 0.98, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "45588 is the best issue for a canonical slot because it is narrowly scoped, reproducible, and tied to a concrete code path rather than a broad regression description.", + "best_issue_reason": null, "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43723", "right": "issue:44554", "accept": false, - "reason": "Different problems: tokenizer loading in v5 vs an MPS attention correctness issue. No shared concrete failure path." + "reason": "Tokenizer loading regression and MPS attention correctness bug are unrelated code paths." }, { "left": "issue:43550", "right": "issue:45083", "accept": false, - "reason": "One is a torch.compile/SDPA failure in Bamba; the other is a helper-length bug in qwen3_omni_moe. Unrelated code paths." + "reason": "Torch.compile/SDPA failure for Bamba is unrelated to the qwen3_omni_moe feature-length helper bug." }, { "left": "issue:32090", "right": "issue:33357", "accept": false, - "reason": "Trainer broadcast TypeError on NoneType vs MacOS bus error with a community CLIP model. Different subsystems and symptoms." + "reason": "Trainer broadcast TypeError and MacOS bus error in CLIP model loading are different failures." }, { "left": "issue:29942", "right": "issue:35141", "accept": false, - "reason": "Flash Attention 2 test failures and embedding reinitialization in post_init are separate bugs." + "reason": "Flash Attention 2 test failures and token-embedding reinitialization are separate subsystems." }, { "left": "issue:37428", "right": "issue:44987", "accept": false, - "reason": "Both involve loading/imports, but one is a missing flash-attention utility symbol and the other is a model load failure for a specific repo. Not the same bug." + "reason": "Missing flash-attention import symbol is not the same as a generic model-loading failure for physical-intelligence/fast." }, { "left": "issue:44561", "right": "issue:44779", "accept": false, - "reason": "Both are v5 regressions, but one is a removed API breaking trust_remote_code and the other is a DeepSeek tokenizer behavior change. Different underlying issues." + "reason": "FX helper removal breaking trust_remote_code and Deepseek tokenizer regression are different bugs." }, { "left": "issue:43873", "right": "issue:45229", "accept": false, - "reason": "Quantization/offloading behavior vs multi-GPU OOM for Gemma4. Same broad area, different failure mode." + "reason": "Quantization/offloading behavior and Gemma4 multi-GPU OOM are related only broadly by memory usage, not the same bug." }, { - "left": "issue:44462", - "right": "issue:44561", + "left": "issue:43927", + "right": "issue:45698", "accept": false, - "reason": "Tokenizer repository selection bug vs removed is_torch_fx_available breaking remote-code models. No duplicate evidence." + "reason": "Custom token IDs being lost on save/load and loading the wrong custom module after save_pretrained are distinct serialization problems." }, { - "left": "issue:43927", - "right": "issue:45698", + "left": "issue:44462", + "right": "issue:44561", "accept": false, - "reason": "Both are save/load related, but one is config token IDs being lost and the other is a wrong custom module being loaded. Related theme, not the same concrete bug." + "reason": "AutoTokenizer ignoring tokenizer.json is unrelated to the FX-availability API removal regression." }, { "left": "issue:41720", "right": "issue:45229", "accept": false, - "reason": "Qwen3 auto device-mapping CUDA assert vs Gemma4 multi-GPU OOM. Different models and distinct root causes." + "reason": "Qwen3 auto-device-map CUDA assert and Gemma4 multi-GPU OOM are different inference failures." }, { - "left": "issue:43278", - "right": "issue:43329", + "left": "issue:44977", + "right": "issue:45563", "accept": false, - "reason": "Embedding dtype drift in eval vs undefined variables in the video branch of multimodal token counting. Unrelated." + "reason": "Flash-attention generation bug and stale num_return_sequences warning do not share the same failure mode." }, { - "left": "issue:44977", - "right": "issue:45563", + "left": "issue:44568", + "right": "issue:45491", "accept": false, - "reason": "Flash-attention generation failure for Qwen3.5 vs a stale warning in paged generate(). Different code paths." + "reason": "Tokenizer BOS/EOS omission and Gemma3 NaN embeddings are unrelated." }, { - "left": "issue:10000", - "right": "issue:15354", + "left": "issue:43278", + "right": "issue:43329", "accept": false, - "reason": "German DistilBertModel issue vs torch.jit GeneratorExp export limitation. No substantive overlap." + "reason": "Embedding dtype drift in eval and undefined multimodal-token helpers in the video branch are separate bugs." }, { - "left": "issue:44568", - "right": "issue:45491", + "left": "issue:10000", + "right": "issue:15354", "accept": false, - "reason": "Tokenizer special-token handling vs NaN embeddings in sliding-window attention. Different classes of bugs." + "reason": "German DistilBertModel issue and torch.jit GeneratorExp export failure are unrelated." }, { "left": "issue:42490", "right": "issue:43575", "accept": false, - "reason": "Trailing-slash local path warning vs tensor-parallel OOM on a large Qwen model. Not duplicates." + "reason": "Trailing-slash import warning and Qwen2 tensor-parallel OOM are not the same bug." }, { "left": "issue:45372", "right": "issue:45588", "accept": false, - "reason": "Gemma4 processor loading import error vs flash-attention crash on sink-less models. Same general family only, but not the same bug." + "reason": "Gemma4 processor import dependency failure and flash_attention.py s_aux=None crash are different code paths." }, { "left": "issue:43412", "right": "issue:45588", "accept": false, - "reason": "Executorch export/erfinv support issue vs flash-attention runtime attribute crash. Distinct issues." + "reason": "Executorch export issues for gemma3n and sink-less flash-attention AttributeError are unrelated." }, { - "left": "issue:44857", - "right": "issue:45588", + "left": "issue:45198", + "right": "issue:45803", "accept": false, - "reason": "AMP/CUDA loss crash vs flash-attention AttributeError. Different layers and failures." + "reason": "Wav2Vec2 save/tokenization failure and Fuyu bare-except masking errors are different bugs." }, { - "left": "issue:45198", - "right": "issue:45803", + "left": "issue:44857", + "right": "issue:45588", "accept": false, - "reason": "Wav2Vec2 save/tokenization failure vs a bare-except bug in FuyuBatchFeature. Related to tokenization only at a very high level." + "reason": "CUDA AMP crash in LwDetrImageLoss is unrelated to the flash_attention s_aux=None AttributeError." }, { "left": "issue:43976", "right": "issue:45440", "accept": false, - "reason": "Python version compatibility issue vs DeepSeekV3MoE divergence from remote implementation. Unrelated." + "reason": "Python version compatibility issue and DeepSeekV3MoE divergence are unrelated." }, { "left": "issue:45372", "right": "issue:45406", "accept": false, - "reason": "Both concern Gemma4 processor loading/serving, but one is an external dependency import error and the other is a missing _tokenizer attribute. Not the same concrete defect." + "reason": "Gemma4 processor import error and Gemma4Processor missing _tokenizer in serve are related by model family only, not the same underlying defect." }, { "left": "issue:45459", "right": "issue:45712", "accept": false, - "reason": "Tokenizer error masking when protobuf is absent vs leftover dummy classes leaking into dir(transformers). Different maintenance issues." - }, - { - "left": "issue:45588", - "right": "issue:45823", - "accept": false, - "reason": "Flash-attention sink-less crash vs Gemma4 PLE device mismatch under device_map='auto'. Different code paths and symptoms." + "reason": "Tokenizer protobuf exception handling and leftover dummy classes are separate housekeeping bugs." }, { "left": "issue:43278", "right": "issue:43381", "accept": false, - "reason": "Eval-mode dtype change and gradient-checkpointing-in-eval are not the same bug." + "reason": "Eval dtype change and gradient checkpointing in eval mode are unrelated." }, { - "left": "issue:43653", - "right": "issue:44743", + "left": "issue:45588", + "right": "issue:45823", "accept": false, - "reason": "BigBirdTokenizer special-token registration vs Qwen3_5 recurrent state reset with cache. Unrelated." + "reason": "Flash-attention s_aux=None crash and Gemma4 PLE device mismatch are different forward-pass bugs." }, { "left": "issue:44448", "right": "issue:44466", "accept": false, - "reason": "Different output for Pegasus across versions vs inconsistent lm_head.weight serialization by device. Both are regressions, but not the same underlying defect." + "reason": "Pegasus v4/v5 output divergence and lm_head serialization inconsistency are separate regressions." + }, + { + "left": "issue:43653", + "right": "issue:44743", + "accept": false, + "reason": "BigBirdTokenizer special-token registration and Qwen3 recurrent-state reset are unrelated." }, { "left": "issue:45245", "right": "issue:45563", "accept": false, - "reason": "Category-count limit error vs stale warning in paged generate(). No shared failure path." + "reason": "2^24 category limit error and stale generate warning are not the same issue." }, { "left": "issue:25251", "right": "issue:32090", "accept": false, - "reason": "Pipeline top_k shape change vs trainer broadcast NoneType error. Completely different issues." + "reason": "Pipeline top_k nesting behavior and Trainer GPU broadcast TypeError are unrelated." } ] }, @@ -8455,7 +8421,7 @@ "component_index": 1, "component_count": 17, "review_unit_index": 27, - "review_unit_count": 50, + "review_unit_count": 49, "status": "reviewed", "reason": null, "source": "llm", @@ -8517,8 +8483,8 @@ "issue:43065|issue:43994", "issue:45372|issue:45459", "issue:44186|issue:45198", - "issue:43065|issue:45412", "issue:44871|issue:44977", + "issue:43065|issue:45412", "issue:43611|issue:46121", "issue:44361|issue:45198", "issue:42915|issue:42994", @@ -8535,223 +8501,216 @@ "issue:43421|issue:45491", "issue:43653|issue:44442", "issue:43381|issue:43701", - "issue:44610|issue:45412", "issue:43824|issue:44279", + "issue:44610|issue:45412", "issue:45440|issue:45593", "issue:29127|issue:33357", "issue:29942|issue:36010", "issue:43526|issue:45834", - "issue:43761|issue:45859", - "issue:43526|issue:44186" + "issue:43761|issue:45859" ], - "prepared_review_unit_hash": "6cff26d722d28029ba9168a4ad527f4a4695db93aa54adb8100e2456e25b7610", + "prepared_review_unit_hash": "796b20125a79070af577c3e58d35d8d42e82efe2a0bf52a9b8fea327e45922c0", "component_budget": { "node_count": 426, "item_count": 426, - "soft_pair_count": 1572, - "serialized_chars": 814130, - "estimated_input_tokens": 203533, - "estimated_eval_tokens": 407322 + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 30, - "serialized_chars": 36626, - "estimated_input_tokens": 9157, - "estimated_eval_tokens": 18570 + "soft_pair_count": 29, + "serialized_chars": 36334, + "estimated_input_tokens": 9084, + "estimated_eval_tokens": 18424 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These issues are thematically adjacent in the broad sense of Transformers regressions, but none of the soft pairs look like true duplicates of the same concrete bug. The closest overall representative is the open quantization/offloading report, but the set is too diverse to collapse further.", - "confidence": 0.93, - "canonical_issue_reason": "issue:43873 is the most active open report and the broadest representative of the cluster\u2019s recurring model-loading/quantization/offloading failure mode.", + "summary": "This cluster is highly heterogeneous: the items span tokenizers, trainer state, model loading, quantization, vision processors, export, and memory/offload bugs. I do not see any pair that clearly describes the same underlying bug or fix, so all soft edges are rejected. If a representative issue is needed, issue 44462 is the strongest general loader-path bug among the set.", + "confidence": 0.95, + "canonical_issue_reason": "No single canonical duplicate emerges; the issues are unrelated across many subsystems and failure modes.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43873 is the best general triage anchor because it is open, has inbound references, and sits in the most recurring operational area among these items.", + "best_issue_reason": "issue:44462 is a clear, user-facing, reproducible loading bug with broad impact in a common path, so it is the best representative issue in this mixed set.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44206", "right": "issue:44479", "accept": false, - "reason": "Different regressions: one is a feature-extractor argument crash, the other is a video-input regression for Qwen VL models." + "reason": "Different regressions in different code paths: feature extractor argument handling vs multimodal video-input behavior." }, { "left": "issue:43065", "right": "issue:43994", "accept": false, - "reason": "Unrelated models and failures; dummy Conv2d in Sam3PixelDecoder is not the same bug as SigLIP2 producing nonsensical outputs." + "reason": "Unrelated bugs: a dummy Conv2d in Sam3PixelDecoder vs incorrect SigLIP2 outputs with AutoModel/pipeline." }, { "left": "issue:45372", "right": "issue:45459", "accept": false, - "reason": "Both involve loading-time errors, but one is a mistral_common import breakage while the other is protobuf-exception masking in tokenizer error handling." + "reason": "Different tokenizer-loading failures: Gemma 4 processor import issue vs protobuf-missing decode error masking." }, { "left": "issue:44186", "right": "issue:45198", "accept": false, - "reason": "Different tokenizer code paths and symptoms; NER/padding crashes in LayoutLMv2Tokenizer are not the same as Wav2Vec2 save/tokenization failures." - }, - { - "left": "issue:43065", - "right": "issue:45412", - "accept": false, - "reason": "Memory reclamation in RT-DETR is a different problem than the Sam3PixelDecoder dummy layer." + "reason": "Different model families and failures: LayoutLMv2 tokenizer NER/padding crash vs Wav2Vec2 save/tokenization failure." }, { "left": "issue:44871", "right": "issue:44977", "accept": false, - "reason": "Distinct model-specific configuration and generation bugs; eos_token_id mismatch is unrelated to flash-attention generation failure." + "reason": "Different bugs: eos_token_id config mismatch vs flash-attention generation regression." }, { - "left": "issue:43611", + "left": "issue:43065", + "right": "issue:45412", + "accept": false, + "reason": "Different problem classes: model layer placeholder vs RT-DETR memory release/GC issue." + }, + { + "left": "issue:43611", "right": "issue:46121", "accept": false, - "reason": "One is a Transformers 5 base_model_prefix loading regression; the other is a rope-parameter helper type error." + "reason": "Different loading/config bugs: base_model_prefix breakage vs rope-validation TypeError on JSON-loaded lists." }, { "left": "issue:44361", "right": "issue:45198", "accept": false, - "reason": "Different tokenizer/model classes and failure mechanisms; MLuke task errors do not match Wav2Vec2 save/tokenization bugs." + "reason": "Different tokenizers and failure modes: MLuke task AttributeError vs Wav2Vec2 save/tokenization failure." }, { "left": "issue:42915", "right": "issue:42994", "accept": false, - "reason": "Both mention quantization-adjacent workflows, but one is MoE weight conversion with accelerate/deepspeed and the other is saving quantized models." + "reason": "Different quantization issues: FineGrainedFP8Config failure vs quantized model saving." }, { "left": "issue:42490", "right": "issue:43873", "accept": false, - "reason": "A trailing-slash local path import warning is not the same underlying issue as quantization offloading not working." + "reason": "Different paths: local-path slash import warning vs quantization/offloading behavior." }, { "left": "issue:32090", "right": "issue:33290", "accept": false, - "reason": "Trainer broadcast NoneType and Adafactor OOM are unrelated training failures." + "reason": "Different failures: NoneType broadcast error in Trainer vs OOM with Adafactor and DeepSpeed." }, { "left": "issue:33357", "right": "issue:39692", "accept": false, - "reason": "MacOS bus error in CLIP loading is different from SigLIP2 documentation example errors." + "reason": "Different scope: MacOS bus error for a pretrained CLIP model vs documentation example/model mismatch and quantization errors for SigLIP2." }, { "left": "issue:43066", "right": "issue:44462", "accept": false, - "reason": "Both concern tokenizer loading, but one is a decoder type mismatch in v5 and the other is AutoTokenizer ignoring tokenizer.json." + "reason": "Both involve tokenizer loading, but the concrete bugs differ: decoder type in v5 vs AutoTokenizer ignoring repository tokenizer.json." }, { "left": "issue:29942", "right": "issue:30064", "accept": false, - "reason": "Flash Attention 2 test failures and void segmentation-map processing are unrelated subsystems." + "reason": "Unrelated test failure vs image processor void-segmentation-map handling." }, { "left": "issue:29127", "right": "issue:34567", "accept": false, - "reason": "LayoutLMv3 error-message clarity and TrainerState token counting are not the same bug." + "reason": "Different concerns: clearer error messaging for LayoutLMv3 boxes vs TrainerState token-count tracking." }, { "left": "issue:43994", "right": "issue:45412", "accept": false, - "reason": "SigLIP2 inference quality issues are unrelated to RT-DETR memory release behavior." + "reason": "Different issues: SigLIP2 bad outputs vs RT-DETR memory leak." }, { "left": "issue:44265", "right": "issue:45198", "accept": false, - "reason": "torch.export failing on torch_compilable_check is a different code path from Wav2Vec2 tokenization/save failures." + "reason": "Different code paths: torch.export compatibility with torch_compilable_check vs Wav2Vec2 save/tokenization." }, { "left": "issue:43257", "right": "issue:43278", "accept": false, - "reason": "MoE weight conversion under distributed loading is not the same as embedding dtype drift during evaluation." + "reason": "Different failures: Qwen3 MoE weights not converting under accelerate+deepspeed vs dtype drift between train and eval." }, { "left": "issue:43756", "right": "issue:45440", "accept": false, - "reason": "Different model implementations and different bugs: RoPE layer loss vs DeepSeekV3MoE divergence." + "reason": "Different model-specific bugs: Smollm3 RoPE layer mismatch vs DeepseekV3MoE divergence from remote implementation." }, { "left": "issue:43421", "right": "issue:45491", "accept": false, - "reason": "Runtime special-token post-processing updates are unrelated to Gemma3 NaN embeddings on padded windows." + "reason": "Different subsystems: runtime special-token post-processing vs Gemma3 NaN embeddings with sliding-window attention." }, { "left": "issue:43653", "right": "issue:44442", "accept": false, - "reason": "BigBird special-token registration and AutoTokenizer loading of FastSpeech2ConformerTokenizer are different tokenizer issues." + "reason": "Different tokenizer problems: BigBird mask-token registration vs AutoTokenizer loading FastSpeech2ConformerTokenizer." }, { "left": "issue:43381", "right": "issue:43701", "accept": false, - "reason": "Gradient checkpointing in eval mode and resume_from_checkpoint key mismatch are separate trainer bugs." + "reason": "Different behaviors: gradient checkpointing in eval mode vs resume_from_checkpoint key mismatch." }, { - "left": "issue:44610", - "right": "issue:45412", + "left": "issue:43824", + "right": "issue:44279", "accept": false, - "reason": "Processor image-size mismatch is unrelated to RT-DETR memory not being released." + "reason": "Different import/dependency issues: missing Qwen2.5-VL symbol vs generic transformers dependency problem." }, { - "left": "issue:43824", - "right": "issue:44279", + "left": "issue:44610", + "right": "issue:45412", "accept": false, - "reason": "Missing Qwen2_5_VL export and a generic dependency issue do not describe the same failure." + "reason": "Different bugs: processor output size mismatch vs model memory not being released." }, { "left": "issue:45440", "right": "issue:45593", "accept": false, - "reason": "DeepSeekV3MoE implementation divergence is not the same as missing auxiliary losses in D-FINE." + "reason": "Different model behavior: DeepseekV3MoE implementation divergence vs D-FINE auxiliary-loss omission." }, { "left": "issue:29127", "right": "issue:33357", "accept": false, - "reason": "These are unrelated: LayoutLMv3 error messaging versus a CLIP bus error on macOS." + "reason": "Different domains and failures: LayoutLMv3 error clarity vs MacOS bus error for CLIP." }, { "left": "issue:29942", "right": "issue:36010", "accept": false, - "reason": "Flash Attention test failures are unrelated to the GenerationMixin import error." + "reason": "Different issues: Flash Attention 2 test failures vs GenerationMixin import error." }, { "left": "issue:43526", "right": "issue:45834", "accept": false, - "reason": "BeitImageProcessorFast reduce_labels behavior and Kosmos2.5 OCR index errors are different bugs." + "reason": "Different processor bugs: BeitImageProcessorFast reduce_labels behavior vs Kosmos2.5 long-OCR index error." }, { "left": "issue:43761", "right": "issue:45859", "accept": false, - "reason": "CLIPVisionModel hidden_states regression is unrelated to Qwen3_5Moe CPU offload incompatibility." - }, - { - "left": "issue:43824", - "right": "issue:44279", - "accept": false, - "reason": "A missing model class import is not the same as a dependency-resolution problem." + "reason": "Different regressions: CLIPVisionModel hidden_states not returned vs Qwen3_5Moe CPU offload incompatibility." } ] }, @@ -8767,7 +8726,7 @@ "component_index": 1, "component_count": 17, "review_unit_index": 28, - "review_unit_count": 50, + "review_unit_count": 49, "status": "reviewed", "reason": null, "source": "llm", @@ -8799,6 +8758,7 @@ "issue:44062", "issue:44077", "issue:44079", + "issue:44186", "issue:44206", "issue:44387", "issue:44442", @@ -8825,6 +8785,7 @@ ], "soft_pairs": [ "issue:45103|issue:45372", + "issue:43526|issue:44186", "issue:44625|issue:45584", "issue:29127|issue:35141", "issue:43334|issue:43531", @@ -8839,195 +8800,201 @@ "issue:43575|issue:44387", "issue:44743|issue:44811", "issue:43646|issue:44077", - "issue:43441|issue:45588", "issue:44062|issue:45907", + "issue:43441|issue:45588", "issue:43761|issue:44079", "issue:43723|issue:45479", "issue:45003|issue:45440", "issue:44479|issue:45290", - "issue:44442|issue:44898", "issue:43232|issue:43388", + "issue:44442|issue:44898", "issue:33290|issue:34567", "issue:33290|issue:34689" ], - "prepared_review_unit_hash": "011418054d331e3bee2bfdb414ec4a990724e5978b365292673278d450b37565", + "prepared_review_unit_hash": "0c003826f099711a1df0ae44136d52b8a3bc9871b6cf8afef872dbaf053ab1e7", "component_budget": { "node_count": 426, "item_count": 426, - "soft_pair_count": 1572, - "serialized_chars": 814130, - "estimated_input_tokens": 203533, - "estimated_eval_tokens": 407322 + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 }, "budget": { - "node_count": 47, - "item_count": 47, - "soft_pair_count": 25, - "serialized_chars": 34816, - "estimated_input_tokens": 8704, - "estimated_eval_tokens": 17664 + "node_count": 48, + "item_count": 48, + "soft_pair_count": 26, + "serialized_chars": 35690, + "estimated_input_tokens": 8923, + "estimated_eval_tokens": 18102 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is heterogeneous and appears to be a similarity false-positive bucket rather than a true duplicate set. None of the soft pairs describe the same underlying bug/change closely enough to merge, and there are no PRs in the cluster.", - "confidence": 0.93, - "canonical_issue_reason": "issue:43575 is the clearest single bug report in the set and the best anchor for follow-up: a concrete OOM failure tied to model loading/tensor-parallel inference. It is still not a duplicate of the other issues, but it is the most representative issue-shaped artifact here.", + "summary": "This cluster is heterogeneous: the soft pairs mostly connect unrelated bugs that only share broad themes like model loading, generation, or tokenizer behavior. I would not merge any of the proposed soft edges.", + "confidence": 0.66, + "canonical_issue_reason": "If one issue must represent the cluster, issue 45584 is the strongest standalone tracker: it is open, has inbound references, and describes a concrete reproducible runtime failure. That said, the items here are not true duplicates of one another.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43575 is the best overall issue candidate because it is a specific, actionable failure mode with a clear reproduction path and it sits near the only other memory-related reports in the cluster.", + "best_issue_reason": "Issue 45584 is the best single issue to keep as a representative because it has clear user impact and the most discussion/inbound references among the open issues. It still should not absorb the others as duplicates.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45103", "right": "issue:45372", "accept": false, - "reason": "Different failures: auto-docstring annotation handling vs a mistral_common import error in Gemma 4 processor loading." + "reason": "Different failures in different loading paths: docstring parsing with future annotations vs Gemma processor import dependency breakage." + }, + { + "left": "issue:43526", + "right": "issue:44186", + "accept": false, + "reason": "Different tokenizer/image-processor bugs affecting different models and inputs; no same concrete code path." }, { "left": "issue:44625", "right": "issue:45584", "accept": false, - "reason": "Unrelated bugs: Qwen3.5 label propagation vs Whisper empty-transcription generation after token alignment." + "reason": "Qwen3.5 config propagation bug vs Whisper empty-transcription generation failure are unrelated." }, { "left": "issue:29127", "right": "issue:35141", "accept": false, - "reason": "Different code paths: LayoutLMv3 error messaging vs token-embedding resize/post_init reinitialization." + "reason": "One is a LayoutLMv3 error-message clarity issue, the other is an embedding reinitialization bug during resizing." }, { "left": "issue:43334", "right": "issue:43531", "accept": false, - "reason": "Same broad Qwen family, but different concrete problems: missing pad_token_id vs sliding-window behavior in Qwen3-MoE." + "reason": "Missing pad_token_id in Qwen3-VL text config is unrelated to Qwen3-MoE sliding-window behavior." }, { "left": "issue:44442", "right": "issue:45072", "accept": false, - "reason": "Tokenizer loading failure and bfloat16 dtype mismatch are distinct issues, even though both are inference-related." + "reason": "AutoTokenizer loading failure and bfloat16 dtype mismatch are different bugs in different subsystems." }, { "left": "issue:42915", "right": "issue:43931", "accept": false, - "reason": "Both involve Qwen3 variants, but one is FineGrainedFP8Config failure and the other is a weight-shape mismatch during model loading." + "reason": "Different Qwen model families and different failure modes: FP8 config handling vs weight-shape mismatch." }, { "left": "issue:44977", "right": "issue:45245", "accept": false, - "reason": "Completely different topics: Qwen3.5 flash-attention generation vs category-cardinality limits in another subsystem." + "reason": "Flash-attention generation regression is unrelated to the category-count runtime error." }, { "left": "issue:43526", "right": "issue:44857", "accept": false, - "reason": "Different components and failures: BeitImageProcessorFast label reduction vs LwDetrImageLoss AMP crash." + "reason": "BeitImageProcessorFast label reduction bug and LwDetr AMP crash are distinct issues." }, { "left": "issue:42490", "right": "issue:43381", "accept": false, - "reason": "A local-path import warning is not the same bug as gradient checkpointing being disallowed in eval mode." + "reason": "Local-path import warning and gradient-checkpointing eval-mode restriction are not the same bug." }, { "left": "issue:44464", "right": "issue:45491", "accept": false, - "reason": "Different runtime issues: compiled chunked generation inconsistency vs Gemma3 NaNs from sliding-window/all-padding behavior." + "reason": "Compiled-forward chunked generation inconsistency is separate from Gemma3 sliding-window NaN embeddings." }, { "left": "issue:43575", "right": "issue:43701", "accept": false, - "reason": "OOM on tensor-parallel model loading is not the same as a checkpoint key mismatch during resume." + "reason": "Tensor-parallel OOM on model load and resume-from-checkpoint key mismatch are unrelated." }, { "left": "issue:43262", "right": "issue:44206", "accept": false, - "reason": "Audio chat-template sample-rate defaulting and a feature-extractor center-arg crash are unrelated." + "reason": "Audio chat-template sampling-rate defaulting and LaserFeatureExtractor center-arg crash are different code paths." }, { "left": "issue:43575", "right": "issue:44387", "accept": false, - "reason": "Both are memory-related, but one is TP load OOM for a specific Qwen2 model and the other is increased reserved memory under int4 quantization; not the same bug path." + "reason": "Both mention OOM, but one is tp load failure and the other is int4 quantization memory growth; not the same underlying bug." }, { "left": "issue:44743", "right": "issue:44811", "accept": false, - "reason": "Recurrent-state reset with cache and Whisper batch_decode special-token handling are unrelated." + "reason": "Recurrent-state cache reset in Qwen3.5 is unrelated to Whisper batch_decode skip_special_tokens behavior." }, { "left": "issue:43646", "right": "issue:44077", "accept": false, - "reason": "General custom-model initialization regression and patchtsmixer post_init policy are not the same code-path problem." + "reason": "General Transformers 5.0 custom-init breakage is broader than the specific patchtsmixer post_init policy issue." }, { - "left": "issue:43441", - "right": "issue:45588", + "left": "issue:44062", + "right": "issue:45907", "accept": false, - "reason": "Both mention flash attention, but they target different crashes in different model setups and are not mergeable into one fix." + "reason": "AddedToken special-flag collision and list_repo_files revision forwarding are unrelated." }, { - "left": "issue:44062", - "right": "issue:45907", + "left": "issue:43441", + "right": "issue:45588", "accept": false, - "reason": "AddedToken constructor conflict and list_repo_files revision forwarding are unrelated." + "reason": "Different flash-attention crashes in different models/code paths; not mergeable as one fix." }, { "left": "issue:43761", "right": "issue:44079", "accept": false, - "reason": "One is a CLIPVisionModel hidden_states regression; the other is ModelOutput key assignment when a prior value was None." + "reason": "Missing hidden_states output in CLIPVisionModel is not the same as ModelOutput key assignment for previously-None keys." }, { "left": "issue:43723", "right": "issue:45479", "accept": false, - "reason": "Tokenizer loading failure and single-label-classification zero-loss behavior are unrelated." + "reason": "Tokenizer loading failure and degenerate zero-loss classification setup are unrelated." }, { "left": "issue:45003", "right": "issue:45440", "accept": false, - "reason": "Unsafe sys.modules access in modeling_utils is unrelated to DeepseekV3MoE diverging from the remote implementation." + "reason": "Unsafe sys.modules access in modeling_utils is unrelated to DeepseekV3MoE parity with remote implementation." }, { "left": "issue:44479", "right": "issue:45290", "accept": false, - "reason": "Video-input regression across Qwen VL models is a different bug from apply_chat_template crashing on tool-call assistant messages." + "reason": "Video-input regression in Qwen VL models and chat-template crash on tool-call assistant messages are different bugs." }, { - "left": "issue:44442", - "right": "issue:44898", + "left": "issue:43232", + "right": "issue:43388", "accept": false, - "reason": "Both affect model loading/inference, but one is AutoTokenizer loading and the other is Perceiver image classification at non-default resolution; no shared fix path." + "reason": "Generation bookkeeping after sync_gpus is unrelated to gather_for_metrics dropping tuple labels." }, { - "left": "issue:43232", - "right": "issue:43388", + "left": "issue:44442", + "right": "issue:44898", "accept": false, - "reason": "Generation-kwargs update after sync_gpus and gather_for_metrics tuple-label truncation are distinct training/generation bugs." + "reason": "Tokenizer loading failure is unrelated to Perceiver interpolation at non-default resolution." }, { "left": "issue:33290", "right": "issue:34567", "accept": false, - "reason": "DeepSpeed Adafactor OOM and TrainerState num_input_tokens_seen not updating are unrelated." + "reason": "AdaFactor/deepspeed OOM and TrainerState token counter not updating are unrelated." }, { "left": "issue:33290", "right": "issue:34689", "accept": false, - "reason": "An optimizer OOM report and a model-loading regression are different underlying problems despite both being runtime failures." + "reason": "Optimizer memory issue and Llama 3.2 Vision loading regression do not share the same bug." } ] }, @@ -9043,7 +9010,7 @@ "component_index": 1, "component_count": 17, "review_unit_index": 29, - "review_unit_count": 50, + "review_unit_count": 49, "status": "reviewed", "reason": null, "source": "llm", @@ -9123,209 +9090,209 @@ "issue:44855|issue:45137", "issue:25251|issue:28282", "issue:43540|issue:44514", - "issue:43295|issue:43825", "issue:44162|issue:44792", - "issue:43316|issue:44877", + "issue:43295|issue:43825", "issue:43065|issue:44617", + "issue:43316|issue:44877", "issue:43316|issue:43716", "issue:43404|issue:43441" ], - "prepared_review_unit_hash": "5bcbd86ca1ba366f0a007bc33701240c37ece2852983d3722a3adc3226900380", + "prepared_review_unit_hash": "e12d8853e66d89f46dd445453e31cbe7e618e1abaa47f7ef3785c0bf68e8e667", "component_budget": { "node_count": 426, "item_count": 426, - "soft_pair_count": 1572, - "serialized_chars": 814130, - "estimated_input_tokens": 203533, - "estimated_eval_tokens": 407322 + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 28, - "serialized_chars": 36192, - "estimated_input_tokens": 9048, - "estimated_eval_tokens": 18352 + "serialized_chars": 36197, + "estimated_input_tokens": 9050, + "estimated_eval_tokens": 18356 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is not a duplicate set; it mixes many unrelated Transformers issues across configs, loading/saving, multimodal processors, training, and docs. None of the soft-edge pairs look like the same underlying bug/change.", - "confidence": 0.94, - "canonical_issue_reason": "No single canonical issue: the items are heterogeneous and do not converge on one underlying bug.", + "summary": "The cluster is heterogeneous: most items are unrelated model-loading, config, tokenizer, or multimodal regressions that only share broad Transformers vocabulary. I do not see any pair that is the same underlying bug/change, so no soft edges should be merged.", + "confidence": 0.41, + "canonical_issue_reason": "Issue 43611 is the broadest representative of the set because it describes a general v5 model-loading regression, but the cluster as a whole is too mixed to have a strong true duplicate nucleus.", "canonical_pr_reason": null, - "best_issue_reason": "No global best issue; none is a clear representative of the others because the cluster is not semantically cohesive.", + "best_issue_reason": "Issue 43611 is the best single issue to anchor on because it is broad and systemic, whereas most others are narrower model-, config-, or processor-specific regressions.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43316", "right": "issue:43525", "accept": false, - "reason": "Both are config API mismatches, but one is Gemma3TextConfig-specific while the other is a missing pad_token_id on Llama4Config; different bugs." + "reason": "Both are config-related, but they affect different models and different fields; not the same bug." }, { "left": "issue:42490", "right": "issue:44743", "accept": false, - "reason": "Local path warning vs recurrent-state reset with cache; unrelated subsystems and failure modes." + "reason": "A local-path import warning and a recurrent-state cache bug are unrelated." }, { "left": "issue:33290", "right": "issue:35141", "accept": false, - "reason": "OOM with Adafactor/DeepSpeed is a training-memory issue, while the other is embedding reinitialization after resizing; not the same bug." + "reason": "Deepspeed/Adafactor OOM and embedding reinitialization in post_init are different code paths." }, { "left": "issue:43381", "right": "issue:43873", "accept": false, - "reason": "Eval-mode gradient checkpointing and quantized offloading are separate runtime behaviors with different root causes." + "reason": "Gradient-checkpointing eval-mode behavior and quantization offloading are separate issues." }, { "left": "issue:44442", "right": "issue:45803", "accept": false, - "reason": "Tokenizer loading failure and bare except in tensor conversion are unrelated." + "reason": "Tokenizer-loading failure and a bare-except bug in tensor conversion are not the same defect." }, { "left": "issue:43761", "right": "issue:44188", "accept": false, - "reason": "Hidden_states being None in CLIPVisionModel is unrelated to attention kernel divergence under torch.compile." + "reason": "CLIP hidden_states regression and attention-kernel divergence under torch.compile are unrelated." }, { "left": "issue:34689", "right": "issue:43611", "accept": false, - "reason": "Both concern loading, but one is a model-specific regression and the other is a base_model_prefix loading regression; different code paths." + "reason": "Both are loading regressions, but for different causes and model families; not mergeable as one fix." }, { "left": "issue:43720", "right": "issue:44442", "accept": false, - "reason": "Packed-weight unloading during accelerate loading is unrelated to tokenizer class loading." + "reason": "BitNet packed-weight loading and tokenizer class loading are unrelated." }, { "left": "issue:44077", "right": "issue:44479", "accept": false, - "reason": "Post-init allowance in patchtsmixer and Qwen video-input regression are separate issues." + "reason": "A model post_init policy issue and a video-input regression are different bugs." }, { "left": "issue:43931", "right": "issue:45216", "accept": false, - "reason": "Both mention Qwen3-VL/Qwen3.5 checkpoints, but one is a weight-shape loading mismatch and the other is incorrect save_pretrained output." + "reason": "Shape mismatch on load and incorrect save_pretrained checkpoint contents are different failure modes." }, { "left": "issue:43493", "right": "issue:43827", "accept": false, - "reason": "SigLIP2 implementation discrepancy and stale docs referencing pipeline() are unrelated." + "reason": "SigLIP2 implementation mismatch and outdated docs referencing pipeline are unrelated." }, { "left": "issue:42907", "right": "issue:45698", "accept": false, - "reason": "Saving dequantized Ministral/Devstral models is not the same as loading the wrong custom module after save_pretrained." + "reason": "Both touch save/load flows, but one is dequantized-model saving and the other is wrong custom-module loading after save_pretrained." }, { "left": "issue:34689", "right": "issue:36010", "accept": false, - "reason": "Model-loading regression vs import error for GenerationMixin; different failures." + "reason": "A model-loading regression and a missing GenerationMixin import are not the same issue." }, { "left": "issue:43316", "right": "issue:45759", "accept": false, - "reason": "Both involve config handling, but one is API discrepancy for Gemma3TextConfig and the other is unwrapping text_config in composite Qwen configs; not the same bug." + "reason": "Both involve config APIs, but they concern different models and different handling of text_config." }, { "left": "issue:33290", "right": "issue:36331", "accept": false, - "reason": "OOM in optimizer/deepspeed is unrelated to a custom trainer signature mismatch." + "reason": "DeepSpeed memory use and a custom trainer signature change are unrelated." }, { "left": "issue:43749", "right": "issue:43931", "accept": false, - "reason": "FSDP CPU RAM efficient loading and Qwen3-VL weight-shape mismatch are distinct loading problems." + "reason": "FSDP CPU RAM efficient loading and Qwen3-VL shape mismatch are different loading bugs." }, { "left": "issue:43873", "right": "issue:44871", "accept": false, - "reason": "Quantization offloading bug and Gemma-3 eos_token_id inconsistency are unrelated." + "reason": "Quantization offloading and eos_token_id config inconsistency do not share a concrete fix." }, { "left": "issue:43643", "right": "issue:44913", "accept": false, - "reason": "Missing fields with trust_remote_code and rotary_pct not persisting on reload are separate config issues." + "reason": "Missing fields with trust_remote_code and GPTNeoX rotary_pct not persisting are different config-serialization problems." }, { "left": "issue:44190", "right": "issue:44496", "accept": false, - "reason": "Local dataset loading failure and unrecognized model_type are unrelated." + "reason": "Local dataset loading in a training script and an unrecognized model config are unrelated." }, { "left": "issue:44855", "right": "issue:45137", "accept": false, - "reason": "Python 3.13 torch.jit parsing error is unrelated to DeepSpeed ZeRO3 deque exhaustion." + "reason": "Python 3.13 import parsing and DeepSpeed ZeRO3 deque underflow are unrelated." }, { "left": "issue:25251", "right": "issue:28282", "accept": false, - "reason": "Pipeline top_k nesting behavior and missing PyTorch import are completely different issues." + "reason": "Pipeline top_k shape change and missing PyTorch ImportError are unrelated." }, { "left": "issue:43540", "right": "issue:44514", "accept": false, - "reason": "Both involve Qwen multimodal inputs, but one is video processing in Qwen3OmniMoe and the other is batched chat-template handling in Qwen2_5_VLProcessor; not the same bug." - }, - { - "left": "issue:43295", - "right": "issue:43825", - "accept": false, - "reason": "Processor/tokenizer regression and pipeline translation error messaging are unrelated v5 regressions." + "reason": "Both are Qwen multimodal processing bugs, but one is video-input handling and the other is batched chat-template padding behavior." }, { "left": "issue:44162", "right": "issue:44792", "accept": false, - "reason": "ESM2 breakage and janus test failure do not share an underlying code path." + "reason": "ESM2 breakage and a janus image-generation test failure are unrelated." }, { - "left": "issue:43316", - "right": "issue:44877", + "left": "issue:43295", + "right": "issue:43825", "accept": false, - "reason": "Both are config-related, but Gemma3TextConfig API discrepancy is not the same as strict config blocking granite_speech loading." + "reason": "Processor/tokenizer regression and pipeline error-message wording are different issues." }, { "left": "issue:43065", "right": "issue:44617", "accept": false, - "reason": "Dummy Conv2d in Sam3PixelDecoder and Sam3Video CUDA OOM are different Sam3 issues." + "reason": "A dummy Conv2d in Sam3PixelDecoder and Sam3Video OOM are not the same bug." + }, + { + "left": "issue:43316", + "right": "issue:44877", + "accept": false, + "reason": "Both are config-related, but they affect different model families and different loading logic." }, { "left": "issue:43316", "right": "issue:43716", "accept": false, - "reason": "Gemma3TextConfig API inconsistency and Mistral-3 image preprocessor dtype mismatch are unrelated." + "reason": "Gemma3 config API inconsistency and Mistral-3 dtype mismatch are unrelated." }, { "left": "issue:43404", "right": "issue:43441", "accept": false, - "reason": "Both are Mistral-family, but one is tied lm_head weights and the other is FlashAttention failure; distinct fixes." + "reason": "One is lm_head weight tying, the other is FlashAttention failure; different code paths." } ] }, @@ -9341,7 +9308,7 @@ "component_index": 1, "component_count": 17, "review_unit_index": 30, - "review_unit_count": 50, + "review_unit_count": 49, "status": "reviewed", "reason": null, "source": "llm", @@ -9400,15 +9367,15 @@ ], "soft_pairs": [ "issue:43550|issue:43701", + "issue:16998|issue:22355", "issue:45803|issue:46170", "issue:43606|issue:43701", - "issue:16998|issue:22355", "issue:44077|issue:44534", "issue:44871|issue:45584", - "issue:42617|issue:43575", "issue:43673|issue:44743", - "issue:43575|issue:43742", + "issue:42617|issue:43575", "issue:41950|issue:43475", + "issue:43575|issue:43742", "issue:43295|issue:44871", "issue:43388|issue:43688", "issue:43761|issue:44479", @@ -9429,20 +9396,20 @@ "issue:44479|issue:44734", "issue:44464|issue:45563" ], - "prepared_review_unit_hash": "de53d5618cdca28c0f3ff52f08c897375d56bd671a9ae198d2fe6a25e343910a", + "prepared_review_unit_hash": "fe936d6a3630a5c35d3ffbb0a03c2135a82934e771528c21ba14614057e613ff", "component_budget": { "node_count": 426, "item_count": 426, - "soft_pair_count": 1572, - "serialized_chars": 814130, - "estimated_input_tokens": 203533, - "estimated_eval_tokens": 407322 + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 29, - "serialized_chars": 36441, + "serialized_chars": 36443, "estimated_input_tokens": 9111, "estimated_eval_tokens": 18478 }, @@ -9451,186 +9418,186 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is heterogeneous: the items span unrelated bugs in model loading, generation, training, multimodal processors, config serialization, and serve/runtime paths. The soft-similarity links do not look like true duplicate pairs, so I would not merge any of them as the same underlying issue.", - "confidence": 0.97, - "canonical_issue_reason": null, + "summary": "The cluster is mostly a set of unrelated Transformers bug reports with superficial title similarity around loading, generation, and model-specific regressions. None of the soft pairs look like the same underlying issue, so all candidate duplicate edges are rejected.", + "confidence": 0.91, + "canonical_issue_reason": "issue:45698 is the best anchor in this set: it is open, has substantial discussion, and describes a concrete reproducible save/load regression affecting custom modules.", "canonical_pr_reason": null, - "best_issue_reason": null, + "best_issue_reason": "issue:45698 is the most suitable representative issue because it is broad, actionable, and centered on a specific loading regression rather than a model-specific one-off failure.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43550", "right": "issue:43701", "accept": false, - "reason": "Torch.compile/SDPA failure and checkpoint resume key mismatch are different code paths and symptoms." + "reason": "Different failures: Bamba torch.compile/SDPA runtime bug vs resume_from_checkpoint key mismatch." + }, + { + "left": "issue:16998", + "right": "issue:22355", + "accept": false, + "reason": "Unrelated topics: model_max_length question vs missing transformers.onnx import." }, { "left": "issue:45803", "right": "issue:46170", "accept": false, - "reason": "Bare except in tensor conversion is unrelated to Windows ROCm import/distributed dependency failures." + "reason": "Different subsystems: Fuyu tensor conversion exception handling vs Windows/ROCm distributed import failure." }, { "left": "issue:43606", "right": "issue:43701", "accept": false, - "reason": "CPU offload device mismatch is a runtime placement bug, not a resume_from_checkpoint key mismatch." - }, - { - "left": "issue:16998", - "right": "issue:22355", - "accept": false, - "reason": "Model max length behavior and missing transformers.onnx module are unrelated." + "reason": "CPU offload device mismatch is unrelated to checkpoint key mismatches." }, { "left": "issue:44077", "right": "issue:44534", "accept": false, - "reason": "Optional post_init deprecation and non-persistent buffer junk are distinct model-initialization issues." + "reason": "One is a post_init API restriction, the other is non-persistent buffer initialization corruption." }, { "left": "issue:44871", "right": "issue:45584", "accept": false, - "reason": "EOS token config inconsistency is unrelated to Whisper empty transcription after align_special_tokens." - }, - { - "left": "issue:42617", - "right": "issue:43575", - "accept": false, - "reason": "3d_parallel.py execution failure and TP OOM on Qwen2-57B are different operational problems." + "reason": "Gemma EOS config mismatch and Whisper empty-transcription crash are different bugs." }, { "left": "issue:43673", "right": "issue:44743", "accept": false, - "reason": "Chunked_prefill cache handling and Qwen3.5 recurrent state reset are separate cache bugs." + "reason": "Both involve cache semantics, but the models and concrete failure modes differ materially." }, { - "left": "issue:43575", - "right": "issue:43742", + "left": "issue:42617", + "right": "issue:43575", "accept": false, - "reason": "TP OOM on a large model and a MobileLLM loading key error do not share the same failure mode." + "reason": "3d_parallel.py execution error vs tensor-parallel OOM while loading a Qwen model." }, { "left": "issue:41950", "right": "issue:43475", "accept": false, - "reason": "Video-classification processor selection and SAM3 vision encoder attribute errors are unrelated multimodal bugs." + "reason": "Pipeline processor lookup bug is unrelated to SAM3 vision encoder missing attribute." + }, + { + "left": "issue:43575", + "right": "issue:43742", + "accept": false, + "reason": "OOM on tp load is unrelated to a MobileLLM config/key loading error." }, { "left": "issue:43295", "right": "issue:44871", "accept": false, - "reason": "Processor/tokenizer regression is unrelated to EOS token configuration mismatch." + "reason": "Custom processor/tokenizer regression and Gemma EOS list-vs-scalar mismatch are distinct issues." }, { "left": "issue:43388", "right": "issue:43688", "accept": false, - "reason": "Metric label truncation and auxiliary loss normalization are different training/eval issues." + "reason": "Metric gathering label truncation is unrelated to auxiliary loss normalization." }, { "left": "issue:43761", "right": "issue:44479", "accept": false, - "reason": "CLIP hidden_states regression and video-input regression affect different model families and paths." + "reason": "CLIPVision hidden_states bug and Qwen VL video-input regression affect different code paths." }, { "left": "issue:43452", "right": "issue:44496", "accept": false, - "reason": "gguf_file tokenizer/model loading breakage is not the same as an unrecognized config.json model type." + "reason": "Both are loading-related, but gguf_file handling and unrecognized model_type are different root causes." }, { "left": "issue:45127", "right": "issue:45200", "accept": false, - "reason": "LoRA extended-vocab collapse and Gemma 4 mm_token_type_ids defaults are unrelated." + "reason": "LoRA/tied-embeddings collapse is unrelated to Gemma 4 mm_token_type_ids defaults." }, { "left": "issue:44871", "right": "issue:46050", "accept": false, - "reason": "EOS token mismatch and sinq quantization save/load breakage are distinct bugs." + "reason": "EOS token config mismatch is unrelated to sinq save/load quantization breakage." }, { "left": "issue:43646", "right": "issue:44913", "accept": false, - "reason": "Custom model initialization breakage and GPTNeoX rotary_pct reload are different config/init regressions." + "reason": "General custom model init regression is not the same as GPTNeoX rotary_pct persistence on reload." }, { "left": "issue:43504", "right": "issue:45698", "accept": false, - "reason": "Legacy field preset loading and wrong custom-module reload after save_pretrained are not the same defect." + "reason": "Both involve loading, but BEiT legacy-field preset loading and wrong custom-module selection are different bugs." }, { "left": "issue:32090", "right": "issue:35141", "accept": false, - "reason": "NoneType GPU broadcast error and embedding reinitialization in post_init are unrelated." + "reason": "Trainer gpu_broadcast NoneType error is unrelated to embedding reinitialization in post_init." }, { "left": "issue:44734", "right": "issue:45406", "accept": false, - "reason": "Serve KV-cache indexing crash and Gemma4Processor missing _tokenizer are different serve/runtime failures." + "reason": "Serve KV-cache indexing crash and Gemma4Processor missing _tokenizer are unrelated." }, { "left": "issue:44617", "right": "issue:45584", "accept": false, - "reason": "SAM3Video OOM and Whisper empty-transcription handling are not the same underlying bug." + "reason": "CUDA OOM in Sam3Video is unrelated to Whisper empty transcription after align_special_tokens." }, { "left": "issue:45563", "right": "issue:45584", "accept": false, - "reason": "Stale num_return_sequences warning and Whisper empty transcription failure do not overlap." + "reason": "A stale warning about num_return_sequences is not the same as a transcription crash." }, { "left": "issue:43525", "right": "issue:43531", "accept": false, - "reason": "Missing pad_token_id on Llama4Config and Qwen3-MoE sliding_window behavior are unrelated config bugs." + "reason": "Missing pad_token_id and sliding_window behavior in different models are not the same defect." }, { "left": "issue:43761", "right": "issue:44242", "accept": false, - "reason": "CLIP hidden_states regression and load-balancing loss omission are different model behaviors." + "reason": "Vision hidden_states regression is unrelated to load balancing loss being omitted." }, { "left": "issue:43931", "right": "issue:44060", "accept": false, - "reason": "Qwen3-VL weight-shape mismatch and tied-weights warning in Qwen3-Next are separate loading/configuration issues." + "reason": "Qwen3-VL weight shape mismatch and Qwen3-Next tied-weights warning are different model-loading issues." }, { "left": "issue:44792", "right": "issue:44977", "accept": false, - "reason": "Janus image-generation test failure and Qwen3.5 flash-attention generation bug are unrelated." + "reason": "Janus image-generation test failure is unrelated to Qwen3.5 flash-attention generation issues." }, { "left": "issue:43906", "right": "issue:45290", "accept": false, - "reason": "A reproduction of another issue is not the same as the assistant/chat template tool-call crash." + "reason": "An isolated repro of another issue is not the same as the assistant-tool-call crash in apply_chat_template." }, { "left": "issue:44479", "right": "issue:44734", "accept": false, - "reason": "Video input regression and serve KV-cache continuation indexing are different code paths." + "reason": "Video input regression for Qwen VL models and serve KV-cache continuation indexing are distinct problems." }, { "left": "issue:44464", "right": "issue:45563", "accept": false, - "reason": "Compiled forward inconsistency and a stale generation warning are not the same bug." + "reason": "Chunked generation inconsistency with compiled forward is unrelated to a stale warning on num_return_sequences." } ] }, @@ -9646,7 +9613,7 @@ "component_index": 1, "component_count": 17, "review_unit_index": 31, - "review_unit_count": 50, + "review_unit_count": 49, "status": "reviewed", "reason": null, "source": "llm", @@ -9656,23 +9623,24 @@ "nodes": [ "issue:35141", "issue:36246", - "issue:39692", "issue:41628", "issue:41720", "issue:42490", "issue:43065", - "issue:43295", + "issue:43116", "issue:43335", "issue:43479", "issue:43540", "issue:43550", "issue:43644", + "issue:43650", "issue:43827", "issue:43976", "issue:43994", "issue:44077", "issue:44112", "issue:44190", + "issue:44220", "issue:44462", "issue:44521", "issue:44534", @@ -9711,223 +9679,223 @@ "issue:41720|issue:45245", "issue:44855|issue:45588", "issue:43065|issue:44610", - "issue:35141|issue:42490", "issue:44933|issue:45362", + "issue:35141|issue:42490", "issue:44077|issue:44877", "issue:43550|issue:44190", "issue:43335|issue:44855", "issue:43644|issue:44913", "issue:43976|issue:44855", - "issue:44737|issue:45803", "issue:45127|issue:45357", + "issue:44737|issue:45803", "issue:43540|issue:44734", "issue:44462|issue:45907", - "issue:44112|issue:45600", "issue:44938|issue:45542", + "issue:44112|issue:45600", "issue:43335|issue:44190", "issue:46082|issue:46121", - "issue:39692|issue:44190", + "issue:43116|issue:44190", "issue:45276|issue:45584", "issue:43479|issue:45375", "issue:36246|issue:41628", - "issue:43295|issue:45276" + "issue:43650|issue:44220" ], - "prepared_review_unit_hash": "8fa482c1c576df9ad55ef4d4239a82da53835718dad417126bbd561dfa9bfe9b", + "prepared_review_unit_hash": "e5791d792266579718867401756dce094cc096600cdfcab5a8d5ee5e82ba224b", "component_budget": { "node_count": 426, "item_count": 426, - "soft_pair_count": 1572, - "serialized_chars": 814130, - "estimated_input_tokens": 203533, - "estimated_eval_tokens": 407322 + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 }, "budget": { - "node_count": 47, - "item_count": 47, + "node_count": 48, + "item_count": 48, "soft_pair_count": 28, - "serialized_chars": 35738, - "estimated_input_tokens": 8935, - "estimated_eval_tokens": 18126 + "serialized_chars": 36181, + "estimated_input_tokens": 9046, + "estimated_eval_tokens": 18348 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is highly heterogeneous: it mixes import errors, config/serialization regressions, multimodal processor bugs, docs issues, and several unrelated model-specific defects. None of the soft pairs look like the same underlying bug/change, so I would not merge any of them as duplicates.", - "confidence": 0.88, - "canonical_issue_reason": "No single issue cleanly represents the whole set; the cluster is too broad and the soft matches are only superficial.", - "canonical_pr_reason": "No PRs are present in the cluster.", - "best_issue_reason": "If forced to pick an anchor, issue 35141 is the clearest concrete bug report in the embedding-resize area, but it still does not represent the other unrelated items well.", - "best_pr_reason": "No PRs are present.", + "summary": "This cluster is highly heterogeneous: the soft pairs span unrelated bugs in imports, configs, multimodal models, CI, and docs. I would not merge any of the suggested soft pairs as duplicates. If a single issue must represent the cluster, issue 44521 is the most active and broadly scoped, but the set is not really cohesive.", + "confidence": 0.9, + "canonical_issue_reason": "Issue 44521 is the strongest representative by activity and breadth (open, many discussion entries, multiple inbound refs), though the overall cluster is not semantically unified.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 44521 is the best duplicate-triage anchor among this set because it has the most ongoing discussion and cross-references; the other issues are mostly one-off, narrow reports in unrelated areas.", + "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44190", "right": "issue:44964", "accept": false, - "reason": "Different problems: local dataset loading in a training script vs loading the Phi-4 multimodal model." + "reason": "Local dataset loading in an example script vs a model import failure; different code paths and symptoms." }, { "left": "issue:44534", "right": "issue:44913", "accept": false, - "reason": "Both involve config persistence, but one is a global non-persistent buffer bug and the other is GPTNeoX rotary_pct reload behavior." + "reason": "Both involve configuration/state, but one is about non-persistent buffers being filled with junk and the other about a config field not surviving reload." }, { "left": "issue:43827", "right": "issue:43994", "accept": false, - "reason": "Both are related to broken examples, but one is a pipeline-docs reference issue and the other is a model/pipeline correctness problem for SigLIP2." + "reason": "Docs still mentioning pipeline() vs incorrect model outputs from AutoModel/pipeline; unrelated bug classes." }, { "left": "issue:44792", "right": "issue:45600", "accept": false, - "reason": "Different bugs: janus image-generation test failure vs stale auto_mappings references to removed configs." + "reason": "Janus image-generation test failure vs auto_mappings referencing removed Sam3LiteText configs; different models and failures." }, { "left": "issue:44521", "right": "issue:45381", "accept": false, - "reason": "Both are multimodal, but one is assistant mask generation and the other is Qwen2.5-VL video position IDs." + "reason": "Both are multimodal, but one is all-zero assistant masks in chat templating and the other is wrong vision_position_ids for Qwen2.5-VL video input." }, { "left": "issue:41720", "right": "issue:45245", "accept": false, - "reason": "Unrelated: Qwen3 auto device mapping cuda assert vs category-count overflow at 2^24." + "reason": "CUDA assert during Qwen3 auto device mapping is unrelated to the category-count limit error." }, { "left": "issue:44855", "right": "issue:45588", "accept": false, - "reason": "Python 3.13 import parsing issue vs flash_attention crash on s_aux=None; different code paths and failure modes." + "reason": "Python 3.13 import parse failure in DebertaV2 vs flash_attention crash on s_aux=None; distinct root causes." }, { "left": "issue:43065", "right": "issue:44610", "accept": false, - "reason": "Both are model/processor related, but one is a dummy Conv2d implementation issue and the other is a processor input-size mismatch." + "reason": "Dummy Conv2d in Sam3PixelDecoder vs processor output size mismatch in OmDet-Turbo; different subsystems." }, { - "left": "issue:35141", - "right": "issue:42490", + "left": "issue:44933", + "right": "issue:45362", "accept": false, - "reason": "Embedding resize/post-init bug vs path-trailing-slash import warning; no substantive overlap." + "reason": "Missing import from image_utils vs Qwen3.5 chat crash; unrelated issues." }, { - "left": "issue:44933", - "right": "issue:45362", + "left": "issue:35141", + "right": "issue:42490", "accept": false, - "reason": "Different import regressions in different APIs: image_utils vs transformers chat/model loading." + "reason": "Embedding reinit after resize_token_embeddings vs import warning for paths ending with a slash; no shared code path." }, { "left": "issue:44077", "right": "issue:44877", "accept": false, - "reason": "Both mention config handling, but one is a model-specific optional post_init rule and the other is strict config blocking granite_speech loading." + "reason": "Optional post_init support vs strict config loading for granite_speech; related only at a very broad config level." }, { "left": "issue:43550", "right": "issue:44190", "accept": false, - "reason": "Torch.compile/SDPA failure in Bamba vs dataset loading in a training script; unrelated bugs." + "reason": "torch.compile/SDPA failure in Bamba vs local dataset loading in an example script; unrelated." }, { "left": "issue:43335", "right": "issue:44855", "accept": false, - "reason": "SwitchTransformers sparse-layer construction bug vs Python 3.13 indentation/script parsing bug." + "reason": "SwitchTransformers sparse-layer construction bug vs Python 3.13 indentation/import error; different areas entirely." }, { "left": "issue:43644", "right": "issue:44913", "accept": false, - "reason": "Both are serialization/reload regressions, but one is about non-persistent buffers and the other is a GPTNeoX config field." + "reason": "Non-persistent buffers initialized with junk vs rotary_pct not persisting on reload; both serialization-adjacent but not the same bug." }, { "left": "issue:43976", "right": "issue:44855", "accept": false, - "reason": "Python version compatibility vs JIT comment parsing; different root causes." + "reason": "General Python-version compatibility issue vs a specific decorator/comment parsing failure; not the same underlying defect." }, { - "left": "issue:44737", - "right": "issue:45803", + "left": "issue:45127", + "right": "issue:45357", "accept": false, - "reason": "XLNet device placement bug vs bare except swallowing interrupts/errors in FuyuBatchFeature." + "reason": "LoRA merge collapse with extended vocab vs incorrect visual encoder keys on save_pretrained; both Qwen-related but different code paths." }, { - "left": "issue:45127", - "right": "issue:45357", + "left": "issue:44737", + "right": "issue:45803", "accept": false, - "reason": "Both involve Qwen/embeddings-adjacent wording, but one is LoRA + extended vocabulary collapse and the other is saving visual encoder keys." + "reason": "XLNet device placement bug vs bare except swallowing errors in FuyuBatchFeature; unrelated." }, { "left": "issue:43540", "right": "issue:44734", "accept": false, - "reason": "Qwen3OmniMoe video-input validation vs serve KV-cache tensor indexing crash; not the same bug." + "reason": "Qwen3OmniMoe video-input ValueError vs transformers serve KV-cache continuation crash; different execution paths." }, { "left": "issue:44462", "right": "issue:45907", "accept": false, - "reason": "AutoTokenizer repo-file selection vs list_repo_files not forwarding revision; different API layers and failures." + "reason": "AutoTokenizer ignoring tokenizer.json vs list_repo_files failing to forward revision; different Hub API behaviors." }, { - "left": "issue:44112", - "right": "issue:45600", + "left": "issue:44938", + "right": "issue:45542", "accept": false, - "reason": "CI device-override test flake vs stale auto_mappings references; unrelated maintenance issues." + "reason": "Python 3.14 loading failure vs TensorBoard/TensorFlow backend mismatch; environment-related but distinct." }, { - "left": "issue:44938", - "right": "issue:45542", + "left": "issue:44112", + "right": "issue:45600", "accept": false, - "reason": "Python 3.14 loading failure vs tensorboard-without-TensorFlow backend error; different compatibility issues." + "reason": "GraniteSpeech CI device override flake vs Sam3LiteText auto-mapping breakage; no shared bug." }, { "left": "issue:43335", "right": "issue:44190", "accept": false, - "reason": "SwitchTransformers config bug vs dataset loading failure; no shared underlying defect." + "reason": "SwitchTransformers config bug vs example script dataset loading issue; unrelated." }, { "left": "issue:46082", "right": "issue:46121", "accept": false, - "reason": "LlamaConfig head_dim validation vs rope parameter conversion on JSON-loaded list; unrelated config-validation problems." + "reason": "Llama head_dim validation vs rope param validation on JSON-loaded lists; both config bugs, but different validation logic and symptoms." }, { - "left": "issue:39692", + "left": "issue:43116", "right": "issue:44190", "accept": false, - "reason": "SigLIP2 docs/model mismatch vs dataset loading script failure; not the same bug." + "reason": "Multi-label classification example returns empty results vs local dataset loading failure; separate example-script bugs." }, { "left": "issue:45276", "right": "issue:45584", "accept": false, - "reason": "Embedding resize/output-embedding behavior vs Whisper empty-transcription generation failure; different models and code paths." + "reason": "Gemma4 resize_token_embeddings not updating outputs vs Whisper empty-transcription generation failure; unrelated." }, { "left": "issue:43479", "right": "issue:45375", "accept": false, - "reason": "Default multimodal config initialization vs strict-dropping a new vision field; related area but distinct defects." + "reason": "Phi4MultimodalConfig defaulting None fields vs missing deepstack_visual_indexes in Qwen3_5MoeVisionConfig; both config issues, but not the same one." }, { "left": "issue:36246", "right": "issue:41628", "accept": false, - "reason": "Both are import errors, but for different symbols and different modules." + "reason": "Two different import errors for different symbols; same broad symptom, but not the same missing export." }, { - "left": "issue:43295", - "right": "issue:45276", + "left": "issue:43650", + "right": "issue:44220", "accept": false, - "reason": "Processor.tokenizer regression vs Gemma4 resize_token_embeddings/output_embeddings behavior; distinct changes and failure modes." + "reason": "A non-informative placeholder issue vs an audio feature extraction bug; no duplicate evidence." } ] }, @@ -9943,7 +9911,7 @@ "component_index": 1, "component_count": 17, "review_unit_index": 32, - "review_unit_count": 50, + "review_unit_count": 49, "status": "reviewed", "reason": null, "source": "llm", @@ -9960,13 +9928,13 @@ "issue:42907", "issue:43065", "issue:43295", + "issue:43299", "issue:43335", "issue:43404", "issue:43441", "issue:43526", "issue:43531", "issue:43643", - "issue:43650", "issue:43673", "issue:43824", "issue:43854", @@ -9974,14 +9942,12 @@ "issue:43976", "issue:44016", "issue:44062", - "issue:44075", "issue:44112", "issue:44186", "issue:44220", "issue:44246", "issue:44279", "issue:44336", - "issue:44368", "issue:44496", "issue:44561", "issue:44683", @@ -9993,6 +9959,7 @@ "issue:45042", "issue:45216", "issue:45245", + "issue:45276", "issue:45335", "issue:45446", "issue:45584", @@ -10001,11 +9968,11 @@ "issue:45632" ], "soft_pairs": [ - "issue:43650|issue:44220", + "issue:43295|issue:45276", "issue:45245|issue:45584", "issue:41950|issue:44016", - "issue:36246|issue:38617", "issue:42175|issue:44279", + "issue:36246|issue:38617", "issue:44186|issue:44734", "issue:44683|issue:45446", "issue:44246|issue:45003", @@ -10016,8 +9983,8 @@ "issue:43643|issue:44496", "issue:44964|issue:45600", "issue:44561|issue:45632", - "issue:44112|issue:44964", "issue:43854|issue:44964", + "issue:44112|issue:44964", "issue:43976|issue:44987", "issue:42907|issue:43441", "issue:42490|issue:45632", @@ -10025,214 +9992,214 @@ "issue:43295|issue:45335", "issue:43526|issue:43901", "issue:44496|issue:45042", - "issue:43673|issue:44336", "issue:44016|issue:44062", - "issue:43335|issue:45593", + "issue:43673|issue:44336", "issue:43824|issue:44933", - "issue:44075|issue:44368" + "issue:43335|issue:45593", + "issue:43299|issue:45216" ], - "prepared_review_unit_hash": "8b724b2069f26b37f83fedd88075c10413ecaa79d228ec3f0197f270dc490454", + "prepared_review_unit_hash": "b2be04d4bdff9f6e54de0ce8086bf879923d9db3ae803b3008e0d41c466f979f", "component_budget": { "node_count": 426, "item_count": 426, - "soft_pair_count": 1572, - "serialized_chars": 814130, - "estimated_input_tokens": 203533, - "estimated_eval_tokens": 407322 + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 }, "budget": { - "node_count": 48, - "item_count": 48, + "node_count": 47, + "item_count": 47, "soft_pair_count": 29, - "serialized_chars": 36325, - "estimated_input_tokens": 9082, - "estimated_eval_tokens": 18420 + "serialized_chars": 35785, + "estimated_input_tokens": 8947, + "estimated_eval_tokens": 18150 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The cluster is heterogeneous: it mixes unrelated import errors, model-loading regressions, pipeline bugs, and a few flex_attention/trust_remote_code/dependency issues. No soft pair looks like the same underlying bug or change, so none should be merged as duplicates.", - "confidence": 0.21, - "canonical_issue_reason": "No true canonical duplicate emerges from this set; if one issue must anchor the cluster, issue:43824 is the strongest broad user-facing import failure with the most discussion/inbound references.", + "summary": "This set is mostly heterogeneous: import regressions, model-specific loading bugs, tokenizer/pipeline issues, and unrelated docs/CI reports. None of the soft pairs look like true duplicates; the closest themes still fail the strict same-code-path test. The clearest representative issue is the Qwen2.5-VL import regression.", + "confidence": 0.93, + "canonical_issue_reason": "Issue 43824 is the most concrete, well-scoped representative: a specific top-level import regression with clear affected symbol and inbound references.", "canonical_pr_reason": null, - "best_issue_reason": "Issue:43824 is the best representative because it is a high-signal, user-facing top-level import regression with broader impact than the more model-specific reports.", + "best_issue_reason": "Issue 43824 is the best single issue to anchor this cluster because it is explicit, reproducible, and broadly recognizable as a Transformers export/import regression.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:43650", - "right": "issue:44220", + "left": "issue:43295", + "right": "issue:45276", "accept": false, - "reason": "Unrelated: a low-quality title/body issue vs a specific `_torch_extract_fbank_features()` failure." + "reason": "Both mention resize/tokenization-related regressions, but they affect different model stacks and different code paths." }, { "left": "issue:45245", "right": "issue:45584", "accept": false, - "reason": "Different bugs: category cardinality limit vs Whisper empty-transcription generation after token alignment." + "reason": "One is a category-cardinality runtime error; the other is a Whisper generation failure after token alignment. Not the same bug." }, { "left": "issue:41950", "right": "issue:44016", "accept": false, - "reason": "One is a pipeline processor mismatch, the other is a notebook syntax error." + "reason": "A pipeline processor lookup bug and a notebook syntax error are unrelated." }, { - "left": "issue:36246", - "right": "issue:38617", + "left": "issue:42175", + "right": "issue:44279", "accept": false, - "reason": "Both are ImportErrors, but they target different missing symbols in different modules and code paths." + "reason": "Dependency/backend packaging trouble vs a generic dependency issue; too broad and not the same concrete failure." }, { - "left": "issue:42175", - "right": "issue:44279", + "left": "issue:36246", + "right": "issue:38617", "accept": false, - "reason": "General dependency/install backend issue versus an unspecified dependency problem; not the same concrete failure." + "reason": "Both are ImportErrors, but for different missing symbols in different modules and different fixes." }, { "left": "issue:44186", "right": "issue:44734", "accept": false, - "reason": "Tokenizer padding/truncation crash is unrelated to serve/KV-cache continuation indexing." + "reason": "Tokenizer padding/NER crashes and serve KV-cache continuation indexing are unrelated code paths." }, { "left": "issue:44683", "right": "issue:45446", "accept": false, - "reason": "Both mention flex_attention, but one is a torch>=2.9 compiled-path failure and the other is a version-check/import guard bug." + "reason": "Both involve flex_attention, but one is a torch-2.9 compiled failure and the other is a version-check/import guard issue." }, { "left": "issue:44246", "right": "issue:45003", "accept": false, - "reason": "Import latency and unsafe sys.modules access are related at a high level, but not the same bug or fix." + "reason": "Import slowness and unsafe sys.modules access are adjacent, but not clearly the same defect." }, { "left": "issue:44016", "right": "issue:44220", "accept": false, - "reason": "Notebook syntax error and audio feature extraction failure are unrelated." + "reason": "Notebook syntax error vs audio feature extraction issue; unrelated." }, { "left": "issue:43404", "right": "issue:45216", "accept": false, - "reason": "Different models and different symptoms: untied lm_head weights vs incorrect saved checkpoint behavior." + "reason": "Different model families and different bugs: untied weights vs incorrect save_pretrained output." }, { "left": "issue:38175", "right": "issue:43065", "accept": false, - "reason": "Zero probabilities in SigLIP2 is unrelated to a dummy Conv2d in Sam3PixelDecoder." + "reason": "A probability anomaly in SIGLIP2 is unrelated to a dummy Conv2d in Sam3PixelDecoder." }, { "left": "issue:44734", "right": "issue:45632", "accept": false, - "reason": "Serve/KV-cache indexing and trust_remote_code cache-path collisions are different code paths." + "reason": "Serve KV-cache continuation bug and local trust_remote_code cache collision are different subsystems." }, { "left": "issue:43643", "right": "issue:44496", "accept": false, - "reason": "Missing fields from trust_remote_code config loading is not the same as unrecognized model_type in config.json." + "reason": "One is missing fields with trust_remote_code; the other is an unrecognized model lacking model_type. Similar category, different bug." }, { "left": "issue:44964", "right": "issue:45600", "accept": false, - "reason": "Model loading failure for Phi-4 multimodal is unrelated to stale auto_mappings references for Sam3LiteText." + "reason": "Both concern model loading, but one is a Phi-4 multimodal load failure and the other is stale auto_mappings references." }, { "left": "issue:44561", "right": "issue:45632", "accept": false, - "reason": "Removal of `is_torch_fx_available` affecting trust_remote_code models is a different failure than cache-path collisions." + "reason": "Both mention trust_remote_code, but one is a removed API break and the other is a cache-path collision." }, { - "left": "issue:44112", + "left": "issue:43854", "right": "issue:44964", "accept": false, - "reason": "A CI test device-override failure is not the same as a model loading regression." + "reason": "Two different model-loading failures for different models; no shared concrete code path." }, { - "left": "issue:43854", + "left": "issue:44112", "right": "issue:44964", "accept": false, - "reason": "Loading GLM-4.7-Flash in tests is a separate model-specific issue from Phi-4 multimodal loading." + "reason": "A CI test flake in GraniteSpeech is unrelated to a multimodal model loading failure." }, { "left": "issue:43976", "right": "issue:44987", "accept": false, - "reason": "Python version compatibility and loading a specific physical-intelligence model are distinct problems." + "reason": "Python compatibility and a model-load regression are not the same underlying issue." }, { "left": "issue:42907", "right": "issue:43441", "accept": false, - "reason": "Saving dequantized Ministral/Devstral models is unrelated to FlashAttention failures in Ministral-3." + "reason": "Both involve Ministral, but one is save/dequantization and the other is FlashAttention loading; different fixes." }, { "left": "issue:42490", "right": "issue:45632", "accept": false, - "reason": "Trailing-slash local path warnings and trust_remote_code cache collisions are different path-resolution bugs." + "reason": "Both involve local paths, but one is a warning on trailing slash and the other is a leaf-directory cache collision." }, { "left": "issue:43531", "right": "issue:45600", "accept": false, - "reason": "Qwen3-MoE sliding_window behavior is unrelated to auto_mappings referring to removed configs." + "reason": "Qwen3-MoE sliding-window behavior and Sam3LiteText auto-mapping cleanup are unrelated." }, { "left": "issue:43295", "right": "issue:45335", "accept": false, - "reason": "Processor/tokenizer regression and t5gemma embedding resize bug are different model-code issues." + "reason": "Both mention regression around embeddings/tokenizer APIs, but the affected models and failure modes differ." }, { "left": "issue:43526", "right": "issue:43901", "accept": false, - "reason": "Image processor label reduction bug is unrelated to documentation wording for text classification pipeline behavior." + "reason": "An image-processor label bug and a documentation mismatch in text classification are unrelated." }, { "left": "issue:44496", "right": "issue:45042", "accept": false, - "reason": "Unrecognized model_type in config loading is not the same as PIL backend processors incorrectly requiring torchvision." - }, - { - "left": "issue:43673", - "right": "issue:44336", - "accept": false, - "reason": "Generation cache regression and ANSI-code logging behavior have no shared bug path." + "reason": "Missing model_type in config and PIL image processor torchvision requirements are different load-time problems." }, { "left": "issue:44016", "right": "issue:44062", "accept": false, - "reason": "Notebook syntax error and AddedToken constructor argument collision are unrelated." + "reason": "Notebook syntax error and AddedToken argument collision are unrelated." }, { - "left": "issue:43335", - "right": "issue:45593", + "left": "issue:43673", + "right": "issue:44336", "accept": false, - "reason": "Sparse-layer creation in SwitchTransformersConfig is not the same as D-FINE auxiliary-loss suppression." + "reason": "Generation cache behavior and ANSI log formatting are entirely separate concerns." }, { "left": "issue:43824", "right": "issue:44933", "accept": false, - "reason": "Qwen2.5-VL export/import failure and a missing image_utils import are different missing-symbol issues." + "reason": "Both are import-related, but one is a missing top-level export and the other is a nonexistent image_utils import; not the same bug." }, { - "left": "issue:44075", - "right": "issue:44368", + "left": "issue:43335", + "right": "issue:45593", + "accept": false, + "reason": "SwitchTransformers config construction and D-FINE auxiliary-loss behavior are unrelated model-specific issues." + }, + { + "left": "issue:43299", + "right": "issue:45216", "accept": false, - "reason": "Unused SGD args and tie_word_embeddings warning are separate configuration warnings/bugs." + "reason": "Qwen3VL loading regression and Qwen3.5 save_pretrained regression are different failure modes." } ] }, @@ -10248,7 +10215,7 @@ "component_index": 1, "component_count": 17, "review_unit_index": 33, - "review_unit_count": 50, + "review_unit_count": 49, "status": "reviewed", "reason": null, "source": "llm", @@ -10272,11 +10239,13 @@ "issue:43901", "issue:43976", "issue:44016", + "issue:44075", "issue:44242", "issue:44279", "issue:44291", "issue:44297", "issue:44351", + "issue:44368", "issue:44462", "issue:44496", "issue:44521", @@ -10295,8 +10264,6 @@ "issue:45542", "issue:45584", "issue:45593", - "issue:45600", - "issue:45712", "issue:45759", "issue:45812", "issue:45823", @@ -10305,20 +10272,20 @@ "issue:46132" ], "soft_pairs": [ - "issue:43299|issue:45216", + "issue:44075|issue:44368", "issue:43976|issue:44279", "issue:43381|issue:45593", "issue:43901|issue:44496", - "issue:45335|issue:45584", "issue:44496|issue:45759", + "issue:45335|issue:45584", "issue:43299|issue:45357", "issue:43479|issue:43525", "issue:44521|issue:45325", "issue:44242|issue:45200", "issue:44351|issue:45278", "issue:41950|issue:45020", - "issue:43577|issue:45216", "issue:41950|issue:45823", + "issue:43577|issue:45216", "issue:44779|issue:45812", "issue:44683|issue:45542", "issue:43976|issue:45405", @@ -10330,227 +10297,220 @@ "issue:43650|issue:46132", "issue:44462|issue:45081", "issue:44297|issue:44462", - "issue:43450|issue:44279", "issue:45081|issue:45907", + "issue:43450|issue:44279", "issue:44016|issue:44291", "issue:41950|issue:43650", - "issue:43441|issue:45834", - "issue:45600|issue:45712" + "issue:43441|issue:45834" ], - "prepared_review_unit_hash": "c37f1c7711576ee751b8f05e52edaa472b0a078746dbd57fdb491d0c77b29192", + "prepared_review_unit_hash": "2a7d4e526d59f7d5f90c48ddc3ba6db3d68d076605ce6c32347da593e2366f2c", "component_budget": { "node_count": 426, "item_count": 426, - "soft_pair_count": 1572, - "serialized_chars": 814130, - "estimated_input_tokens": 203533, - "estimated_eval_tokens": 407322 + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 }, "budget": { "node_count": 47, "item_count": 47, - "soft_pair_count": 31, - "serialized_chars": 36448, - "estimated_input_tokens": 9112, - "estimated_eval_tokens": 18480 + "soft_pair_count": 30, + "serialized_chars": 36151, + "estimated_input_tokens": 9038, + "estimated_eval_tokens": 18332 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These items are mostly unrelated bug reports that only share broad transformer/tokenizer/model-loading themes. None of the soft pairs looks like the same underlying bug or a mergeable PR-equivalent change, so I would not accept any duplicate edges.", - "confidence": 0.91, - "canonical_issue_reason": "Issue 44462 is the clearest representative if a single issue must anchor the cluster: it is concrete, user-facing, and about a core tokenizer-loading path; the other items are too heterogeneous to collapse into it as duplicates.", + "summary": "No convincing duplicate cluster here: the soft pairs mostly connect unrelated bugs across different models, pipelines, tokenizers, and docs. I would not merge any of the candidate pairs.", + "confidence": 0.94, + "canonical_issue_reason": "issue:44521 is the strongest representative item in this set: it is active, has multiple inbound references, and describes a concrete reproducible multimodal bug.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 44462 is the best overall representative for this set because it is the most actionable and centrally scoped regression, but the cluster is not a true duplicate group.", + "best_issue_reason": "issue:44521 is the best global issue candidate because it has the clearest ongoing impact and discussion footprint among otherwise unrelated reports.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:43299", - "right": "issue:45216", + "left": "issue:44075", + "right": "issue:44368", "accept": false, - "reason": "Related Qwen3.5/Qwen3VL area, but one is a load-time model-init failure and the other is incorrect save_pretrained serialization; different bugs and code paths." + "reason": "Different problems: optimizer argument handling vs a tie_word_embeddings warning in a Qwen fine-tuning workflow." }, { "left": "issue:43976", "right": "issue:44279", "accept": false, - "reason": "One is a Python-version compatibility complaint, the other a vague dependency problem; too broad and not the same defect." + "reason": "One is a Python/version compatibility report; the other is a vague dependency issue. No shared code-path bug." }, { "left": "issue:43381", "right": "issue:45593", "accept": false, - "reason": "Gradient checkpointing in eval mode and D-FINE auxiliary-loss behavior are unrelated model-training issues." + "reason": "Gradient checkpointing in eval mode is unrelated to D-FINE auxiliary-loss behavior when denoising is off." }, { "left": "issue:43901", "right": "issue:44496", "accept": false, - "reason": "Docs/behavior mismatch for text classification vs unrecognized model/config loading failure; different subsystems and symptoms." + "reason": "Docs mismatch for text classification is unrelated to model loading failing for a specific repository." }, { - "left": "issue:45335", - "right": "issue:45584", + "left": "issue:44496", + "right": "issue:45759", "accept": false, - "reason": "Resize-token-embeddings bug in t5gemma is unrelated to Whisper empty-transcription generation after align_special_tokens." + "reason": "Both involve model loading, but different mechanisms and configs: unrecognized model vs unwrapping text_config for multimodal Qwen configs." }, { - "left": "issue:44496", - "right": "issue:45759", + "left": "issue:45335", + "right": "issue:45584", "accept": false, - "reason": "Both concern model loading, but one is missing model_type recognition and the other is composite-config unwrapping for AutoModelForCausalLM; not the same bug." + "reason": "Unrelated bugs: embedding resize propagation in t5gemma vs Whisper empty-transcription generation after token alignment." }, { "left": "issue:43299", "right": "issue:45357", "accept": false, - "reason": "Same family, but one is a load failure and the other is a save_pretrained key regression; distinct concrete code paths." + "reason": "Both are Qwen-related but distinct regressions affecting loading vs save_pretrained output structure." }, { "left": "issue:43479", "right": "issue:43525", "accept": false, - "reason": "Phi4Multimodal default config initialization and Llama4Config missing pad_token_id are unrelated config issues." + "reason": "Different config construction bugs: missing default multimodal subconfigs vs missing pad_token_id on Llama4Config." }, { "left": "issue:44521", "right": "issue:45325", "accept": false, - "reason": "apply_chat_template assistant-mask bug is unrelated to Qwen2.5-VL rope-index temporal scaling." + "reason": "Different multimodal chat-template/mask bug vs a Qwen2.5-VL rope index scaling issue." }, { "left": "issue:44242", "right": "issue:45200", "accept": false, - "reason": "MoE load-balancing loss gating and Gemma 4 mm_token_type_ids defaults are different model-training semantics." + "reason": "Load-balancing loss omission is a different training-loss bug from Gemma 4 mm_token_type_ids defaults." }, { "left": "issue:44351", "right": "issue:45278", "accept": false, - "reason": "Specific missing HybridCache import versus broad import breakage after upgrade; too different to be duplicates." + "reason": "Importing HybridCache is a specific symbol regression; the other is a broad import-error report after upgrade." }, { "left": "issue:41950", "right": "issue:45020", "accept": false, - "reason": "Video-classification processor lookup issue is unrelated to remote_code regressions across recent versions." + "reason": "Video-classification processor lookup is unrelated to remote_code breakage." }, { - "left": "issue:43577", - "right": "issue:45216", + "left": "issue:41950", + "right": "issue:45823", "accept": false, - "reason": "BLIP2 dtype propagation and Qwen3.5 save_pretrained serialization are different bugs." + "reason": "Different Gemma/PLE device-placement bug vs video-classification processor selection bug." }, { - "left": "issue:41950", - "right": "issue:45823", + "left": "issue:43577", + "right": "issue:45216", "accept": false, - "reason": "One is pipeline processor selection, the other a Gemma4 PLE device-mapping mismatch; not the same underlying problem." + "reason": "Blip2 dtype propagation and Qwen3.5 save_pretrained corruption are separate code paths and regressions." }, { "left": "issue:44779", "right": "issue:45812", "accept": false, - "reason": "Both are tokenizer regressions, but they affect different model families and likely different tokenizer code paths." + "reason": "Both are tokenizer regressions, but for different model families and likely different failure modes; too broad to merge." }, { "left": "issue:44683", "right": "issue:45542", "accept": false, - "reason": "Compiled flex_attention on torch>=2.9 and tensorboard/TF backend import handling are unrelated." + "reason": "Compiled flex_attention on torch 2.9 is unrelated to a TensorFlow backend installation error." }, { "left": "issue:43976", "right": "issue:45405", "accept": false, - "reason": "Python support mismatch and MIN_PEFT_VERSION bump are separate dependency/version issues." + "reason": "Python compatibility issue vs an unmet PEFT version requirement; not the same bug." }, { "left": "issue:38175", "right": "issue:44617", "accept": false, - "reason": "Unexpected zero probabilities in SigLIP2 is unrelated to Sam3Video CUDA OOM." + "reason": "Unexpected zero probabilities in SigLIP2 is unrelated to a Sam3Video CUDA OOM report." }, { "left": "issue:44279", "right": "issue:45405", "accept": false, - "reason": "Generic dependency issue versus a specific minimum PEFT version bump; not the same bug." + "reason": "Generic dependency issue and a specific minimum-version bump problem are not the same underlying change." }, { "left": "issue:39692", "right": "issue:43901", "accept": false, - "reason": "SigLIP2 doc example problems and TextClassificationPipeline docs wording are different documentation issues." + "reason": "Both touch docs, but one is a SigLIP2 example with model/processor and quantization errors, the other is a text-classification API documentation mismatch." }, { "left": "issue:43299", "right": "issue:43854", "accept": false, - "reason": "Qwen3VL loading failure and GLM-4.7-Flash unit-test loading failure are separate model-specific regressions." + "reason": "Qwen3VL loading regression and GLM-4.7-Flash test loading failure are unrelated model-specific issues." }, { "left": "issue:43295", "right": "issue:44779", "accept": false, - "reason": "Processor.tokenizer regression and DeepSeek tokenizer output regression are both tokenizer-related, but they are not the same concrete failure." + "reason": "Processor/tokenizer API regression for custom code is unrelated to DeepSeek tokenizer output changes." }, { "left": "issue:43650", "right": "issue:46132", "accept": false, - "reason": "ADD THE DATA is not comparable to AttentionInterface.register changing behavior; no shared bug." + "reason": "Placeholder issue title and an AttentionInterface.register behavior change are clearly unrelated." }, { "left": "issue:44462", "right": "issue:45081", "accept": false, - "reason": "AutoTokenizer ignoring tokenizer.json and Mistral regex patch crashing are both tokenizer-loading issues, but they arise in different mechanisms." + "reason": "AutoTokenizer ignoring tokenizer.json and a Mistral regex patch crash are different tokenizer-loading failures." }, { "left": "issue:44297", "right": "issue:44462", "accept": false, - "reason": "Tokenizer_class mismatch on save_pretrained and ignoring tokenizer.json on load are related, but they are different serialization vs loading defects." + "reason": "Tokenizer save_pretrained metadata mismatch is a different side of the pipeline than AutoTokenizer ignoring tokenizer.json; not the same concrete bug." }, { - "left": "issue:43450", - "right": "issue:44279", + "left": "issue:45081", + "right": "issue:45907", "accept": false, - "reason": "Video processor batch-shape bug and generic dependency issue are unrelated." + "reason": "Tokenizer regex patching crash vs list_repo_files revision forwarding are unrelated." }, { - "left": "issue:45081", - "right": "issue:45907", + "left": "issue:43450", + "right": "issue:44279", "accept": false, - "reason": "Tokenizer regex crash and list_repo_files revision forwarding are completely different concerns." + "reason": "Video processor batch shape bug is unrelated to a generic dependency issue." }, { "left": "issue:44016", "right": "issue:44291", "accept": false, - "reason": "Notebook syntax error and init_empty_weights unexpected-argument TypeError are unrelated." + "reason": "Notebook syntax error and init_empty_weights argument compatibility are unrelated." }, { "left": "issue:41950", "right": "issue:43650", "accept": false, - "reason": "No overlap beyond being issue reports; these are unrelated." + "reason": "Video-classification processor lookup bug is unrelated to the placeholder data request issue." }, { "left": "issue:43441", "right": "issue:45834", "accept": false, - "reason": "FlashAttention regression for Ministral-3 and Kosmos2.5 OCR index error are different model-specific bugs." - }, - { - "left": "issue:45600", - "right": "issue:45712", - "accept": false, - "reason": "Stale auto_mappings references and leftover dummy classes are both repo-maintenance issues, but not the same defect." + "reason": "Ministral FlashAttention failure and Kosmos2.5 OCR index error are different model-specific runtime bugs." } ] }, @@ -10566,7 +10526,7 @@ "component_index": 1, "component_count": 17, "review_unit_index": 34, - "review_unit_count": 50, + "review_unit_count": 49, "status": "reviewed", "reason": null, "source": "llm", @@ -10605,7 +10565,6 @@ "issue:44361", "issue:44466", "issue:44521", - "issue:44589", "issue:44617", "issue:44623", "issue:44671", @@ -10616,29 +10575,31 @@ "issue:44987", "issue:45276", "issue:45290", - "issue:45362", "issue:45397", "issue:45542", "issue:45593", + "issue:45600", "issue:45656", + "issue:45712", "issue:45823" ], "soft_pairs": [ + "issue:45600|issue:45712", "issue:44466|issue:45276", "issue:42994|issue:43317", "issue:33290|issue:45656", - "issue:43653|issue:45593", "issue:43650|issue:45290", + "issue:43653|issue:45593", "issue:44623|issue:44749", "issue:44623|issue:45397", "issue:43065|issue:44898", - "issue:43441|issue:44683", "issue:44261|issue:44263", + "issue:43441|issue:44683", "issue:43723|issue:44162", - "issue:42175|issue:43976", "issue:44016|issue:45823", - "issue:44361|issue:45593", + "issue:42175|issue:43976", "issue:43262|issue:44521", + "issue:44361|issue:45593", "issue:44671|issue:44779", "issue:43901|issue:43994", "issue:43673|issue:43761", @@ -10646,209 +10607,208 @@ "issue:43784|issue:45542", "issue:43824|issue:43976", "issue:43525|issue:43881", - "issue:43450|issue:44617", "issue:44297|issue:45397", - "issue:43901|issue:44246", + "issue:43450|issue:44617", "issue:43525|issue:43643", - "issue:41628|issue:43976", - "issue:44589|issue:45362" + "issue:43901|issue:44246", + "issue:41628|issue:43976" ], - "prepared_review_unit_hash": "2feb5ca02ebcf36c7c360c9cdbc2afc302c49088c48cdbda74d66efbc4afe79d", + "prepared_review_unit_hash": "0edc0dac4594e6e99dece354fb13aaadeb27303ceb9e430b7f1c5a0ff49af12f", "component_budget": { "node_count": 426, "item_count": 426, - "soft_pair_count": 1572, - "serialized_chars": 814130, - "estimated_input_tokens": 203533, - "estimated_eval_tokens": 407322 + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 28, - "serialized_chars": 35773, - "estimated_input_tokens": 8944, - "estimated_eval_tokens": 18144 + "serialized_chars": 35862, + "estimated_input_tokens": 8966, + "estimated_eval_tokens": 18188 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is highly heterogeneous: the soft pairs span unrelated issues across tokenizer loading, multimodal chat templates, config/serialization, attention backends, packaging, and model-specific regressions. I do not see a real duplicate group here, so all soft edges are rejected.", - "confidence": 0.95, - "canonical_issue_reason": "No clear canonical issue: the items are mostly unrelated, so there is no single underlying bug to collapse them into one duplicate set.", + "summary": "These items are a mixed bag of unrelated Transformers issues across loading, tokenization, processors, models, docs, and build/runtime regressions. None of the soft-edge pairs look like the same concrete underlying bug/change.", + "confidence": 0.97, + "canonical_issue_reason": "issue:44521 is the best representative because it is a substantive user-facing bug with multiple references and broad multimodal/chat-template impact, unlike the more narrow or highly specific reports in the set.", "canonical_pr_reason": null, - "best_issue_reason": "If forced to pick a representative, issue 43643 is the broadest core-loading/config regression in the set, but it is still not a true duplicate anchor for the cluster.", + "best_issue_reason": "issue:44521 has the strongest signal for reuse as a canonical issue: clear reproduction, active discussion, and broad downstream relevance.", "best_pr_reason": null, "soft_edge_verdicts": [ + { + "left": "issue:45600", + "right": "issue:45712", + "accept": false, + "reason": "Both are CI/repo hygiene problems, but one is stale auto_mappings references and the other is leftover dummy classes; different files and failure modes." + }, { "left": "issue:44466", "right": "issue:45276", "accept": false, - "reason": "Different problems: tied-weight serialization/device dependence vs resize_token_embeddings not propagating in Gemma4." + "reason": "Both involve Gemma4, but one is serialization/tied-weights behavior while the other is resize_token_embeddings not updating embeddings." }, { "left": "issue:42994", "right": "issue:43317", "accept": false, - "reason": "Quantized model saving failure is unrelated to dequantized model loading with gpu/cpu offload." + "reason": "One is saving quantized models failing; the other is loading dequantized models with device_map/offload. Related area, different bug." }, { "left": "issue:33290", "right": "issue:45656", "accept": false, - "reason": "Both mention deepspeed, but one is Adafactor OOM while the other is double optimizer stepping." + "reason": "Deepspeed is common, but adafactor OOM and optimizer.step being called twice are distinct optimizer/runtime issues." }, { - "left": "issue:43653", - "right": "issue:45593", + "left": "issue:43650", + "right": "issue:45290", "accept": false, - "reason": "Tokenizer special-token registration and D-FINE auxiliary-loss behavior are unrelated code paths." + "reason": "Completely different chat-template failures: one is a vague data-related issue, the other is a specific tool-call/content crash." }, { - "left": "issue:43650", - "right": "issue:45290", + "left": "issue:43653", + "right": "issue:45593", "accept": false, - "reason": "Completely unrelated: a vague data request versus chat-template crashing on empty tool-call assistant messages." + "reason": "BigBirdTokenizer special-token registration and D-FINE auxiliary-loss behavior are unrelated." }, { "left": "issue:44623", "right": "issue:44749", "accept": false, - "reason": "Processor save_pretrained missing files is not the same as a v5 data-filtering slowdown regression." + "reason": "One is missing files in processor.save_pretrained; the other is a performance regression during data filtering." }, { "left": "issue:44623", "right": "issue:45397", "accept": false, - "reason": "Different failure modes: missing processor files versus Gemma-4 Zero3 from_pretrained." + "reason": "Different areas: processor serialization vs Gemma-4 ZeRO-3 from_pretrained loading." }, { "left": "issue:43065", "right": "issue:44898", "accept": false, - "reason": "Sam3PixelDecoder dummy Conv2d and Perceiver interpolation failure are separate model bugs." - }, - { - "left": "issue:43441", - "right": "issue:44683", - "accept": false, - "reason": "FlashAttention failure for Ministral-3 and compiled flex_attention failure on torch>=2.9 are distinct backend issues." + "reason": "Sam3PixelDecoder dummy conv and Perceiver non-default-resolution failure are separate vision-model bugs." }, { "left": "issue:44261", "right": "issue:44263", "accept": false, - "reason": "One is an MLA layernorm config precision issue; the other is a torch.split return-value issue in a different class." + "reason": "MLA rms_norm_eps precision issue and GlmMoeDsaIndexer torch.split return values are unrelated." }, { - "left": "issue:43723", - "right": "issue:44162", + "left": "issue:43441", + "right": "issue:44683", "accept": false, - "reason": "Tokenizer loading regression and ESM2 model breakage are not the same underlying bug." + "reason": "Both mention attention, but one is a model/backend incompatibility and the other is a torch 2.9 compiled flex_attention failure." }, { - "left": "issue:42175", - "right": "issue:43976", + "left": "issue:43723", + "right": "issue:44162", "accept": false, - "reason": "Backend dependency packaging and Python version compatibility are unrelated." + "reason": "Tokenizer loading and ESM2 model breakage are distinct failures." }, { "left": "issue:44016", "right": "issue:45823", "accept": false, - "reason": "Notebook syntax error is unrelated to Gemma4 device-mismatch behavior." + "reason": "Notebook syntax error and Gemma4 device mismatch are unrelated." }, { - "left": "issue:44361", - "right": "issue:45593", + "left": "issue:42175", + "right": "issue:43976", "accept": false, - "reason": "MLukeTokenizer AttributeError and D-FINE auxiliary-loss behavior do not share a code path." + "reason": "One is missing TensorFlow in the torch extra; the other is Python version compatibility for Transformers 5.1.0." }, { "left": "issue:43262", "right": "issue:44521", "accept": false, - "reason": "Both involve apply_chat_template, but audio sampling-rate defaults and multimodal assistant mask generation are different bugs." + "reason": "Both involve apply_chat_template, but audio sampling-rate defaulting and multimodal assistant_masks are different code paths and bugs." + }, + { + "left": "issue:44361", + "right": "issue:45593", + "accept": false, + "reason": "MLukeTokenizer task AttributeError and D-FINE auxiliary-loss omission are unrelated." }, { "left": "issue:44671", "right": "issue:44779", "accept": false, - "reason": "CamemBERT masked-LM predictions and Deepseek tokenizer output regressions are separate issues." + "reason": "Both are v5 regressions with incorrect outputs, but they affect different models and tokenizer/model code paths." }, { "left": "issue:43901", "right": "issue:43994", "accept": false, - "reason": "Documentation mismatch for a pipeline option is not the same as a wrong-output model/pipeline bug." + "reason": "Docs mismatch for TextClassificationPipeline and bad SigLIP2 inference are not the same issue." }, { "left": "issue:43673", "right": "issue:43761", "accept": false, - "reason": "Generation cache/chunked prefill regression is unrelated to CLIPVision hidden_states handling." + "reason": "Generation cache missing in chunked_prefill and CLIPVisionModel hidden_states regression are different model/runtime bugs." }, { "left": "issue:44987", "right": "issue:45542", "accept": false, - "reason": "Model-loading failure for a specific repo is unrelated to a missing TensorFlow backend dependency issue." + "reason": "Loading a specific model failing and TensorFlow backend import issues are unrelated." }, { "left": "issue:43784", "right": "issue:45542", "accept": false, - "reason": "Import-time NameError in sentence-transformers is unrelated to tensorboard/TensorFlow backend setup." + "reason": "Sentence-transformers import NameError and undefined tf backend from missing TensorFlow are different import problems." }, { "left": "issue:43824", "right": "issue:43976", "accept": false, - "reason": "Missing Qwen2_5_VL export and Python version support are different problems." + "reason": "Missing Qwen2.5-VL export and Python compatibility regression do not share the same root cause." }, { "left": "issue:43525", "right": "issue:43881", "accept": false, - "reason": "Llama4Config missing pad_token_id and glm-4v-9b loading failure are distinct model/config issues." - }, - { - "left": "issue:43450", - "right": "issue:44617", - "accept": false, - "reason": "Batched video processor shape bug is not the same as Sam3Video CUDA OOM." + "reason": "Llama4Config missing pad_token_id and glm-4v-9b loading failure are separate model-loading issues." }, { "left": "issue:44297", "right": "issue:45397", "accept": false, - "reason": "Tokenizer save metadata mismatch and Gemma-4 Zero3 loading are unrelated." + "reason": "Tokenizer save_pretrained metadata mismatch and Gemma-4 ZeRO-3 loading are unrelated." }, { - "left": "issue:43901", - "right": "issue:44246", + "left": "issue:43450", + "right": "issue:44617", "accept": false, - "reason": "A docs mismatch does not duplicate an intermittent slow import bug." + "reason": "Batched video processor output shape and Sam3Video CUDA OOM are different problems." }, { "left": "issue:43525", "right": "issue:43643", "accept": false, - "reason": "These are both config-loading related, but one is a specific pad_token_id absence and the other is a broader trust_remote_code field-loss bug; not enough to treat as the same issue." + "reason": "Both involve config fields, but one is a specific Llama4Config attribute omission and the other is a generic trust_remote_code config hydration issue." }, { - "left": "issue:41628", - "right": "issue:43976", + "left": "issue:43901", + "right": "issue:44246", "accept": false, - "reason": "Cannot import AutoImageProcessor and Python 3.9/3.10 compatibility are unrelated." + "reason": "Docs/behavior mismatch in a pipeline and intermittent slow imports are unrelated." }, { - "left": "issue:44589", - "right": "issue:45362", + "left": "issue:41628", + "right": "issue:43976", "accept": false, - "reason": "Float8 storage deserialization error and Qwen3.5 chat crash are different failures." + "reason": "AutoImageProcessor import failure and Python version compatibility regression are not the same bug." } ] }, @@ -10864,7 +10824,7 @@ "component_index": 1, "component_count": 17, "review_unit_index": 35, - "review_unit_count": 50, + "review_unit_count": 49, "status": "reviewed", "reason": null, "source": "llm", @@ -10897,6 +10857,7 @@ "issue:44485", "issue:44496", "issue:44561", + "issue:44589", "issue:44749", "issue:44779", "issue:44857", @@ -10921,6 +10882,7 @@ "issue:46132" ], "soft_pairs": [ + "issue:44589|issue:45362", "issue:44188|issue:44336", "issue:45561|issue:45563", "issue:42491|issue:43299", @@ -10934,8 +10896,8 @@ "issue:43526|issue:45003", "issue:36246|issue:42757", "issue:36246|issue:44351", - "issue:45003|issue:45561", "issue:44336|issue:44485", + "issue:45003|issue:45561", "issue:44246|issue:44871", "issue:43352|issue:45250", "issue:43673|issue:44561", @@ -10947,94 +10909,100 @@ "issue:43901|issue:45003", "issue:45412|issue:46123", "issue:44496|issue:45003", - "issue:45292|issue:45584", "issue:44877|issue:45375", + "issue:45292|issue:45584", "issue:45987|issue:46082", "issue:43262|issue:44246", "issue:45292|issue:45701" ], - "prepared_review_unit_hash": "d721ae199d9b165f79e0d21134ac6dc55a7bc19992090ee00dcdcd3e423273de", + "prepared_review_unit_hash": "5b883c8ce00e74b0a36228c950cf9e8b0de0b36a0c46744bfffa258b55ef8857", "component_budget": { "node_count": 426, "item_count": 426, - "soft_pair_count": 1572, - "serialized_chars": 814130, - "estimated_input_tokens": 203533, - "estimated_eval_tokens": 407322 + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 }, "budget": { - "node_count": 47, - "item_count": 47, - "soft_pair_count": 31, - "serialized_chars": 36241, - "estimated_input_tokens": 9061, - "estimated_eval_tokens": 18378 + "node_count": 48, + "item_count": 48, + "soft_pair_count": 32, + "serialized_chars": 37098, + "estimated_input_tokens": 9275, + "estimated_eval_tokens": 18806 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The set is a loose similarity cluster of mostly unrelated bug reports; they share broad themes like import errors, model-loading regressions, and config issues, but no pair looks like the same concrete defect.", - "confidence": 0.78, - "canonical_issue_reason": "Issue 36246 is the clearest, most concrete import-error report in the set, with an exact traceback/module path and enough discussion to serve as a reasonable anchor.", + "summary": "This cluster looks like a noisy semantic grouping of mostly unrelated issues. A few are thematically similar (import errors, Qwen/MoE regressions, config strictness), but none of the soft pairs look like the same concrete bug/change.", + "confidence": 0.82, + "canonical_issue_reason": "issue:44246 is the broadest and most discussion-heavy report in the set, so it is the least-bad anchor for a noisy cluster.", "canonical_pr_reason": null, - "best_issue_reason": "It has the strongest specificity and reproducibility among the candidate issues, making it the best representative issue for this cluster.", + "best_issue_reason": "issue:44246 is the most generic/high-activity issue here, but the cluster still does not form a true duplicate family.", "best_pr_reason": null, "soft_edge_verdicts": [ + { + "left": "issue:44589", + "right": "issue:45362", + "accept": false, + "reason": "Different failures: a missing storage import vs a Qwen3.5 crash in chat generation; same broad ML stack, not the same bug." + }, { "left": "issue:44188", "right": "issue:44336", "accept": false, - "reason": "Different subsystems and failures: attention-kernel divergence vs ANSI color handling in loading reports." + "reason": "Unrelated code paths: torch.compile attention-kernel divergence vs ANSI codes in loading_report." }, { "left": "issue:45561", "right": "issue:45563", "accept": false, - "reason": "Test-worker file race and stale generation warning are unrelated bugs." + "reason": "One is a pytest-xdist file race, the other a stale generation warning; no shared underlying defect." }, { "left": "issue:42491", "right": "issue:43299", "accept": false, - "reason": "Both involve Qwen3 MoE, but one is LoRA compatibility/loading and the other is Qwen3VL loading; not the same concrete path." + "reason": "Both involve Qwen/MoE version regressions, but one is LoRA compatibility and the other is model loading; not the same concrete issue." }, { "left": "issue:44336", "right": "issue:44373", "accept": false, - "reason": "ANSI output handling and a position_ids docstring issue are unrelated." + "reason": "Docs/formatting bug vs a position_ids docstring issue; unrelated." }, { "left": "issue:44242", "right": "issue:45593", "accept": false, - "reason": "Both mention missing losses, but they affect different models and different conditional paths." + "reason": "Both mention missing losses, but they are different models and different gating logic; not mergeable as one PR." }, { "left": "issue:44295", "right": "issue:45397", "accept": false, - "reason": "A position_ids buffer bug and a ZeRO-3 from_pretrained bug are not the same defect." + "reason": "A buffer/position_ids error and a ZeRO3 Gemma-4 load failure are different bugs." }, { "left": "issue:44246", "right": "issue:44496", "accept": false, - "reason": "Import performance is unrelated to model-type resolution failure." + "reason": "Import slowdown vs unrecognized model/config loading failure; no shared code path." }, { "left": "issue:41950", "right": "issue:44220", "accept": false, - "reason": "Video pipeline processor selection and audio fbank feature extraction are different code paths." + "reason": "Video pipeline processor selection and audio fbank extraction are different subsystems." }, { "left": "issue:43531", "right": "issue:45362", "accept": false, - "reason": "Both are Qwen-related, but one is sliding-window behavior and the other is a chat crash; too different to merge." + "reason": "Both are Qwen-related, but one is a sliding_window bug and the other is a chat crash; not the same defect." }, { "left": "issue:42886", @@ -11046,127 +11014,127 @@ "left": "issue:43526", "right": "issue:45003", "accept": false, - "reason": "Beit image-label reduction and sys.modules access in modeling_utils are unrelated bugs." + "reason": "Beit label reduction and unsafe sys.modules access are separate implementation bugs." }, { "left": "issue:36246", "right": "issue:42757", "accept": false, - "reason": "Both are import errors, but for different symbols from different packages and likely different fixes." + "reason": "Both are import errors, but for different packages and missing symbols." }, { "left": "issue:36246", "right": "issue:44351", "accept": false, - "reason": "Both are missing-name import failures, but the missing APIs and likely causes differ." + "reason": "Different missing exports: Qwen2_5_VLImageProcessor vs HybridCache." }, { - "left": "issue:45003", - "right": "issue:45561", + "left": "issue:44336", + "right": "issue:44485", "accept": false, - "reason": "A modeling_utils import/access issue and a pytest-xdist file race are unrelated." + "reason": "ANSI-code output formatting and GLM-5 RoPE implementation are unrelated." }, { - "left": "issue:44336", - "right": "issue:44485", + "left": "issue:45003", + "right": "issue:45561", "accept": false, - "reason": "Terminal color handling and GLM-5 RoPE implementation are unrelated." + "reason": "sys.modules access in modeling_utils vs a captured_info.txt race in tests are not the same bug." }, { - "left": "issue:44246", + "left": "issue:44242", "right": "issue:44871", "accept": false, - "reason": "Import latency and Gemma eos_token_id inconsistency are unrelated." + "reason": "Load-balancing-loss handling and Gemma eos_token_id config mismatch are unrelated." }, { "left": "issue:43352", "right": "issue:45250", "accept": false, - "reason": "A model-specific Flash Attention unsupported error and a generic Flash Attention topic are not the same bug." + "reason": "A model-specific Flash Attention support gap is not the same as the generic Flash Attention topic." }, { "left": "issue:43673", "right": "issue:44561", "accept": false, - "reason": "Cache behavior during chunked prefill and trust_remote_code breakage from a removed helper are unrelated." + "reason": "Generation cache regression during chunked prefill vs removal of is_torch_fx_available affecting trust_remote_code are different changes." }, { "left": "issue:36246", "right": "issue:43596", "accept": false, - "reason": "ImportError for a processor class and a DeepSpeed/BertModel IndexError are different failure modes." + "reason": "ImportError for a missing processor symbol vs a DeepSpeed/Bert IndexError are unrelated." }, { "left": "issue:44242", "right": "issue:44857", "accept": false, - "reason": "Different model losses and triggers; not the same auxiliary-loss bug." + "reason": "Both are training-loss failures, but on different models and different failure mechanisms." }, { "left": "issue:44779", "right": "issue:45292", "accept": false, - "reason": "Tokenizer regression and resize_token_embeddings/output_embeddings mismatch are different issues." + "reason": "Tokenizer regression vs resize_token_embeddings not updating output embeddings are distinct bugs." }, { "left": "issue:44295", "right": "issue:44749", "accept": false, - "reason": "Position_ids buffer handling and tokenizer slowdown across versions are unrelated." + "reason": "A position_ids buffer error and a version-related performance slowdown are not the same issue." }, { "left": "issue:44162", "right": "issue:45200", "accept": false, - "reason": "ESM2 model breakage and Gemma-4 mm_token_type_ids defaults are unrelated." + "reason": "ESM2 breakage and Gemma-4 mm_token_type_ids defaults are separate model-specific problems." }, { "left": "issue:43901", "right": "issue:45003", "accept": false, - "reason": "A docstring mismatch and unsafe sys.modules access are different classes of bugs." + "reason": "Documentation mismatch for return_all_scores does not match the sys.modules access bug." }, { "left": "issue:45412", "right": "issue:46123", "accept": false, - "reason": "Memory release on deletion and partial-batch masking behavior are unrelated." + "reason": "Memory leak on delete/GC and a final-partial-batch pipeline bug are unrelated." }, { "left": "issue:44496", "right": "issue:45003", "accept": false, - "reason": "Model config resolution failure and sys.modules access are unrelated." + "reason": "Model config recognition failure and unsafe sys.modules access are different code paths." }, { - "left": "issue:45292", - "right": "issue:45584", + "left": "issue:44877", + "right": "issue:45375", "accept": false, - "reason": "Embedding resize behavior and Whisper empty-transcription failure are different code paths." + "reason": "Both touch strict config handling, but one rejects granite_speech loading while the other silently drops a missing field in a different config class; not the same concrete bug." }, { - "left": "issue:44877", - "right": "issue:45375", + "left": "issue:45292", + "right": "issue:45584", "accept": false, - "reason": "Both involve config strictness, but one blocks loading while the other silently drops a field; not the same defect." + "reason": "Embedding resize behavior and Whisper empty-transcription failure are unrelated." }, { "left": "issue:45987", "right": "issue:46082", "accept": false, - "reason": "Cache sequence-length return type and LlamaConfig head_dim validation are unrelated." + "reason": "StaticCache return type contract and LlamaConfig head_dim validation are different areas." }, { "left": "issue:43262", "right": "issue:44246", "accept": false, - "reason": "Audio sampling-rate defaults and import-time slowdown are unrelated." + "reason": "Audio chat-template sampling-rate defaults and import-time slowness are unrelated." }, { "left": "issue:45292", "right": "issue:45701", "accept": false, - "reason": "Embedding resize/output embedding sync and tokenizer-version changes are different bugs." + "reason": "Embedding resizing behavior and version-dependent tokenization changes are not the same defect." } ] }, @@ -11182,7 +11150,7 @@ "component_index": 1, "component_count": 17, "review_unit_index": 36, - "review_unit_count": 50, + "review_unit_count": 49, "status": "reviewed", "reason": null, "source": "llm", @@ -11240,13 +11208,13 @@ "issue:46170" ], "soft_pairs": [ - "issue:43257|issue:45137", "issue:45375|issue:45759", + "issue:43257|issue:45137", "issue:44987|issue:45459", "issue:43232|issue:44336", "issue:44995|issue:45200", - "issue:44479|issue:46170", "issue:43475|issue:43540", + "issue:44479|issue:46170", "issue:43065|issue:43408", "issue:42886|issue:45229", "issue:44466|issue:45335", @@ -11273,220 +11241,226 @@ "issue:44623|issue:45656", "issue:44749|issue:45656" ], - "prepared_review_unit_hash": "99fb2c89c50cde501b8ae482e0355b6e90ce28af461087ed59f8c08b4f965d69", + "prepared_review_unit_hash": "9100cbc4a71d1794d0322bedf61bcae33d0f97b0bdfc41d2bbf13d49ac43aa3a", "component_budget": { "node_count": 426, "item_count": 426, - "soft_pair_count": 1572, - "serialized_chars": 814130, - "estimated_input_tokens": 203533, - "estimated_eval_tokens": 407322 + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 32, - "serialized_chars": 37425, - "estimated_input_tokens": 9357, - "estimated_eval_tokens": 18970 + "serialized_chars": 37429, + "estimated_input_tokens": 9358, + "estimated_eval_tokens": 18972 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This set is mostly a grab bag of unrelated Transformers bug reports, with a loose recurring theme of version-upgrade regressions. None of the soft pairs look like the same underlying issue closely enough to merge.", - "confidence": 0.81, - "canonical_issue_reason": "Issue 45278 is the broadest umbrella regression report here: multiple import errors after upgrading from 4.57.0 to 5.5.0. It is the best anchor among otherwise heterogeneous bug reports.", + "summary": "This set is mostly a collection of unrelated closed issues across tokenizer, generation, multimodal, DeepSpeed, and config-loading bugs. I do not see any soft pair that looks like the same concrete bug/change, so I would not merge any of them as duplicates.", + "confidence": 0.9, + "canonical_issue_reason": "No single issue is a true canonical duplicate anchor here; the set is heterogeneous. If forced to pick a representative, issue 44479 is the broadest single regression report, but it still does not subsume the rest.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 45278 is the most general, user-facing duplicate target in this set and best represents the cluster\u2019s recurring upgrade/regression theme.", + "best_issue_reason": "Issue 44479 is the most representative of the larger recurring regression theme in this set (multi-model Qwen/VL video-input breakage), but it is still only a representative issue, not a duplicate hub.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:43257", - "right": "issue:45137", + "left": "issue:45375", + "right": "issue:45759", "accept": false, - "reason": "Both involve distributed/DeepSpeed behavior, but one is a Qwen3 MOE weight-conversion bug and the other is a ZeRO3 deque IndexError; different failure modes and code paths." + "reason": "Both involve Qwen 3.5/3.6 composite configs, but one is strict-config field dropping and the other is from_config text_config unwrapping; different bugs." }, { - "left": "issue:45375", - "right": "issue:45759", + "left": "issue:43257", + "right": "issue:45137", "accept": false, - "reason": "Both concern composite Qwen configs, but one is a missing config field under @strict and the other is AutoModelForCausalLM.from_config not unwrapping text_config; not the same bug." + "reason": "Both mention DeepSpeed, but one is MoE weight conversion on load and the other is a ZeRO3 deque/index error; not the same code path." }, { "left": "issue:44987", "right": "issue:45459", "accept": false, - "reason": "One is a model-loading failure on transformers>=5.1.0, the other hides tokenizer errors when protobuf is absent; related area, but different underlying problems." + "reason": "One is a model-loading regression for a specific repo/version, the other is tokenizer error masking when protobuf is missing." }, { "left": "issue:43232", "right": "issue:44336", "accept": false, - "reason": "Generation kwargs after sync_gpus and ANSI codes in loading_report are unrelated subsystems." + "reason": "Generation kwargs after sync_gpus is unrelated to ANSI output in loading_report." }, { "left": "issue:44995", "right": "issue:45200", "accept": false, - "reason": "A stale cache crash on second forward is not the same as mm_token_type_ids defaulting behavior for Gemma 4." + "reason": "Different models and failures: stale indexer cache on second forward vs missing mm_token_type_ids defaulting." }, { - "left": "issue:44479", - "right": "issue:46170", + "left": "issue:43475", + "right": "issue:43540", "accept": false, - "reason": "Video-input regression in Qwen multimodal models is unrelated to a Windows/ROCm import dependency issue." + "reason": "Both are video/model issues, but one is a missing output attribute in SAM 3 and the other is a Qwen3OmniMoe video-processing ValueError." }, { - "left": "issue:43475", - "right": "issue:43540", + "left": "issue:44479", + "right": "issue:46170", "accept": false, - "reason": "Both are multimodal/video-related, but they affect different model families and distinct attributes/exceptions." + "reason": "Video-input regression in Qwen models is unrelated to Windows/ROCm import and torch.distributed dependency handling." }, { "left": "issue:43065", "right": "issue:43408", "accept": false, - "reason": "Dummy Conv2d in Sam3PixelDecoder and a sam3_video/sam3_tracker type warning are different SAM3 issues, not the same defect." + "reason": "Both are SAM 3 related, but one is a dummy Conv2d implementation detail and the other is a model-type warning during instantiation." }, { "left": "issue:42886", "right": "issue:45229", "accept": false, - "reason": "Tokenizer offline-cache loading and Gemma4 multi-GPU CUDA OOM are unrelated." + "reason": "Tokenizer offline-cache loading is unrelated to Gemma4 multi-GPU OOM." }, { "left": "issue:44466", "right": "issue:45335", "accept": false, - "reason": "Both touch tied embeddings, but one is inconsistent serialization by device and the other is resize_token_embeddings not propagating; not the same concrete bug." + "reason": "Both touch tied embeddings, but one is serialization inconsistency and the other is resize_token_embeddings not propagating to decoder.embed_tokens." }, { "left": "issue:44060", "right": "issue:44077", "accept": false, - "reason": "These are different model/config problems: a Qwen3-Next tied-weights warning versus a patchtsmixer post_init policy issue." + "reason": "Different subsystems: Qwen3-Next tied-weights warning vs patchtsmixer post_init policy." }, { "left": "issue:43653", "right": "issue:44162", "accept": false, - "reason": "BigBirdTokenizer special-token registration and ESM2 breakage are separate issues." + "reason": "BigBirdTokenizer special-token registration and ESM2 breakage are unrelated." }, { "left": "issue:45397", "right": "issue:45656", "accept": false, - "reason": "Both mention DeepSpeed, but one is a from_pretrained/load-time issue and the other is an optimizer-step duplication bug." + "reason": "Gemma-4 ZeRO3 loading and duplicate optimizer.step behavior are different DeepSpeed problems." }, { "left": "issue:44995", "right": "issue:45588", "accept": false, - "reason": "A stale indexer cache crash is not the same as a flash-attention AttributeError on s_aux=None." + "reason": "Stale cache on second forward is unrelated to flash_attention s_aux=None handling." }, { "left": "issue:44368", "right": "issue:44977", "accept": false, - "reason": "Both involve Qwen3.5, but one is a tied-embeddings warning during LoRA fine-tuning and the other is a flash-attention generation regression." + "reason": "One is a tied-embeddings warning during fine-tuning, the other is a flash-attention generation regression." }, { "left": "issue:45137", "right": "issue:45600", "accept": false, - "reason": "ZeRO3 deque failure and removed Sam3LiteText auto-mapping references are unrelated." + "reason": "ZeRO3 deque error and auto_mappings broken CI are unrelated." }, { "left": "issue:44368", "right": "issue:45563", "accept": false, - "reason": "These are both warnings, but one is about tie_word_embeddings and the other is a stale num_return_sequences warning in paged generate." + "reason": "Tied-embeddings warning is unrelated to the stale num_return_sequences warning." }, { "left": "issue:44877", "right": "issue:46170", "accept": false, - "reason": "Granite_speech strict-config loading and a Windows ROCm import error are unrelated." + "reason": "Strict config loading for granite_speech is unrelated to the Windows ROCm import error." }, { "left": "issue:43673", "right": "issue:44734", "accept": false, - "reason": "Both are cache/generation problems, but chunked_prefill cache missing and serve KV-cache continuation indexing are different code paths and symptoms." + "reason": "Both concern generation/cache, but one is missing cache in chunked_prefill and the other is wrong tensor indexing in the serve API." }, { "left": "issue:43931", "right": "issue:44368", "accept": false, - "reason": "A Qwen3-VL weight-shape mismatch is unrelated to a Qwen3.5 tie-word-embeddings warning." + "reason": "Weight-shape mismatch loading Qwen3-VL is unrelated to a tied-embeddings warning." }, { "left": "issue:44530", "right": "issue:46170", "accept": false, - "reason": "PagedAttentionCache invalid group type on Qwen3.5 is unrelated to the Windows ROCm import dependency issue." + "reason": "PagedAttentionCache invalid group type on Qwen3.5 is unrelated to the ROCm import/dependency issue." }, { "left": "issue:43065", "right": "issue:46123", "accept": false, - "reason": "Sam3PixelDecoder dummy Conv2d and MaskGenerationPipeline final-batch handling are unrelated." + "reason": "SAM 3 pixel decoder dummy layer and MaskGenerationPipeline partial-batch dropping are different bugs." }, { "left": "issue:44933", "right": "issue:45459", "accept": false, - "reason": "Both are import-related, but one is a nonexistent image_utils import and the other is tokenizer error masking without protobuf; different defects." + "reason": "Missing import from image_utils is unrelated to protobuf error masking in tokenizer decode." }, { "left": "issue:44869", "right": "issue:45356", "accept": false, - "reason": "Whisper timestamp decoding and Kimi-K2.5 codec/regex handling are different tokenizer/decode issues." + "reason": "Both are tokenizer-adjacent, but they affect different models and code paths (Whisper timestamp decode vs Kimi-K2.5 codec handling)." }, { "left": "issue:44297", "right": "issue:44623", "accept": false, - "reason": "Both concern save_pretrained, but one is tokenizer_class metadata and the other is missing processor files." + "reason": "tokenizer.save_pretrained mismatch and processor.save_pretrained missing files are separate serialization bugs." }, { - "left": "issue:43673", - "right": "issue:45278", + "left": "issue:43232", + "right": "issue:43673", "accept": false, - "reason": "Generation cache missing in chunked_prefill and broad import failures after upgrade are unrelated regressions." + "reason": "Different generation-cache failures: sync_gpus kwargs update vs chunked_prefill cache availability." }, { "left": "issue:44361", "right": "issue:44749", "accept": false, - "reason": "MLukeTokenizer task-time AttributeError and a post-upgrade filtering slowdown are different bugs." + "reason": "MLukeTokenizer task AttributeError is unrelated to the Chinese report about performance slowing after upgrade." }, { "left": "issue:43646", "right": "issue:44162", "accept": false, - "reason": "Both are broad breakage reports, but one is custom model initialization in v5.0.0 and the other is ESM2-specific breakage." + "reason": "Custom model initialization breakage in Transformers 5.0.0 and ESM2 breakage are not the same issue." }, { "left": "issue:44871", "right": "issue:46132", "accept": false, - "reason": "Gemma eos_token_id inconsistency and AttentionInterface.register behavior changes are unrelated." + "reason": "Gemma eos_token_id mismatch is unrelated to AttentionInterface.register behavior changes." + }, + { + "left": "issue:43673", + "right": "issue:45278", + "accept": false, + "reason": "Missing generation cache during chunked_prefill is not the same as general import errors after upgrade." }, { "left": "issue:44623", "right": "issue:45656", "accept": false, - "reason": "Missing processor save files and double optimizer stepping in DeepSpeed are unrelated." + "reason": "Processor save_pretrained file omissions and duplicate optimizer.step calls are unrelated." }, { "left": "issue:44749", "right": "issue:45656", "accept": false, - "reason": "A filtering performance regression after upgrade is not the same as optimizer.step being called twice under DeepSpeed." + "reason": "Performance regression after upgrade and duplicate optimizer.step are different symptoms and code paths." } ] }, @@ -11502,7 +11476,7 @@ "component_index": 1, "component_count": 17, "review_unit_index": 37, - "review_unit_count": 50, + "review_unit_count": 49, "status": "reviewed", "reason": null, "source": "llm", @@ -11592,235 +11566,235 @@ "issue:43526|issue:45561", "issue:43504|issue:45561" ], - "prepared_review_unit_hash": "23297d3a6fc0d5d398b96c930f9ac2183e030daa8600f871916144a7fc79aacf", + "prepared_review_unit_hash": "0dacca25acb9973197aaba1b0cbaff8aa09705d29a795d57c76de68c085f3e30", "component_budget": { "node_count": 426, "item_count": 426, - "soft_pair_count": 1572, - "serialized_chars": 814130, - "estimated_input_tokens": 203533, - "estimated_eval_tokens": 407322 + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 }, "budget": { "node_count": 47, "item_count": 47, "soft_pair_count": 32, - "serialized_chars": 36483, - "estimated_input_tokens": 9121, - "estimated_eval_tokens": 18498 + "serialized_chars": 36486, + "estimated_input_tokens": 9122, + "estimated_eval_tokens": 18500 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The cluster is mostly heterogeneous: many items mention Transformers, but they describe different bugs, models, or code paths. Only one soft edge is a clear duplicate: the Qwen3ForCausalLM VRAM leak reports.", - "confidence": 0.91, - "canonical_issue_reason": "Issue 45310 is the clearest representative of the duplicated Qwen3.5-MoE from_pretrained failure; it is concrete, cross-referenced, and a later item in the same exact-title repost line (45478).", + "summary": "The cluster is mostly heterogeneous, but it does contain two clear duplicate pockets: the Qwen3ForCausalLM VRAM-leak pair (42673/43704) and an exact-title duplicate pair for '[BUG] transformers>=5.4.0, Qwen3.5 Moe from_pretrained error' (45310/45478). The remaining soft pairs are largely false positives across broadly related subsystems, not the same concrete bug.", + "confidence": 0.84, + "canonical_issue_reason": "issue:42673 is the strongest canonical issue for the cluster because it anchors the clearest confirmed duplicate pair (exact same Qwen3ForCausalLM VRAM-leak report as issue:43704) and has the richer discussion thread; the packet also contains a separate exact-title duplicate pair, issue:45310 and issue:45478.", "canonical_pr_reason": null, - "best_issue_reason": "45310 is the best anchor for the cluster because it is one of the few clearly duplicate-worthy reports and is more substantive than its repost/near-repost counterparts.", + "best_issue_reason": "issue:42673 is the best single representative issue in this packet because it is part of the clearest duplicate pair and has the strongest supporting discussion, while the rest of the set is mixed and the other duplicate pocket is separate.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44749", "right": "issue:45310", "accept": false, - "reason": "Different problems: data filtering slowdown vs Qwen3.5-MoE from_pretrained error." + "reason": "Different failures: data filtering slowdown vs a Qwen3.5-Moe from_pretrained error." }, { "left": "issue:42994", "right": "issue:43482", "accept": false, - "reason": "Quantized model saving failure and Qwen2.5-GGUF loading failure are distinct code paths." + "reason": "Different code paths: quantized model saving vs GGUF loading on a different model/version issue." }, { "left": "issue:43475", "right": "issue:45381", "accept": false, - "reason": "Different model families and failures: SAM 3 Video attribute error vs qwen2.5-vl position id issue." + "reason": "Both are multimodal/video-related, but they describe different models and distinct bugs." }, { "left": "issue:28282", "right": "issue:30990", "accept": false, - "reason": "PyTorch missing ImportError is unrelated to Sentence Transformers hanging during load." + "reason": "Missing-PyTorch import error is unrelated to a Sentence Transformers loading hang." }, { "left": "issue:45230", "right": "issue:45397", "accept": false, - "reason": "Generic bug report provides no concrete overlap with gemma-4 ZeRO3 from_pretrained failure." + "reason": "Generic bug report vs a specific ZeRO3 from_pretrained issue; no evidence they are the same bug." }, { "left": "issue:43475", "right": "issue:43643", "accept": false, - "reason": "SAM 3 Video attribute error is unrelated to AutoConfig trust_remote_code missing fields." + "reason": "Missing model attribute in SAM 3 Video is different from trust_remote_code missing fields in AutoConfig." }, { "left": "issue:44038", "right": "issue:44623", "accept": false, - "reason": "Qwen3-VL-Moe / transformers v5 issue is unrelated to processor.save_pretrained missing files." + "reason": "Qwen3-VL-Moe compatibility and processor.save_pretrained missing files are different failures." }, { "left": "issue:46077", "right": "issue:46123", "accept": false, - "reason": "Gemma4 attention-mask bug is unrelated to MaskGenerationPipeline final-batch dropping." + "reason": "Attention-mask construction bug is unrelated to final-batch result dropping in the pipeline." }, { "left": "issue:41628", "right": "issue:44933", "accept": false, - "reason": "Both involve imports, but they are different missing symbols and likely different fixes." + "reason": "Both are import-related, but they concern different missing symbols and likely different regressions." }, { "left": "issue:42371", "right": "issue:45092", "accept": false, - "reason": "TF backend packaging issue is unrelated to InternVL2 meta-initialization incompatibility." + "reason": "TF32 API guidance is unrelated to remote-code/meta-init incompatibility for InternVL2." }, { "left": "issue:43475", "right": "issue:44661", "accept": false, - "reason": "Different failures: SAM 3 Video output attribute error vs tokenizer mapping-name handling." + "reason": "Different problems: missing encoder output field vs tokenizer-mapping failure in add-new-model-like." }, { "left": "issue:42994", "right": "issue:43525", "accept": false, - "reason": "Quantized saving failure is unrelated to Llama4Config missing pad_token_id." + "reason": "Quantized save failure is unrelated to Llama4Config missing pad_token_id." }, { "left": "issue:45375", "right": "issue:45478", "accept": false, - "reason": "Related model family, but different concrete bugs: missing config field vs from_pretrained error." + "reason": "Related Qwen family, but one is a missing config field and the other is a generic from_pretrained error." }, { "left": "issue:45081", "right": "issue:45103", "accept": false, - "reason": "Tokenizer regex patch crash and auto_docstring annotations crash are unrelated." + "reason": "Tokenizer regex crash and auto_docstring annotation crash are unrelated." }, { "left": "issue:45310", "right": "issue:45375", "accept": false, - "reason": "Same model family, but not the same bug: from_pretrained error vs silently dropped config field." + "reason": "Same model family, but one is a from_pretrained failure and the other is a missing vision config field." }, { "left": "issue:43873", "right": "issue:44945", "accept": false, - "reason": "Quantization offloading behavior and pipeline parallelism output corruption are different issues." + "reason": "Offloading/quantization behavior is unrelated to incorrect output under pipeline parallelism." }, { "left": "issue:41628", "right": "issue:44351", "accept": false, - "reason": "AutoImageProcessor import failure and HybridCache import failure are different missing exports." + "reason": "Both are import errors, but they concern different missing names and likely different export regressions." }, { "left": "issue:43232", "right": "issue:44188", "accept": false, - "reason": "Generation kwargs sync bug and attention-kernel divergence on torch.compile are unrelated." + "reason": "Generation kwargs update after sync_gpus is unrelated to attention-kernel divergence under torch.compile." }, { "left": "issue:42673", "right": "issue:43704", "accept": true, - "reason": "Same title, same Qwen3ForCausalLM VRAM leak with multiple dataloader threads; clear duplicate." + "reason": "Exact same Qwen3ForCausalLM VRAM-leak bug in multi-threaded dataloader use; clear duplicate." }, { "left": "issue:42831", "right": "issue:44222", "accept": false, - "reason": "FP8 accuracy regression is different from FP8 save_pretrained failure." + "reason": "Both involve FP8, but one is an accuracy issue and the other is FP8 save_pretrained for MoE." }, { "left": "issue:44945", "right": "issue:45753", "accept": false, - "reason": "Pipeline-parallel output corruption and Qwen3_5 image infinite loop are different bugs." + "reason": "Incorrect pipeline-parallel output is unrelated to an image-triggered infinite loop in Qwen3_5." }, { "left": "issue:36010", "right": "issue:41628", "accept": false, - "reason": "Both are import errors, but for different symbols and likely unrelated fixes." + "reason": "Different missing imports: GenerationMixin vs AutoImageProcessor." }, { "left": "issue:44222", "right": "issue:45310", "accept": false, - "reason": "FP8 save_pretrained moe is unrelated to the Qwen3.5-MoE from_pretrained error." + "reason": "FP8 save_pretrained MoE is unrelated to Qwen3.5-Moe from_pretrained errors." }, { "left": "issue:42994", "right": "issue:43824", "accept": false, - "reason": "Quantized model saving failure is unrelated to missing Qwen2.5-VL import." + "reason": "Saving a quantized model is unrelated to a missing Qwen2.5-VL class import." }, { "left": "issue:44964", "right": "issue:45137", "accept": false, - "reason": "Phi-4 multimodal loading failure and DeepSpeed ZeRO3 deque error are unrelated." + "reason": "Model loading incompatibility is unrelated to a DeepSpeed ZeRO3 deque error." }, { "left": "issue:42175", "right": "issue:45712", "accept": false, - "reason": "Packaging/backend omission is related only broadly; these are different concrete dummy-object and backend issues." + "reason": "Backend packaging/install issue is different from dummy_pt_objects leaking into dir(transformers)." }, { "left": "issue:43232", "right": "issue:44792", "accept": false, - "reason": "Generation sync bug and janus image-generation test failure are not the same underlying issue." + "reason": "Generation kwargs/sync_gpus bug is unrelated to a janus image-generation test failure." }, { "left": "issue:44038", "right": "issue:45310", "accept": false, - "reason": "Qwen3-VL-Moe transformers v5 bug is unrelated to Qwen3.5-MoE from_pretrained failure." + "reason": "Different Qwen families and different failures: VL-Moe compatibility vs Qwen3.5-Moe from_pretrained." }, { "left": "issue:43673", "right": "issue:44188", "accept": false, - "reason": "Chunked-prefill cache issue and torch.compile attention-kernel divergence are distinct." + "reason": "Chunked_prefill cache regression is unrelated to bidirectional-skip branching in attention kernels." }, { "left": "issue:44496", "right": "issue:45561", "accept": false, - "reason": "Model-type/config loading failure is unrelated to xdist race on captured_info.txt." + "reason": "Unrecognized model/config issue is unrelated to a pytest-xdist race on captured_info.txt." }, { "left": "issue:43526", "right": "issue:45561", "accept": false, - "reason": "BeitImageProcessorFast label reduction bug is unrelated to the xdist file race." + "reason": "BeitImageProcessorFast reduce_labels bug is unrelated to the xdist file race." }, { "left": "issue:43504", "right": "issue:45561", "accept": false, - "reason": "Legacy-field pretrained loading bug is unrelated to the xdist race condition." + "reason": "Legacy-field model loading issue is unrelated to the captured_info.txt race." } ] }, "evaluator_result": { - "accept": true, - "feedback": "The summary is grounded in the packet and stays conservative. The one accepted soft edge (42673/43704) is well supported by the exact same title and issue text, and the other rejected pairs are reasonably explained as different bugs, models, or code paths. No overstatement stands out." + "accept": false, + "feedback": "The summary is not fully grounded in the packet. In addition to the Qwen3ForCausalLM VRAM-leak duplicate pair (42673/43704), the packet also contains an exact-title match between issue:45310 and issue:45478 ('[BUG] transformers>=5.4.0, Qwen3.5 Moe from_pretrained error'), which the analyst omitted. Because the duplicate landscape is understated, the overall cluster summary is too strong." }, "evaluator_used": true, - "retried": false, + "retried": true, "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null @@ -11830,7 +11804,7 @@ "component_index": 1, "component_count": 17, "review_unit_index": 38, - "review_unit_count": 50, + "review_unit_count": 49, "status": "reviewed", "reason": null, "source": "llm", @@ -11847,7 +11821,6 @@ "issue:42222", "issue:42831", "issue:42898", - "issue:43116", "issue:43278", "issue:43377", "issue:43479", @@ -11878,6 +11851,7 @@ "issue:45125", "issue:45137", "issue:45200", + "issue:45245", "issue:45310", "issue:45356", "issue:45405", @@ -11895,8 +11869,8 @@ "issue:43976|issue:44933", "issue:30990|issue:35141", "issue:43479|issue:44869", - "issue:43278|issue:45137", "issue:43479|issue:44079", + "issue:43278|issue:45137", "issue:44623|issue:45310", "issue:42898|issue:45812", "issue:44188|issue:44373", @@ -11904,8 +11878,8 @@ "issue:43784|issue:43824", "issue:44734|issue:45103", "issue:43723|issue:44568", - "issue:42898|issue:44779", "issue:42898|issue:45356", + "issue:42898|issue:44779", "issue:43874|issue:45092", "issue:44060|issue:45125", "issue:45405|issue:45712", @@ -11914,228 +11888,222 @@ "issue:44448|issue:46132", "issue:44933|issue:45712", "issue:44222|issue:44297", - "issue:43116|issue:44190", "issue:44297|issue:45656", "issue:43479|issue:43644", "issue:44190|issue:46082", "issue:42831|issue:44623", + "issue:44222|issue:44623", "issue:30333|issue:36331", - "issue:44222|issue:44623" + "issue:45245|issue:46132" ], - "prepared_review_unit_hash": "5a6a47bedc955f9bd457a9668610489016069e8202faf8795986e922823cab3d", + "prepared_review_unit_hash": "922cf2ceb35bc91764dd684823bc24df138da53a6137f4bbe5dcb1b204568890", "component_budget": { "node_count": 426, "item_count": 426, - "soft_pair_count": 1572, - "serialized_chars": 814130, - "estimated_input_tokens": 203533, - "estimated_eval_tokens": 407322 + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 33, - "serialized_chars": 37729, - "estimated_input_tokens": 9433, - "estimated_eval_tokens": 19122 + "serialized_chars": 37690, + "estimated_input_tokens": 9423, + "estimated_eval_tokens": 19102 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is heterogeneous: the soft pairs connect different bugs in different areas (tokenizers, model configs, DeepSpeed, save_pretrained, examples, multimodal code, etc.), not the same underlying defect. I would not merge any of the proposed soft edges.", - "confidence": 0.98, - "canonical_issue_reason": "If a representative is needed, issue 42898 is the broadest and clearest user-facing regression in the set (v5 tokenization behavior change), but it is not a true duplicate of the others.", + "summary": "The cluster is mostly a mix of unrelated Transformers v5-era bug reports. The soft matches are topical similarities, but none look like true duplicates of the same underlying defect.", + "confidence": 0.62, + "canonical_issue_reason": "issue:42898 is the broadest tokenizer-regression report here and sits near several tokenizer-related soft matches, so it is the best cluster anchor.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 42898 is the strongest standalone representative: concrete, reproducible, and broadly understandable, even though the cluster does not contain a single shared bug.", + "best_issue_reason": "issue:45812 is the clearest high-impact standalone bug report: explicit v4\u2192v5 regression, broad model impact, and a concrete symptom.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43377", "right": "issue:44242", "accept": false, - "reason": "Different failures: batched-vs-single encoder output mismatch vs missing MoE load-balancing loss when router logits are disabled." + "reason": "Different failure modes: MIMI padding-mask parity vs MoE load-balancing loss gating." }, { "left": "issue:42222", "right": "issue:45200", "accept": false, - "reason": "Unrelated models and code paths: broken ViTPose model files vs Gemma 4 mm_token_type_ids defaulting for text-only fine-tuning." + "reason": "Different bugs: vitpose model import breakage vs Gemma 4 mm_token_type_ids defaulting." }, { "left": "issue:42175", "right": "issue:44933", "accept": false, - "reason": "Different problems: packaging backend selection vs a missing import from image_utils." + "reason": "Packaging/backend extra issue vs nonexistent import from image_utils." }, { "left": "issue:42898", "right": "issue:44568", "accept": false, - "reason": "Both tokenizer-related, but different regressions: clean_up_tokenization_spaces behavior vs missing BOS/EOS insertion for one tokenizer." + "reason": "Both are tokenizer regressions in v5, but one is clean_up_tokenization_spaces behavior and the other is add_special_tokens/BOS-EOS handling." }, { "left": "issue:43976", "right": "issue:44933", "accept": false, - "reason": "Python version compatibility issue vs missing import; no shared underlying bug." + "reason": "Python support/version mismatch vs image_utils import error." }, { "left": "issue:30990", "right": "issue:35141", "accept": false, - "reason": "Sentence-transformers loading hang is unrelated to output embedding reinitialization during token embedding resize." + "reason": "Sentence Transformers loading hang is unrelated to embedding reinitialization after resize." }, { "left": "issue:43479", "right": "issue:44869", "accept": false, - "reason": "Different domains: multimodal config defaults vs Whisper timestamp decode crash." + "reason": "Config None-default initialization bug vs Whisper word-timestamp decode crash." }, { - "left": "issue:43278", - "right": "issue:45137", + "left": "issue:43479", + "right": "issue:44079", "accept": false, - "reason": "BF16/FP32 dtype regression during eval is unrelated to DeepSpeed ZeRO3 deque underflow." + "reason": "Config initialization bug vs ModelOutput key assignment bug; different code paths." }, { - "left": "issue:43479", - "right": "issue:44079", + "left": "issue:43278", + "right": "issue:45137", "accept": false, - "reason": "Config default initialization bug vs ModelOutput key assignment bug; different code paths." + "reason": "Embedding dtype drift in eval is unrelated to DeepSpeed ZeRO3 deque underflow." }, { "left": "issue:44623", "right": "issue:45310", "accept": false, - "reason": "save_pretrained file persistence issue vs Qwen3.5 MoE from_pretrained failure; not the same defect." + "reason": "Processor save_pretrained missing files vs Qwen3.5 MoE from_pretrained failure." }, { "left": "issue:42898", "right": "issue:45812", "accept": false, - "reason": "Both are tokenizer regressions, but they affect different behaviors and model families (v5 cleanup behavior vs Granite token IDs)." + "reason": "Related tokenizer regressions, but distinct defects: cleanup-space semantics vs Granite token IDs." }, { "left": "issue:44188", "right": "issue:44373", "accept": false, - "reason": "Attention-kernel divergence under torch.compile is unrelated to a docstring typo." + "reason": "torch.compile attention branching bug vs docstring wording issue." }, { "left": "issue:30333", "right": "issue:33290", "accept": false, - "reason": "MLflow run-state reporting on failure is unrelated to an OOM in DeepSpeed+Adafactor." + "reason": "MLflow reporting on failed jobs is unrelated to Adafactor OOM under DeepSpeed." }, { "left": "issue:43784", "right": "issue:43824", "accept": false, - "reason": "Import error in sentence-transformers due to nn missing is unrelated to missing Qwen2.5-VL symbol export." + "reason": "Sentence-transformers import NameError vs missing Qwen2.5-VL class export." }, { "left": "issue:44734", "right": "issue:45103", "accept": false, - "reason": "Serve KV-cache indexing crash is a runtime path bug; auto_docstring crash is a docs introspection bug." + "reason": "KV-cache continuation indexing crash vs auto-docstring crash with future annotations." }, { "left": "issue:43723", "right": "issue:44568", "accept": false, - "reason": "Tokenizer loading regression and add_special_tokens/BOS-EOS behavior are different tokenizer issues." + "reason": "Tokenizer loading issue vs special-token insertion regression; not the same bug." }, { "left": "issue:42898", - "right": "issue:44779", + "right": "issue:45356", "accept": false, - "reason": "Both involve tokenizer regressions in v5, but the concrete failure modes differ: cleanup behavior vs DeepSeek codec handling." + "reason": "Different tokenizer regressions: cleanup-space behavior vs Kimi codec handling." }, { "left": "issue:42898", - "right": "issue:45356", + "right": "issue:44779", "accept": false, - "reason": "Different tokenizer regressions with different root causes (cleanup semantics vs Kimi codec/regex warning behavior)." + "reason": "Different tokenizer bugs: cleanup-space behavior vs Deepseek incorrect results." }, { "left": "issue:43874", "right": "issue:45092", "accept": false, - "reason": "Missing image-patch count method in a fast processor is unrelated to remote-code checkpoint incompatibility with meta init." + "reason": "Missing image patch-count API vs remote-code checkpoint/meta-init incompatibility." }, { "left": "issue:44060", "right": "issue:45125", "accept": false, - "reason": "Incorrect tied-weights warning in Qwen3-Next is unrelated to missing _tp_plan in Qwen3_5Moe." + "reason": "Incorrect tied-weights warning vs missing tensor-parallel plan." }, { "left": "issue:45405", "right": "issue:45712", "accept": false, - "reason": "Dependency version bump to unreleased peft is unrelated to leftover dummy classes leaking into API surface." + "reason": "PEFT version pinning issue vs dummy-class cleanup/leakage." }, { "left": "issue:42898", "right": "issue:44448", "accept": false, - "reason": "Different v5 regressions: tokenization cleanup behavior vs model output differences for Pegasus." + "reason": "Tokenizer cleanup-space regression vs Pegasus output mismatch; different symptoms and likely causes." }, { "left": "issue:44295", "right": "issue:45656", "accept": false, - "reason": "Position_ids buffer read error is unrelated to DeepSpeed calling optimizer.step() twice." - }, - { - "left": "issue:44222", - "right": "issue:44297", - "accept": false, - "reason": "FP8 save_pretrained behavior and tokenizer_class metadata mismatch are separate save/persist bugs." - }, - { - "left": "issue:43116", - "right": "issue:44190", - "accept": false, - "reason": "Example script multi-label output bug is unrelated to local dataset loading in a different example script." + "reason": "position_ids buffer access error vs optimizer step being called twice under DeepSpeed." }, { "left": "issue:44297", "right": "issue:45656", "accept": false, - "reason": "Tokenizer metadata serialization bug is unrelated to DeepSpeed double optimizer stepping." + "reason": "tokenizer.save_pretrained metadata mismatch vs duplicated optimizer stepping." }, { "left": "issue:43479", "right": "issue:43644", "accept": false, - "reason": "Incorrect default multimodal config initialization is unrelated to non-persistent buffer junk filling." + "reason": "Config default initialization bug vs non-persistent buffer junk-fill bug." }, { "left": "issue:44190", "right": "issue:46082", "accept": false, - "reason": "Local dataset loading failure in an example script is unrelated to LlamaConfig head_dim validation." + "reason": "Local dataset script loading issue vs LlamaConfig head_dim validation." }, { "left": "issue:42831", "right": "issue:44623", "accept": false, - "reason": "FineGrainedFP8 accuracy issue is unrelated to processor.save_pretrained missing files." + "reason": "FineGrainedFP8 accuracy issue vs processor.save_pretrained missing files." + }, + { + "left": "issue:44222", + "right": "issue:44623", + "accept": false, + "reason": "FP8 MoE save_pretrained bug vs processor file persistence bug." }, { "left": "issue:30333", "right": "issue:36331", "accept": false, - "reason": "MLflow reporting on training failure and CustomTrainer compute_loss signature breakage are distinct trainer issues." + "reason": "Failed-job MLflow reporting vs CustomTrainer.compute_loss kwarg incompatibility." }, { - "left": "issue:44222", - "right": "issue:44623", + "left": "issue:45245", + "right": "issue:46132", "accept": false, - "reason": "MoE FP8 save_pretrained bug and processor.save_pretrained missing files affect different serialization paths." + "reason": "Category overflow in embeddings vs AttentionInterface.register behavior change." } ] }, @@ -12151,7 +12119,7 @@ "component_index": 1, "component_count": 17, "review_unit_index": 39, - "review_unit_count": 50, + "review_unit_count": 49, "status": "reviewed", "reason": null, "source": "llm", @@ -12166,6 +12134,7 @@ "issue:34689", "issue:35141", "issue:38175", + "issue:39401", "issue:41628", "issue:42757", "issue:42831", @@ -12189,6 +12158,7 @@ "issue:44484", "issue:44485", "issue:44514", + "issue:44568", "issue:44623", "issue:44829", "issue:44871", @@ -12198,23 +12168,20 @@ "issue:45071", "issue:45092", "issue:45103", - "issue:45245", "issue:45310", "issue:45357", "issue:45406", "issue:45834", "issue:46002", "issue:46050", - "issue:46082", - "issue:46132" + "issue:46082" ], "soft_pairs": [ - "issue:45245|issue:46132", "issue:30333|issue:34689", "issue:22355|issue:28282", "issue:41628|issue:42757", - "issue:44206|issue:45103", "issue:45071|issue:46082", + "issue:44206|issue:45103", "issue:28282|issue:30333", "issue:30333|issue:35141", "issue:44297|issue:45310", @@ -12229,239 +12196,240 @@ "issue:44514|issue:45103", "issue:43723|issue:44871", "issue:44263|issue:44360", - "issue:30990|issue:38175", "issue:45092|issue:45357", + "issue:30990|issue:38175", "issue:42907|issue:46050", "issue:43723|issue:44991", "issue:43723|issue:44442", - "issue:44829|issue:46002", "issue:44261|issue:45092", - "issue:42831|issue:45310", + "issue:44829|issue:46002", "issue:41628|issue:43824", + "issue:42831|issue:45310", "issue:43479|issue:44913", "issue:42831|issue:44038", - "issue:44263|issue:44373" + "issue:44263|issue:44373", + "issue:39401|issue:44568" ], - "prepared_review_unit_hash": "bb6897b8beb790952d19fd731c59160f7c32f48c299af3d29fbe026b0ea6b6de", + "prepared_review_unit_hash": "a4c5eb0b6a9270c2a115f25123527a7d08d11663a41293431ac04670770e0e70", "component_budget": { "node_count": 426, "item_count": 426, - "soft_pair_count": 1572, - "serialized_chars": 814130, - "estimated_input_tokens": 203533, - "estimated_eval_tokens": 407322 + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 32, - "serialized_chars": 37046, - "estimated_input_tokens": 9262, - "estimated_eval_tokens": 18780 + "serialized_chars": 37080, + "estimated_input_tokens": 9270, + "estimated_eval_tokens": 18796 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The cluster is heterogeneous: it mixes unrelated import errors, tokenizer/model-loading regressions, config validation bugs, quantization/save-pretrained issues, and a few documentation or runtime edge cases. I do not see any true duplicate PRs (none are present) or issue pairs that clearly describe the same underlying bug.", - "confidence": 0.96, - "canonical_issue_reason": "issue:43723 is the strongest representative of the cluster\u2019s most repeated theme: v5-era loading regressions around tokenizer/model initialization. It is concrete, user-facing, and has several nearby but still distinct tokenizer-related soft matches.", + "summary": "This cluster is mostly a loose collection of unrelated Transformers bugs; I don\u2019t see any soft pair that looks like the same underlying issue. The best representative issue is the broad v5 tokenizer-loading regression.", + "confidence": 0.84, + "canonical_issue_reason": "issue:43723 is the broadest, most central report in the set: a version-scoped tokenizer loading failure in v5, which is the closest thing to a unifying theme here.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43723 is the best single issue to anchor this cluster because it has a clear repro, broad impact, and the most topical overlap with other items in the set. It is still not a duplicate of the others.", + "best_issue_reason": "issue:43723 is the clearest umbrella issue: it is specific enough to be actionable, but general enough to represent the recurring tokenizer-import/load regressions better than the model-specific reports.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:45245", - "right": "issue:46132", - "accept": false, - "reason": "Different problems: category-limit/runtime error vs AttentionInterface.register behavior change. Same general ML library area only." - }, { "left": "issue:30333", "right": "issue:34689", "accept": false, - "reason": "MLflow job reporting vs Llama vision model loading regression are unrelated code paths." + "reason": "Different failures: MLflow reporting on training job failure vs model loading regression for Llama 3.2 Vision." }, { "left": "issue:22355", "right": "issue:28282", "accept": false, - "reason": "Both are import errors, but for different missing components (onnx module vs PyTorch dependency)." + "reason": "Both are import errors, but one is missing `transformers.onnx` and the other is missing PyTorch; different root causes." }, { "left": "issue:41628", "right": "issue:42757", "accept": false, - "reason": "Different missing top-level imports with different root causes; not the same bug." + "reason": "One is a missing `AutoImageProcessor` import in transformers, the other is a dependency API mismatch in `huggingface_hub`." }, { - "left": "issue:44206", - "right": "issue:45103", + "left": "issue:45071", + "right": "issue:46082", "accept": false, - "reason": "Feature extractor argument handling vs auto-docstring annotation introspection are unrelated." + "reason": "`PretrainedConfig` type-checking regression is unrelated to LlamaConfig rejecting an explicit `head_dim` value." }, { - "left": "issue:45071", - "right": "issue:46082", + "left": "issue:44206", + "right": "issue:45103", "accept": false, - "reason": "Config type-checking regression and LlamaConfig head_dim validation are distinct bugs." + "reason": "Different code paths: a feature extractor crashing on an unsupported arg vs an auto-docstring annotation introspection crash." }, { "left": "issue:28282", "right": "issue:30333", "accept": false, - "reason": "AutoModel/PyTorch import failure is unrelated to MLflow reporting on failed training." + "reason": "PyTorch import failure and MLflow job-reporting failure are unrelated." }, { "left": "issue:30333", "right": "issue:35141", "accept": false, - "reason": "MLflow reporting and embedding reinitialization during resize are unrelated." + "reason": "Training-job reporting bug vs token-embedding resize/post-init reinitialization bug." }, { "left": "issue:44297", "right": "issue:45310", "accept": false, - "reason": "Tokenizer class metadata mismatch on save is not the same as Qwen3.5 MoE from_pretrained failure." + "reason": "Tokenizer config save metadata mismatch is not the same bug as Qwen3.5 MoE from_pretrained failing." }, { "left": "issue:43723", "right": "issue:44977", "accept": false, - "reason": "Both involve loading, but one is tokenizer loading in v5 and the other is flash-attention generation behavior for Qwen3.5." + "reason": "Tokenizer loading regression and flash-attention generation failure are different symptoms in different subsystems." }, { "left": "issue:44222", "right": "issue:44484", "accept": false, - "reason": "FP8 save_pretrained for MoE is unrelated to max_shard_size default behavior." + "reason": "FP8 save_pretrained bug is unrelated to a question about the default `max_shard_size` constant." }, { "left": "issue:43232", "right": "issue:45834", "accept": false, - "reason": "Generation kwarg update after sync_gpus is unrelated to Kosmos2.5 OCR index overflow." + "reason": "Generation kwargs update after sync_gpus is unrelated to Kosmos2.5 OCR index errors." }, { "left": "issue:44263", "right": "issue:45092", "accept": false, - "reason": "torch.split return handling in one indexer is unrelated to remote-code checkpoint/meta-init incompatibility." + "reason": "Torch split/indexing issue in GLM MoE is unrelated to meta-initialization incompatibility for old InternVL2 checkpoints." }, { "left": "issue:43761", "right": "issue:46082", "accept": false, - "reason": "CLIPVisionModel hidden_states regression and LlamaConfig head_dim validation are different code paths." + "reason": "CLIP hidden-states regression and LlamaConfig head-dim validation are different model behaviors with different fixes." }, { "left": "issue:44483", "right": "issue:45406", "accept": false, - "reason": "OpenAI-compatible chat endpoint request parsing is unrelated to Gemma4Processor missing _tokenizer." + "reason": "Chat-completions request handling failure is not the same as Gemma4Processor missing `_tokenizer`." }, { "left": "issue:44295", "right": "issue:44623", "accept": false, - "reason": "Reading registered buffer position_ids and processor.save_pretrained missing files are separate issues." + "reason": "Reading registered `position_ids` as a buffer and missing processor save files are different save/load bugs." }, { "left": "issue:44373", "right": "issue:44485", "accept": false, - "reason": "Docstring mismatch for position_ids is not the same as GLM-5 RoPE implementation behavior." + "reason": "A wrong docstring is documentation-only; GLM-5 RoPE implementation is a code bug." }, { "left": "issue:44514", "right": "issue:45103", "accept": false, - "reason": "Batched chat template padding crash is unrelated to auto-docstring annotation handling." + "reason": "Batched `apply_chat_template` padding crash is unrelated to the auto-docstring annotation AttributeError." }, { "left": "issue:43723", "right": "issue:44871", "accept": false, - "reason": "Tokenizer loading regression and eos_token_id config inconsistency are distinct bugs." + "reason": "Generic tokenizer loading failure and inconsistent `eos_token_id` config are different tokenizer issues." }, { "left": "issue:44263", "right": "issue:44360", "accept": false, - "reason": "Both mention DSA indexer, but one is torch.split return handling and the other is a missing ReLU; not the same bug." + "reason": "Both mention DSA indexers, but one is a `torch.split` return-value problem and the other is a missing ReLU." }, { - "left": "issue:30990", - "right": "issue:38175", + "left": "issue:45092", + "right": "issue:45357", "accept": false, - "reason": "Sentence Transformers loading hang and SigLIP2 zero probabilities are unrelated." + "reason": "Remote-code/meta-init incompatibility is not the same as incorrect visual encoder keys being saved." }, { - "left": "issue:45092", - "right": "issue:45357", + "left": "issue:30990", + "right": "issue:38175", "accept": false, - "reason": "Both touch multimodal checkpoint compatibility, but the concrete failures differ: meta-init incompatibility vs incorrect visual encoder keys on save_pretrained." + "reason": "Sentence-transformers loading stall and SigLIP2 zero-probability classification are unrelated model behaviors." }, { "left": "issue:42907", "right": "issue:46050", "accept": false, - "reason": "Different quantization/save-load bugs with different mechanisms and model families." + "reason": "Both are quantization/save-load related, but they involve different quantization schemes and different failure modes." }, { "left": "issue:43723", "right": "issue:44991", "accept": false, - "reason": "Both are tokenizer-related, but one is a v5 loading regression and the other is a specific model tokenizer failure." + "reason": "Both are tokenizer load failures, but they are model-specific reports with no evidence of the same underlying root cause." }, { "left": "issue:43723", "right": "issue:44442", "accept": false, - "reason": "Different tokenizer loading failures for different models; no shared root cause shown." - }, - { - "left": "issue:44829", - "right": "issue:46002", - "accept": false, - "reason": "Flash-attention training degeneration and NaNs in classification heads are unrelated." + "reason": "FastSpeech2ConformerTokenizer load failure is a specific tokenizer issue, not clearly the same bug as the broader v5 tokenizer regression." }, { "left": "issue:44261", "right": "issue:45092", "accept": false, - "reason": "Layernorm epsilon config precision issue and meta-init checkpoint incompatibility are unrelated." + "reason": "MLA layernorm epsilon precision issue is unrelated to remote-code checkpoint meta-initialization incompatibility." }, { - "left": "issue:42831", - "right": "issue:45310", + "left": "issue:44829", + "right": "issue:46002", "accept": false, - "reason": "FineGrainedFP8 accuracy issue and Qwen3.5 MoE from_pretrained failure do not describe the same bug." + "reason": "Flash-attention training degeneration is not the same underlying problem as NaNs during checkpoint/init." }, { "left": "issue:41628", "right": "issue:43824", "accept": false, - "reason": "Missing AutoImageProcessor import and missing Qwen2_5_VLForConditionalGeneration import are different export gaps." + "reason": "Missing `AutoImageProcessor` import and missing `Qwen2_5_VLForConditionalGeneration` import are different symbol-resolution problems." + }, + { + "left": "issue:42831", + "right": "issue:45310", + "accept": false, + "reason": "FineGrainedFP8 accuracy issues are unrelated to Qwen3.5 MoE from_pretrained errors." }, { "left": "issue:43479", "right": "issue:44913", "accept": false, - "reason": "Phi4Multimodal config default-init behavior and GPTNeoX rotary_pct reload loss are separate config bugs." + "reason": "Default config initialization from `None` is unrelated to `rotary_pct` silently reverting on reload." }, { "left": "issue:42831", "right": "issue:44038", "accept": false, - "reason": "FP8 accuracy concerns and Qwen3-VL-Moe bug are unrelated." + "reason": "FP8 accuracy regression and Qwen3-VL-Moe v5 bug are not the same change." }, { "left": "issue:44263", "right": "issue:44373", "accept": false, - "reason": "Runtime tensor split bug vs docstring typo are not the same underlying issue." + "reason": "Code bug in DSA indexer vs documentation issue for `position_ids`; no shared underlying defect." + }, + { + "left": "issue:39401", + "right": "issue:44568", + "accept": false, + "reason": "Qwen3 tokenizer offset mapping bug and mdeberta BOS/EOS insertion bug are distinct tokenizer behaviors." } ] }, @@ -12477,7 +12445,7 @@ "component_index": 1, "component_count": 17, "review_unit_index": 40, - "review_unit_count": 50, + "review_unit_count": 49, "status": "reviewed", "reason": null, "source": "llm", @@ -12520,10 +12488,11 @@ "issue:44488", "issue:44514", "issue:44521", - "issue:44568", + "issue:44625", "issue:44671", "issue:44704", "issue:44743", + "issue:44829", "issue:44945", "issue:45278", "issue:45290", @@ -12534,13 +12503,12 @@ "issue:46132" ], "soft_pairs": [ - "issue:39401|issue:44568", "issue:42757|issue:45278", "issue:42886|issue:42994", "issue:43232|issue:45593", - "issue:42886|issue:43066", "issue:43756|issue:44060", "issue:42831|issue:45753", + "issue:42886|issue:43066", "issue:22355|issue:28218", "issue:44671|issue:46132", "issue:43122|issue:44521", @@ -12553,8 +12521,8 @@ "issue:43232|issue:44743", "issue:44704|issue:45290", "issue:42907|issue:44945", - "issue:41628|issue:45278", "issue:43122|issue:44514", + "issue:41628|issue:45278", "issue:39401|issue:44488", "issue:43784|issue:45600", "issue:43976|issue:44945", @@ -12562,210 +12530,211 @@ "issue:43408|issue:43475", "issue:42907|issue:43937", "issue:39401|issue:44521", - "issue:44360|issue:44485" + "issue:44360|issue:44485", + "issue:44625|issue:44829" ], - "prepared_review_unit_hash": "c17e97f1d88ec1276c3b1f2938717c7f90a2fb5e67b52cf739d3d85bc7eb107e", + "prepared_review_unit_hash": "46c44fd348fbf681721a9a3cf6292d252171980cf9b96853cb63144ce55022af", "component_budget": { "node_count": 426, "item_count": 426, - "soft_pair_count": 1572, - "serialized_chars": 814130, - "estimated_input_tokens": 203533, - "estimated_eval_tokens": 407322 + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 }, "budget": { - "node_count": 47, - "item_count": 47, + "node_count": 48, + "item_count": 48, "soft_pair_count": 29, - "serialized_chars": 35241, - "estimated_input_tokens": 8811, - "estimated_eval_tokens": 17878 + "serialized_chars": 35827, + "estimated_input_tokens": 8957, + "estimated_eval_tokens": 18170 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a loose set of unrelated Transformers issues that only share broad subsystem keywords (tokenizers, imports, generation, SAM3, etc.). I do not see any true duplicate pair among the soft edges. There are no PRs in the cluster, so issue selection is the only meaningful canonical choice.", - "confidence": 0.88, - "canonical_issue_reason": "issue:45278 is the broadest issue in the set and best represents the recurring version-upgrade import regressions, but it is still just a general umbrella rather than a true duplicate of the others.", + "summary": "This cluster is heterogeneous: most items share only broad subsystem words (imports, tokenizers, multimodal, SAM3) but not the same concrete failure mode. I reject all soft edges; none look like a single fixable duplicate pair. If a hub is needed, the broadest umbrella is the import-regression report 45278.", + "confidence": 0.41, + "canonical_issue_reason": "issue:45278 is the broadest, most umbrella-like report here, covering a version-upgrade import-regression pattern that could subsume related import failures.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45278 is the most suitable global anchor because it captures the widest class of breakages in this cluster (import failures after upgrading Transformers), making it the closest thing to a canonical issue here.", + "best_issue_reason": "issue:45278 is the most suitable triage anchor because it is broad, regression-oriented, and discussion-friendly; the other issues are too specific and mostly unrelated.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:39401", - "right": "issue:44568", - "accept": false, - "reason": "Both are tokenizer bugs, but one is wrong offset_mapping for Qwen3 and the other is missing BOS/EOS insertion for mdeberta-v3-base; different code paths and symptoms." - }, { "left": "issue:42757", "right": "issue:45278", "accept": false, - "reason": "Both mention import failures, but one is a specific huggingface_hub import and the other is a broad Transformers upgrade regression; not the same underlying bug." + "reason": "Both mention import failures, but 42757 is a specific missing symbol from huggingface_hub while 45278 is a broad multi-import regression after upgrade; not the same concrete bug." }, { "left": "issue:42886", "right": "issue:42994", "accept": false, - "reason": "Offline tokenizer cache loading and quantized model saving are unrelated subsystems and failure modes." + "reason": "Offline tokenizer cache loading and quantized model saving are unrelated code paths." }, { "left": "issue:43232", "right": "issue:45593", "accept": false, - "reason": "Generation kwargs handling after sync_gpus is unrelated to D-FINE auxiliary loss behavior." - }, - { - "left": "issue:42886", - "right": "issue:43066", - "accept": false, - "reason": "Tokenizer cache/offline loading and tokenizer decoder type in v5 are distinct bugs." + "reason": "Generation-kwargs handling after sync_gpus is unrelated to D-FINE auxiliary loss behavior." }, { "left": "issue:43756", "right": "issue:44060", "accept": false, - "reason": "SMOLLM3 RoPE layer count and Qwen3-Next tied-weights warning involve different models and different implementation issues." + "reason": "Different Qwen3-related model bugs: RoPE layer dropping versus tied-weights warning/misassignment." }, { "left": "issue:42831", "right": "issue:45753", "accept": false, - "reason": "FineGrainedFP8 accuracy loss and Qwen3_5 infinite-loop behavior are not the same defect." + "reason": "FineGrainedFP8 accuracy drift and Qwen3_5 image-triggered infinite loop are unrelated symptoms and subsystems." + }, + { + "left": "issue:42886", + "right": "issue:43066", + "accept": false, + "reason": "Both are tokenizer-related, but one is offline cache loading and the other is a decoder-type regression in v5; not the same bug." }, { "left": "issue:22355", "right": "issue:28218", "accept": false, - "reason": "Missing transformers.onnx module import and extra tokenizer spacing after added tokens are unrelated." + "reason": "Missing transformers.onnx module import and added-token spacing are unrelated." }, { "left": "issue:44671", "right": "issue:46132", "accept": false, - "reason": "CamemBERT masked LM prediction regressions and AttentionInterface.register behavior changes are different code-path problems." + "reason": "CamemBERT MLM output regression and AttentionInterface.register behavior change affect different components and failures." }, { "left": "issue:43122", "right": "issue:44521", "accept": false, - "reason": "Tokenization differences across versions and multimodal assistant mask handling are both tokenizer-related but not the same bug." + "reason": "Tokenization differs across versions, but 44521 is a multimodal assistant-mask bug; no evidence they share one fix." }, { "left": "issue:43576", "right": "issue:43784", "accept": false, - "reason": "Broken transformers env CLI and a NameError during sentence-transformers import are separate import/CLI regressions." + "reason": "A broken env command and a NameError importing sentence-transformers are different regressions." }, { "left": "issue:43646", "right": "issue:46002", "accept": false, - "reason": "Custom model initialization breakage and NaNs in classification heads are different failure modes." + "reason": "Custom model initialization breakage and NaNs in classification heads are not the same code-path problem." }, { "left": "issue:44117", "right": "issue:44704", "accept": false, - "reason": "Tokenizer mapping fallback bug and AutoProcessor cached_file kwargs forwarding are unrelated." + "reason": "Tokenizer mapping fallback logic and AutoProcessor cached_file kwargs handling are separate issues." }, { "left": "issue:30990", "right": "issue:33357", "accept": false, - "reason": "Sentence Transformers loading hang and MacOS bus error with a CLIP model are distinct issues." + "reason": "Sentence-Transformers loading hang and a MacOS bus error with CLIP are different failure modes." }, { "left": "issue:42371", "right": "issue:44263", "accept": false, - "reason": "TF32 configuration API changes and a torch.split return-value issue in GlmMoeDsaIndexer are unrelated." + "reason": "TF32 configuration guidance and GlmMoeDsaIndexer torch.split output handling are unrelated." }, { "left": "issue:44295", "right": "issue:44297", "accept": false, - "reason": "Position_ids buffer access and tokenizer_class mismatch on save_pretrained are different tokenizer/model serialization bugs." + "reason": "Reading a bufferized position_ids and mismatched tokenizer_class metadata are different bugs." }, { "left": "issue:43232", "right": "issue:44743", "accept": false, - "reason": "Generation state update after sync_gpus and recurrent state reset with cache are separate generation bugs." + "reason": "Both touch generation/cache logic, but one is sync_gpus kwargs propagation and the other is recurrent state reset; not mergeable as one fix." }, { "left": "issue:44704", "right": "issue:45290", "accept": false, - "reason": "AutoProcessor kwargs forwarding and chat-template crashes with tool-call assistant messages are different call paths." + "reason": "Cached-file kwargs forwarding and apply_chat_template crashing on tool-call assistant messages are separate API bugs." }, { "left": "issue:42907", "right": "issue:44945", "accept": false, - "reason": "Quantized/dequantized model saving and incorrect pipeline-parallel LLM output are unrelated." + "reason": "Quantized model saving failure and incorrect pipeline-parallel LLM output are unrelated." }, { - "left": "issue:41628", - "right": "issue:45278", + "left": "issue:43122", + "right": "issue:44514", "accept": false, - "reason": "Missing AutoImageProcessor export is one specific import regression, while 45278 is a broader bundle of import errors; related theme, but not the same bug." + "reason": "Version-dependent tokenization changes and batched chat-template crash with padding=False are distinct issues." }, { - "left": "issue:43122", - "right": "issue:44514", + "left": "issue:41628", + "right": "issue:45278", "accept": false, - "reason": "Version-dependent tokenization differences and batched apply_chat_template crashes are different multimodal/tokenizer issues." + "reason": "Both are import-related, but 41628 is a specific missing AutoImageProcessor export while 45278 is a broad upgrade-wide import regression; too coarse to treat as one bug." }, { "left": "issue:39401", "right": "issue:44488", "accept": false, - "reason": "Qwen3 offset_mapping is unrelated to failing to load cjvt/sleng-bert." + "reason": "Wrong offset_mapping and failure to load a specific model are unrelated tokenizer/model-loading problems." }, { "left": "issue:43784", "right": "issue:45600", "accept": false, - "reason": "sentence-transformers import NameError and removed Sam3LiteText config references are unrelated." + "reason": "sentence-transformers import NameError and stale Sam3LiteText references in auto_mappings.py are different import regressions." }, { "left": "issue:43976", "right": "issue:44945", "accept": false, - "reason": "Python version compatibility failure and pipeline parallelism output corruption are different problems." + "reason": "Python-version compatibility breakage and pipeline-parallel output corruption are unrelated." }, { "left": "issue:44261", "right": "issue:44373", "accept": false, - "reason": "MLA layernorm epsilon precision issue and a wrong docstring for position_ids are not duplicates." + "reason": "Missing rms_norm_eps affecting precision and a wrong docstring for position_ids are not the same underlying bug." }, { "left": "issue:43408", "right": "issue:43475", "accept": false, - "reason": "Wrong model-type warning and missing fpn_position_embeddings are both SAM3-related but involve different breakages." + "reason": "Both are SAM3-related, but one is a model-type mismatch warning and the other is a missing attribute on an output object; separate fixes." }, { "left": "issue:42907", "right": "issue:43937", "accept": false, - "reason": "Failed dequantized model saving and invalid GenerationConfig for GLM-5 are unrelated." + "reason": "Dequantized model saving and invalid GenerationConfig are different parts of the stack." }, { "left": "issue:39401", "right": "issue:44521", "accept": false, - "reason": "Wrong tokenizer offset mapping and all-zero assistant masks for multimodal inputs are different tokenizer/template bugs." + "reason": "Tokenizer offset mapping for Qwen3 and multimodal assistant-mask generation are unrelated." }, { "left": "issue:44360", "right": "issue:44485", "accept": false, - "reason": "Missing ReLU in the DSA indexer and GLM-5 RoPE implementation are unrelated model internals." + "reason": "DSA indexer activation shape/logic and GLM-5 RoPE implementation are separate model internals." + }, + { + "left": "issue:44625", + "right": "issue:44829", + "accept": false, + "reason": "num_labels propagation in Qwen3.5 config and flash_attention_3 training degeneration are different bugs." } ] }, @@ -12781,7 +12750,7 @@ "component_index": 1, "component_count": 17, "review_unit_index": 41, - "review_unit_count": 50, + "review_unit_count": 49, "status": "reviewed", "reason": null, "source": "llm", @@ -12816,15 +12785,15 @@ "issue:43976", "issue:43992", "issue:44038", + "issue:44075", "issue:44222", "issue:44295", "issue:44297", "issue:44351", "issue:44485", - "issue:44625", "issue:44704", "issue:44779", - "issue:44829", + "issue:44936", "issue:44945", "issue:44964", "issue:44977", @@ -12838,7 +12807,6 @@ "issue:46002" ], "soft_pairs": [ - "issue:44625|issue:44829", "issue:43116|issue:45200", "issue:42907|issue:44964", "issue:44038|issue:44297", @@ -12866,180 +12834,175 @@ "issue:43976|issue:45446", "issue:42222|issue:43441", "issue:43116|issue:43493", - "issue:42371|issue:43874" + "issue:42371|issue:43874", + "issue:44075|issue:44936" ], - "prepared_review_unit_hash": "0cf75fc8f549bde9195c17279c29fb5c92c2f81055386e10567cfb223e15a2df", + "prepared_review_unit_hash": "3691502de7ea9dab9b6aac684994ba28b3c0e2a28ec46c1c9b85910c0020aca4", "component_budget": { "node_count": 426, "item_count": 426, - "soft_pair_count": 1572, - "serialized_chars": 814130, - "estimated_input_tokens": 203533, - "estimated_eval_tokens": 407322 + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 }, "budget": { "node_count": 47, "item_count": 47, "soft_pair_count": 29, - "serialized_chars": 35407, - "estimated_input_tokens": 8852, - "estimated_eval_tokens": 17960 + "serialized_chars": 35277, + "estimated_input_tokens": 8820, + "estimated_eval_tokens": 17896 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly heterogeneous and does not look like a true duplicate set. The only loose theme is tokenizer/version-regression reports, so I\u2019d use the broad tokenization regression issue as the canonical anchor and reject all soft-edge merges.", - "confidence": 0.71, - "canonical_issue_reason": "issue:43122 is the broadest and most general tokenizer-regression report in the set, making it the safest duplicate anchor among several narrower tokenizer bugs.", + "summary": "This cluster is a loose mix of unrelated Transformers issues; none of the suggested soft pairs are strong enough to treat as true duplicates. The most representative issue is the broad tokenizer regression report around the v4\u2192v5 behavior change.", + "confidence": 0.34, + "canonical_issue_reason": "issue:43122 is the clearest, most general regression report in the set: it has a concrete version boundary (4.57.3 \u2192 5.0), a reproducible tokenizer behavior change, and broader relevance than the model-specific reports.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43122 is the best overall representative because it describes a general cross-version tokenizer behavior change rather than a model-specific or API-specific symptom.", + "best_issue_reason": "issue:43122 is the best representative issue because it describes a concrete, user-facing tokenizer regression with a clear before/after comparison and is broad enough to summarize the tokenizer-related subset of the cluster.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:44625", - "right": "issue:44829", - "accept": false, - "reason": "Different bugs: one is Qwen3.5 config propagation, the other is flash-attention training degeneration." - }, { "left": "issue:43116", "right": "issue:45200", "accept": false, - "reason": "Different subsystems and symptoms: example-script multi-label output vs Gemma 4 multimodal token defaults." + "reason": "Different problems: example script multi-label output vs Gemma 4 multimodal token-type handling. No shared code path." }, { "left": "issue:42907", "right": "issue:44964", "accept": false, - "reason": "One is a save_pretrained/dequantization problem; the other is a model loading failure for a different model family." + "reason": "Saving dequantized Ministral/Devstral weights is unrelated to loading Phi-4 multimodal with latest transformers." }, { "left": "issue:44038", "right": "issue:44297", "accept": false, - "reason": "Qwen3-VL-Moe bug versus tokenizer save_pretrained metadata mismatch; not the same underlying defect." + "reason": "Qwen3-VL-Moe bug is not the same as tokenizer save_pretrained metadata mismatch." }, { "left": "issue:28218", "right": "issue:30990", "accept": false, - "reason": "Tokenizer spacing behavior and sentence-transformers loading hang are unrelated." + "reason": "Tokenizer added-space bug and Sentence Transformers loading hang are unrelated." }, { "left": "issue:44038", "right": "issue:44295", "accept": false, - "reason": "Different code paths: multimodal Qwen3-VL-Moe issue versus position_ids buffer read error." + "reason": "Different symptoms and code paths: Qwen3-VL-Moe failure vs position_ids buffer read error." }, { "left": "issue:43122", "right": "issue:44779", "accept": false, - "reason": "Both are tokenizer regressions, but the titles point to different models and failure modes; not enough to treat as one bug." + "reason": "Both are tokenizer regressions in v5, but one is generic tokenization drift and the other is Deepseek-specific incorrect outputs; not enough evidence of the same root cause." }, { "left": "issue:42907", "right": "issue:43452", "accept": false, - "reason": "Saving dequantized Ministral/Devstral models is unrelated to gguf_file loading breakage." + "reason": "Dequantized save failure and gguf_file loading failure are different loading/saving paths and different bugs." }, { "left": "issue:36296", "right": "issue:43116", "accept": false, - "reason": "Tensor-parallel training bug and example-script classification output issue are different problems." + "reason": "Tensor-parallel training bug is unrelated to the classification example script issue." }, { "left": "issue:43824", "right": "issue:44351", "accept": false, - "reason": "Separate import failures for different symbols and packages." + "reason": "Both are import errors, but for different symbols and different underlying missing exports." }, { "left": "issue:44038", "right": "issue:44222", "accept": false, - "reason": "Qwen3-VL-Moe issue versus FP8 save_pretrained issue; different model behavior." + "reason": "Different Qwen3-VL-Moe inference bug vs FP8 save_pretrained-on-MoE bug." }, { "left": "issue:45216", "right": "issue:45478", "accept": false, - "reason": "Both involve Qwen3.5 save/load regressions, but the reported failures are not specific enough to confirm one shared concrete bug." + "reason": "Same model family, but save_pretrained checkpoint corruption and from_pretrained loading error are distinct code paths and not clearly the same defect." }, { "left": "issue:42757", "right": "issue:44351", "accept": false, - "reason": "Different missing imports from different modules." + "reason": "Different import failures from different libraries/modules; not a duplicate." }, { "left": "issue:43278", "right": "issue:44945", "accept": false, - "reason": "Dtype mismatch in eval vs incorrect output under pipeline parallelism are different symptoms and likely different bugs." + "reason": "BF16-to-FP32 dtype shift in eval is unrelated to incorrect outputs under pipeline parallelism." }, { "left": "issue:43784", "right": "issue:45446", "accept": false, - "reason": "Sentence-transformers import NameError and flex_attention version-check bug do not share the same code path." + "reason": "Sentence-transformers import NameError and flex_attention version-check bug are unrelated." }, { "left": "issue:22355", "right": "issue:29159", "accept": false, - "reason": "ONNX import failure and slow/fast tokenizer inconsistency are unrelated." + "reason": "Missing transformers.onnx module and slow/fast tokenizer inconsistency are separate issues." }, { "left": "issue:43122", "right": "issue:44977", "accept": false, - "reason": "Tokenizer regression versus flash-attention generation bug; different failure modes." + "reason": "Tokenizer regression and flash-attention generation failure are different symptoms and likely different paths." }, { "left": "issue:43493", "right": "issue:45753", "accept": false, - "reason": "SigLIP2 implementation discrepancy and Qwen3_5 infinite-loop-on-image are not the same bug." + "reason": "SigLIP2 implementation discrepancy is unrelated to Qwen3_5 image-induced infinite loop." }, { "left": "issue:42673", "right": "issue:45072", "accept": false, - "reason": "VRAM leak in multi-threaded Qwen3ForCausalLM usage is unrelated to dtype mismatch in inference." + "reason": "VRAM leak in multi-threaded Qwen3ForCausalLM use is unrelated to dtype mismatch in bfloat16 inference." }, { "left": "issue:43066", "right": "issue:44779", "accept": false, - "reason": "Both mention tokenizer issues in v5, but one is decoder-type metadata and the other is model-specific tokenization output." + "reason": "Wrong tokenizer decoder type and Deepseek tokenization regression are related only at a high level; not the same concrete bug." }, { "left": "issue:38617", "right": "issue:45278", "accept": false, - "reason": "A specific missing import and a broad upgrade-related import breakage are too different to merge." + "reason": "Specific missing import versus broad import errors after upgrade; not the same issue." }, { "left": "issue:43992", "right": "issue:44704", "accept": false, - "reason": "Missing embed_tokens.weight in UMT5Encoder and AutoProcessor kwargs passthrough are unrelated loading issues." + "reason": "Missing embed_tokens.weight in UMT5Encoder and AutoProcessor kwargs forwarding are unrelated." }, { "left": "issue:22355", "right": "issue:29127", "accept": false, - "reason": "Importing transformers.onnx and LayoutLMv3 error-message clarity are unrelated." + "reason": "transformers.onnx import failure and LayoutLMv3 error-message clarity are unrelated." }, { "left": "issue:43937", "right": "issue:44485", "accept": false, - "reason": "GenerationConfig validation failure and RoPE implementation discussion are distinct issues." + "reason": "Invalid GenerationConfig and GLM-5 RoPE discussion do not describe the same defect." }, { "left": "issue:43611", @@ -13051,25 +13014,31 @@ "left": "issue:43976", "right": "issue:45446", "accept": false, - "reason": "Python version compatibility and a PyTorch version check bug are unrelated." + "reason": "Python version support issue and flex_attention version check bug are unrelated." }, { "left": "issue:42222", "right": "issue:43441", "accept": false, - "reason": "VitPose breakage and Ministral flash-attention failure affect different models and code paths." + "reason": "VitPose model breakage and Ministral-3 FlashAttention failure are different model-specific bugs." }, { "left": "issue:43116", "right": "issue:43493", "accept": false, - "reason": "Example-script classification bug versus SigLIP2 model discrepancy are unrelated." + "reason": "Classification example script issue and SigLIP2 implementation discrepancy are unrelated." }, { "left": "issue:42371", "right": "issue:43874", "accept": false, - "reason": "TF32 API settings warning and missing multimodal patch-count method are different issues." + "reason": "TF32 API guidance and missing image patch count method are unrelated." + }, + { + "left": "issue:44075", + "right": "issue:44936", + "accept": false, + "reason": "Unused SGD args and trainer.evaluate() failing after train() are different training-loop bugs." } ] }, @@ -13085,7 +13054,7 @@ "component_index": 1, "component_count": 17, "review_unit_index": 42, - "review_unit_count": 50, + "review_unit_count": 49, "status": "reviewed", "reason": null, "source": "llm", @@ -13103,17 +13072,18 @@ "issue:42673", "issue:43064", "issue:43116", + "issue:43408", "issue:43421", "issue:43450", "issue:43582", "issue:43673", + "issue:43742", "issue:43825", "issue:43827", "issue:43906", "issue:43937", "issue:43986", "issue:44038", - "issue:44075", "issue:44117", "issue:44261", "issue:44291", @@ -13123,7 +13093,6 @@ "issue:44484", "issue:44779", "issue:44829", - "issue:44936", "issue:44977", "issue:45071", "issue:45081", @@ -13142,7 +13111,6 @@ "issue:46077" ], "soft_pairs": [ - "issue:44075|issue:44936", "issue:43064|issue:43673", "issue:44038|issue:45478", "issue:45357|issue:45478", @@ -13151,14 +13119,14 @@ "issue:43937|issue:45563", "issue:44380|issue:46077", "issue:33453|issue:39401", - "issue:43827|issue:45446", "issue:39401|issue:44779", + "issue:43827|issue:45446", "issue:44117|issue:45588", "issue:42673|issue:43906", "issue:43937|issue:44977", "issue:44291|issue:45071", - "issue:44466|issue:46002", "issue:43450|issue:43986", + "issue:44466|issue:46002", "issue:45405|issue:45446", "issue:45341|issue:45561", "issue:44117|issue:45081", @@ -13169,211 +13137,212 @@ "issue:42673|issue:43582", "issue:42617|issue:45341", "issue:44484|issue:45310", + "issue:42371|issue:44261", "issue:42490|issue:42673", - "issue:42371|issue:44261" + "issue:43408|issue:43742" ], - "prepared_review_unit_hash": "a70ded6d29f573cedc1f3dbc16ed8457805c8ef01ea003aaa3d7b062d5df61d3", + "prepared_review_unit_hash": "934fba96391f7caf15c0c8d523d2fcfb183bb63406167fd4f44c255349c1cd46", "component_budget": { "node_count": 426, "item_count": 426, - "soft_pair_count": 1572, - "serialized_chars": 814130, - "estimated_input_tokens": 203533, - "estimated_eval_tokens": 407322 + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 }, "budget": { "node_count": 47, "item_count": 47, "soft_pair_count": 29, - "serialized_chars": 35789, - "estimated_input_tokens": 8948, - "estimated_eval_tokens": 18152 + "serialized_chars": 35850, + "estimated_input_tokens": 8963, + "estimated_eval_tokens": 18182 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "Mostly a heterogeneous mix of unrelated Transformers issues; there are no PRs here. The only clear duplicate anchor is the Qwen3.5 \"from_pretrained\" report (45310/45478), while the provided soft edges mostly connect distinct bugs in adjacent subsystems.", - "confidence": 0.84, - "canonical_issue_reason": "Issue 45310 is the strongest canonical anchor because it is the earlier report for the exact same Qwen3.5 Moe \"from_pretrained\" error later re-reported as 45478, giving the clearest duplicate relationship in this set.", + "summary": "This is not a tight duplicate cluster; it mixes many unrelated bugs across tokenizers, Qwen/Qwen3.5, training, imports, docs, and testing. I would not merge any of the proposed soft edges.", + "confidence": 0.61, + "canonical_issue_reason": "No single true canonical exists here because the items are heterogeneous; if a placeholder is required, 43825 is the cleanest user-facing runtime bug and has some discussion/inbound references.", "canonical_pr_reason": null, - "best_issue_reason": "45310 is the best global representative: it has the clearest duplication signal and sits in the most coherent recurring bug thread, whereas the rest of the cluster is largely unrelated one-off issues.", + "best_issue_reason": "43825 is the most concrete and broadly understandable issue in the set, but it still does not represent the rest of the unrelated cluster.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:44075", - "right": "issue:44936", - "accept": false, - "reason": "Different code paths: optimizer argument handling vs trainer evaluate-after-train failure." - }, { "left": "issue:43064", "right": "issue:43673", "accept": false, - "reason": "Unrelated bugs: FSDP/PEFT optimizer state mismatch vs generation cache missing during chunked prefill." + "reason": "FSDP/PEFT optimizer-state corruption is a different bug than missing GenerationMixin cache during chunked prefill." }, { "left": "issue:44038", "right": "issue:45478", "accept": false, - "reason": "Both are Qwen/Moe-related, but they describe different failures and model variants, not the same concrete bug." + "reason": "Both mention Qwen3.5-era model loading, but one is a generic Qwen3-VL-Moe v5 issue and the other is a specific from_pretrained failure; not the same bug." }, { "left": "issue:45357", "right": "issue:45478", "accept": false, - "reason": "Save_pretrained checkpoint-key regression vs from_pretrained load error; related model family, but different bug." + "reason": "One is incorrect save_pretrained visual encoder keys; the other is a from_pretrained load error. Different code paths and symptoms." }, { "left": "issue:43116", "right": "issue:43825", "accept": false, - "reason": "Different surfaces: example script multi-label output vs pipeline() v5 error message/docs behavior." + "reason": "Multi-label classification example behavior is unrelated to the pipeline translation-task message regression." }, { "left": "issue:45083", "right": "issue:45446", "accept": false, - "reason": "Different subsystems: Qwen3 Omni feature extraction helper vs flex_attention import/version check." + "reason": "qwen3_omni_moe feature-length behavior and a flex_attention PyTorch version check are unrelated." }, { "left": "issue:43937", "right": "issue:45563", "accept": false, - "reason": "GenerationConfig validation error vs stale warning in paged generate; not the same underlying issue." + "reason": "Invalid GenerationConfig handling is not the same as a stale num_return_sequences warning in paged generate." }, { "left": "issue:44380", "right": "issue:46077", "accept": false, - "reason": "Both involve attention behavior, but one is GPT2 scaling under backends and the other is Gemma4 mask construction." + "reason": "GPT-2 attention scaling under SDPA/FlashAttention and Gemma4 bidirectional mask construction are different model/backend bugs." }, { "left": "issue:33453", "right": "issue:39401", "accept": false, - "reason": "Both are tokenizer bugs, but the symptoms and affected paths differ (loading regression vs offset_mapping)." + "reason": "Tokenizer loading regression and Qwen3 offset_mapping are both tokenizer-related but are different failures with different models/symptoms." }, { - "left": "issue:43827", - "right": "issue:45446", + "left": "issue:39401", + "right": "issue:44779", "accept": false, - "reason": "Docs/reference cleanup vs a runtime PyTorch version check bug." + "reason": "Different tokenizer bugs affecting different models: offset_mapping vs incorrect DeepSeek tokenization results." }, { - "left": "issue:39401", - "right": "issue:44779", + "left": "issue:43827", + "right": "issue:45446", "accept": false, - "reason": "Different tokenizer correctness bugs in different model families; too broad to merge." + "reason": "Docs still referencing pipeline() is unrelated to the flex_attention import/version check." }, { "left": "issue:44117", "right": "issue:45588", "accept": false, - "reason": "Tokenizer mapping fallback bug vs flash_attention crash on s_aux=None; unrelated code paths." + "reason": "TOKENIZER_MAPPING_NAMES returning None is unrelated to flash_attention crash on sink-less models." }, { "left": "issue:42673", "right": "issue:43906", "accept": false, - "reason": "VRAM leak in multi-dataloader-thread use vs an isolated reproduction of another issue; no evidence they are the same bug." + "reason": "VRAM leak across dataloader threads is not the same as an isolated reproduction of a different issue." }, { "left": "issue:43937", "right": "issue:44977", "accept": false, - "reason": "GenerationConfig invalid vs flash-attention generation failure; different model/config paths." + "reason": "Generic invalid generation config handling is different from Qwen3.5 flash-attention generation failure." }, { "left": "issue:44291", "right": "issue:45071", "accept": false, - "reason": "init_empty_weights argument incompatibility vs PretrainedConfig type-checking regression; not the same defect." + "reason": "Unexpected _is_hf_initialized argument in init_empty_weights is unrelated to v5.4.0 PretrainedConfig type checking." }, { - "left": "issue:44466", - "right": "issue:46002", + "left": "issue:43450", + "right": "issue:43986", "accept": false, - "reason": "Serialization inconsistency for tied weights vs NaNs in classification heads; unrelated symptoms and code paths." + "reason": "Batched video processor shape bug and missing torchvision import crash are separate video-processing problems." }, { - "left": "issue:43450", - "right": "issue:43986", + "left": "issue:44466", + "right": "issue:46002", "accept": false, - "reason": "Batched video shape bug vs missing torchvision crash during AutoProcessor load; related area, different failure mode." + "reason": "lm_head.weight serialization inconsistency is unrelated to NaNs in classification heads after checkout/init." }, { "left": "issue:45405", "right": "issue:45446", "accept": false, - "reason": "PEFT minimum-version packaging issue vs flex_attention import guard bug; unrelated." + "reason": "A PEFT version bump problem is unrelated to the flex_attention AuxRequest version check." }, { "left": "issue:45341", "right": "issue:45561", "accept": false, - "reason": "A small testing_utils bug vs a pytest-xdist race in captured_info.txt; not clearly the same underlying problem." + "reason": "A small testing_utils bug is not the same as the pytest-xdist captured_info.txt race." }, { "left": "issue:44117", "right": "issue:45081", "accept": false, - "reason": "Token mapping None handling vs Mistral regex patch crash; both tokenizer-related but different code paths." + "reason": "Tokenizer mapping returning None is a different failure from _patch_mistral_regex crashing on backend_tokenizer access." }, { "left": "issue:38617", "right": "issue:41628", "accept": false, - "reason": "Both are import errors, but for different symbols and modules; no duplicate evidence." + "reason": "Both are import errors, but they concern different missing symbols from different modules and code paths." }, { "left": "issue:44410", "right": "issue:45310", "accept": false, - "reason": "Missing layer projections in qwen3next vs Qwen3.5 from_pretrained error; different bugs." + "reason": "Missing qkv/gate projections in qwen3next is a model-architecture issue, not the same as Qwen3.5 Moe from_pretrained failure." }, { "left": "issue:44829", "right": "issue:45216", "accept": false, - "reason": "FlashAttention training degeneration vs Qwen3.5 save_pretrained regression; different subsystems and symptoms." + "reason": "FlashAttention-3 causing degenerate training is unrelated to Qwen3.5 save_pretrained checkpoint regression." }, { "left": "issue:39401", "right": "issue:43421", "accept": false, - "reason": "Tokenizer offset mapping bug vs runtime post-processor update feature request; not the same issue." + "reason": "Qwen3 tokenizer offset_mapping wrong results and runtime post-processor updates for special tokens are different tokenizer behaviors." }, { "left": "issue:42673", "right": "issue:43582", "accept": false, - "reason": "VRAM leak in threaded loading vs Apple Silicon TypeError in caching warmup; unrelated." + "reason": "A VRAM leak in multi-threaded loading is unrelated to an Apple Silicon TypeError in caching_allocator_warmup." }, { "left": "issue:42617", "right": "issue:45341", "accept": false, - "reason": "3d_parallel.py execution failure vs a minor testing_utils bug; no overlap in root cause." + "reason": "Failure to run 3d_parallel.py is unrelated to a minor testing_utils issue." }, { "left": "issue:44484", "right": "issue:45310", "accept": false, - "reason": "Shard-size question vs Qwen3.5 from_pretrained regression; unrelated." + "reason": "save_pretrained max_shard_size behavior is unrelated to Qwen3.5 Moe loading errors." + }, + { + "left": "issue:42371", + "right": "issue:44261", + "accept": false, + "reason": "TF32 behavior settings and MLA rms_norm_eps precision mismatch are unrelated issues." }, { "left": "issue:42490", "right": "issue:42673", "accept": false, - "reason": "Local-path import warning vs VRAM leak in dataloader threads; different bugs." + "reason": "Trailing-slash import warning when loading a local model is not the same as a VRAM leak across dataloader threads." }, { - "left": "issue:42371", - "right": "issue:44261", + "left": "issue:43408", + "right": "issue:43742", "accept": false, - "reason": "TF32 API guidance issue vs MLA q_a_layernorm precision/config bug; not the same code path." + "reason": "sam3_video/sam3_tracker warning is unrelated to a KeyError while loading facebook/MobileLLM-125M." } ] }, @@ -13389,7 +13358,7 @@ "component_index": 1, "component_count": 17, "review_unit_index": 43, - "review_unit_count": 50, + "review_unit_count": 49, "status": "reviewed", "reason": null, "source": "llm", @@ -13413,7 +13382,6 @@ "issue:43452", "issue:43577", "issue:43673", - "issue:43742", "issue:43874", "issue:43931", "issue:43937", @@ -13425,6 +13393,7 @@ "issue:44261", "issue:44263", "issue:44360", + "issue:44368", "issue:44410", "issue:44483", "issue:44484", @@ -13447,17 +13416,16 @@ "issue:46172" ], "soft_pairs": [ - "issue:43408|issue:43742", - "issue:42222|issue:45542", "issue:44933|issue:45278", "issue:38617|issue:44220", + "issue:42222|issue:45542", "issue:43937|issue:44496", "issue:42175|issue:42222", "issue:44483|issue:45464", "issue:43066|issue:43122", "issue:42371|issue:42831", - "issue:43012|issue:43577", "issue:44075|issue:45656", + "issue:43012|issue:43577", "issue:44410|issue:45478", "issue:43673|issue:45341", "issue:44484|issue:45478", @@ -13478,228 +13446,243 @@ "issue:43097|issue:43452", "issue:45003|issue:45341", "issue:43986|issue:45341", - "issue:44117|issue:44987" + "issue:44117|issue:44987", + "issue:43408|issue:44368", + "issue:43986|issue:44484", + "issue:43874|issue:44263" ], - "prepared_review_unit_hash": "fc939f48166c39983f54601495f405e9bb0b048529fa2f2d258ed4c4803b791b", + "prepared_review_unit_hash": "03800beb213eaf1f5d42c9a0cdd0e3db072ca85468a2dcceb8eb4bb0be1b4301", "component_budget": { "node_count": 426, "item_count": 426, - "soft_pair_count": 1572, - "serialized_chars": 814130, - "estimated_input_tokens": 203533, - "estimated_eval_tokens": 407322 + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 32, - "serialized_chars": 36870, - "estimated_input_tokens": 9218, - "estimated_eval_tokens": 18692 + "soft_pair_count": 34, + "serialized_chars": 37583, + "estimated_input_tokens": 9396, + "estimated_eval_tokens": 19048 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is heterogeneous: most items are distinct Transformers regressions spanning import errors, tokenizer changes, model-loading failures, precision warnings, and backend/dependency issues. I reject all soft edges as duplicates; none clearly point to the same concrete bug or change.", - "confidence": 0.31, - "canonical_issue_reason": "issue:45278 is the broadest representative of the cluster\u2019s repeated import/API breakage reports after upgrading Transformers, so it best serves as the umbrella issue even though the rest are mostly separate bugs.", + "summary": "These items mostly form a noisy similarity cluster: many are Transformers regressions, but the proposed edges usually connect different models, different code paths, or only broadly related symptom classes. I found no pair that clearly looks like the same underlying bug or change.", + "confidence": 0.82, + "canonical_issue_reason": "Issue 45278 is the broadest, most umbrella-like report in the set: a post-upgrade import-regression with multiple failures, making it the best representative issue among these mostly unrelated reports.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45278 is the most general and duplicate-prone report in the set; it captures the shared upgrade/regression theme better than the narrower model-specific or single-symbol import issues.", + "best_issue_reason": "Issue 45278 has the clearest generalized regression framing and the widest apparent scope, so it is the strongest single issue to keep as the cluster representative.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:43408", - "right": "issue:43742", - "accept": false, - "reason": "Different model families and failure modes: SAM3 model-type warning vs MobileLLM key error." - }, - { - "left": "issue:42222", - "right": "issue:45542", - "accept": false, - "reason": "VitPose model breakage is unrelated to the TensorBoard-without-TF backend error." - }, { "left": "issue:44933", "right": "issue:45278", "accept": false, - "reason": "One is a specific missing import from image_utils; the other is a broad upgrade-related import regression report." + "reason": "Both mention imports, but one is a specific missing import from image_utils while the other is a broad upgrade-wide import regression; not the same bug." }, { "left": "issue:38617", "right": "issue:44220", "accept": false, - "reason": "Missing configuration_utils export vs audio feature extraction bug; no shared code path." + "reason": "Different failures in different areas: a missing configuration utility symbol vs. a feature-extraction helper issue." + }, + { + "left": "issue:42222", + "right": "issue:45542", + "accept": false, + "reason": "One is about broken vitpose models, the other about TensorFlow backend detection when only tensorboard is installed." }, { "left": "issue:43937", "right": "issue:44496", "accept": false, - "reason": "GLM generation-config validation and unrecognized-model config loading are different bugs." + "reason": "Different model-loading/validation failures affecting unrelated models and error paths." }, { "left": "issue:42175", "right": "issue:42222", "accept": false, - "reason": "Packaging/dependency mismatch is not the same as a broken VitPose model implementation." + "reason": "TensorFlow installation/backend packaging problem vs. vitpose model breakage; too different." }, { "left": "issue:44483", "right": "issue:45464", "accept": false, - "reason": "Both touch chat APIs, but one is a request acceptance issue and the other is a streaming-inference failure on a specific model." + "reason": "Both touch chat/completions, but one is a generic API rejection and the other is a specific streaming-inference failure for a particular model; not clearly the same bug." }, { "left": "issue:43066", "right": "issue:43122", "accept": false, - "reason": "Related tokenizer regressions, but the concrete symptoms differ: decoder type mismatch vs changed tokenization output." + "reason": "Both are tokenizer regressions in v5, but one is decoder-type mismatch and the other is changed tokenization output; not enough to treat as the same defect." }, { "left": "issue:42371", "right": "issue:42831", "accept": false, - "reason": "TF32 configuration warning and FineGrainedFP8 accuracy loss are separate precision-related problems." + "reason": "TF32 API-setting guidance is unrelated to the FineGrainedFP8 accuracy regression." }, { - "left": "issue:43012", - "right": "issue:43577", + "left": "issue:44075", + "right": "issue:45656", "accept": false, - "reason": "Both mention dtype/precision, but one is a warning and the other is incorrect dtype propagation in Blip2 loading." + "reason": "SGD argument handling and DeepSpeed double-stepping are different optimizer bugs." }, { - "left": "issue:44075", - "right": "issue:45656", + "left": "issue:43012", + "right": "issue:43577", "accept": false, - "reason": "SGD argument handling and double optimizer stepping with deepspeed are different optimizer bugs." + "reason": "Both involve dtype behavior, but one is a PyTorch warning during compilation and the other is Blip2 load-time dtype propagation; different concrete problems." }, { "left": "issue:44410", "right": "issue:45478", "accept": false, - "reason": "Different Qwen-family issues with different symptoms: missing projections vs from_pretrained loading error." + "reason": "Different models and different symptoms: missing projections in qwen3next vs. Qwen3.5 MoE from_pretrained error." }, { "left": "issue:43673", "right": "issue:45341", "accept": false, - "reason": "Generation cache bug and a testing_utils bug are unrelated." + "reason": "Generation cache/chunked-prefill bug is unrelated to a testing_utils issue." }, { "left": "issue:44484", "right": "issue:45478", "accept": false, - "reason": "Sharding-size behavior is unrelated to Qwen3.5 MoE loading failures." + "reason": "save_pretrained sharding-size question is unrelated to Qwen3.5 MoE loading failure." }, { "left": "issue:43097", "right": "issue:44493", "accept": false, - "reason": "Both are v5 regressions, but one is an API removal and the other is unexpected position-id keys during model loading." + "reason": "Removed tie_embeddings behavior and unexpected position-id keys are different regression classes." }, { "left": "issue:41628", "right": "issue:45071", "accept": false, - "reason": "Importing AutoImageProcessor and PretrainedConfig type checking are unrelated breakages." + "reason": "AutoImageProcessor import failure and PretrainedConfig type-checking regression are distinct issues." }, { "left": "issue:43874", "right": "issue:44261", "accept": false, - "reason": "Different models and different root causes: missing image-patch helper vs missing rms_norm_eps config usage." + "reason": "Missing image-patch counting helper vs. rms_norm_eps precision issue; different models and paths." }, { "left": "issue:44263", "right": "issue:44485", "accept": false, - "reason": "Torch.split return handling and RoPE implementation are separate GLM internals." + "reason": "DSA indexer split behavior and GLM-5 RoPE implementation are not the same concrete bug." }, { "left": "issue:45071", "right": "issue:45542", "accept": false, - "reason": "Type-checking regression is not the same as the TensorBoard-without-TF backend error." + "reason": "Type-checking regression in config handling is unrelated to TensorFlow backend import/installation behavior." }, { "left": "issue:42222", "right": "issue:45459", "accept": false, - "reason": "VitPose model breakage and protobuf-related tokenizer error masking are unrelated." + "reason": "Vitpose model breakage and tokenizer error-hiding when protobuf is absent are unrelated." }, { "left": "issue:44263", "right": "issue:46172", "accept": false, - "reason": "Different models and code paths: GLM MoE indexing vs DeepSeek-V4 tensor transpose." + "reason": "Both are tensor/indexing-related, but the concrete code paths and symptoms differ." }, { "left": "issue:44360", "right": "issue:46172", "accept": false, - "reason": "A missing ReLU in the DSA indexer is not the same bug as DeepSeek-V4 dtype/transpose handling." + "reason": "ReLU missing in the DSA indexer and a dtype/transpose issue in deepseek-v4 are different bugs." }, { "left": "issue:43012", "right": "issue:43408", "accept": false, - "reason": "Precision-warning behavior and SAM3 model-type warning are unrelated." + "reason": "A bfloat16 precision warning is unrelated to the sam3_video/sam3_tracker model-type mismatch." }, { "left": "issue:43441", "right": "issue:44829", "accept": false, - "reason": "Ministral FlashAttention failure and flash_attention_3 degenerate training are different attention-path bugs." + "reason": "Both mention FlashAttention, but they affect different models and failure modes; not the same underlying change." }, { "left": "issue:43931", "right": "issue:44410", "accept": false, - "reason": "Weight-shape mismatch and missing projections are different model-loading/model-architecture issues." + "reason": "Qwen3-VL weight-shape mismatch and qwen3next missing projections are different model-specific bugs." }, { "left": "issue:43673", "right": "issue:46123", "accept": false, - "reason": "Generation cache handling and MaskGenerationPipeline batch-finalization are separate pipeline bugs." + "reason": "Cache handling during chunked prefill is unrelated to the final partial-batch masking pipeline bug." }, { "left": "issue:43931", "right": "issue:44038", "accept": false, - "reason": "Both involve Qwen3-VL, but one is a weight mismatch and the other is a MoE loading issue." + "reason": "Different Qwen3-VL-related failures with different symptoms and likely different fixes." }, { "left": "issue:43335", "right": "issue:45071", "accept": false, - "reason": "SwitchTransformers sparse-layer construction is unrelated to PretrainedConfig type checking." + "reason": "Sparse-layer construction bug and config type-checking regression are unrelated." }, { "left": "issue:43097", "right": "issue:43452", "accept": false, - "reason": "API removal of tie_embeddings_and_encoder_decoder is unrelated to gguf_file tokenizer/model loading failures." + "reason": "Embedding-tie removal and gguf_file loading failures are different code paths." }, { "left": "issue:45003", "right": "issue:45341", "accept": false, - "reason": "Unsafe sys.modules access and a testing_utils bug do not describe the same defect." + "reason": "Unsafe sys.modules access in modeling_utils is unrelated to a testing_utils bug." }, { "left": "issue:43986", "right": "issue:45341", "accept": false, - "reason": "AutoProcessor/torchvision dependency crash is unrelated to the testing_utils bug." + "reason": "AutoProcessor crashing without torchvision and a testing_utils issue are unrelated." }, { "left": "issue:44117", "right": "issue:44987", "accept": false, - "reason": "TOKENIZER_MAPPING_NAMES returning None is a different issue from loading physical-intelligence/fast." + "reason": "Tokenizer mapping None handling and a broad model-loading failure are too different to merge." + }, + { + "left": "issue:43408", + "right": "issue:44368", + "accept": false, + "reason": "sam3 model-type warning and a Qwen3.5 LoRA tie_word_embeddings warning are unrelated." + }, + { + "left": "issue:43986", + "right": "issue:44484", + "accept": false, + "reason": "Video-model processor crash and max_shard_size behavior are unrelated." + }, + { + "left": "issue:43874", + "right": "issue:44263", + "accept": false, + "reason": "Missing image-patch helper and torch.split return-value behavior are different implementation bugs." } ] }, @@ -13715,7 +13698,7 @@ "component_index": 1, "component_count": 17, "review_unit_index": 44, - "review_unit_count": 50, + "review_unit_count": 49, "status": "reviewed", "reason": null, "source": "llm", @@ -13732,7 +13715,6 @@ "issue:43097", "issue:43208", "issue:43352", - "issue:43408", "issue:43502", "issue:43531", "issue:43550", @@ -13740,7 +13722,6 @@ "issue:43824", "issue:43825", "issue:43873", - "issue:43874", "issue:43906", "issue:43976", "issue:43986", @@ -13750,14 +13731,15 @@ "issue:44117", "issue:44162", "issue:44261", - "issue:44263", "issue:44315", "issue:44351", "issue:44368", "issue:44380", "issue:44410", + "issue:44483", "issue:44484", "issue:44485", + "issue:44514", "issue:44655", "issue:44704", "issue:45103", @@ -13765,6 +13747,7 @@ "issue:45372", "issue:45375", "issue:45440", + "issue:45479", "issue:45588", "issue:45753", "issue:46050", @@ -13773,11 +13756,8 @@ "issue:46172" ], "soft_pairs": [ - "issue:43408|issue:44368", - "issue:43986|issue:44484", - "issue:43874|issue:44263", - "issue:43531|issue:45341", "issue:45440|issue:46172", + "issue:43531|issue:45341", "issue:43986|issue:45372", "issue:38617|issue:42757", "issue:44261|issue:44485", @@ -13805,234 +13785,243 @@ "issue:44162|issue:46172", "issue:43792|issue:46123", "issue:43352|issue:43976", - "issue:44380|issue:46132" + "issue:44380|issue:46132", + "issue:44380|issue:45479", + "issue:43064|issue:44380", + "issue:44483|issue:44514" ], - "prepared_review_unit_hash": "bd3a59c91a531a3f9ecf0d18f4e160588cee11245c608b8a84797f9c5bc59d23", + "prepared_review_unit_hash": "313f5ec691155985b7999297ba66fea9145d7134ffdacdee780b99daf5af1e91", "component_budget": { "node_count": 426, "item_count": 426, - "soft_pair_count": 1572, - "serialized_chars": 814130, - "estimated_input_tokens": 203533, - "estimated_eval_tokens": 407322 + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 33, - "serialized_chars": 37425, - "estimated_input_tokens": 9357, - "estimated_eval_tokens": 18970 + "serialized_chars": 37467, + "estimated_input_tokens": 9367, + "estimated_eval_tokens": 18990 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "These issues are only loosely related by broad Transformers/HF themes, but none of the soft pairs look like true duplicates or the same fixable code-path bug. I would not merge any of the soft edges.", - "confidence": 0.89, - "canonical_issue_reason": "issue:43097 is the broadest, most central v5 regression report in the set and a reasonable anchor if one issue must represent the cluster, though the set is not a true duplicate cluster.", + "summary": "These items are mostly unrelated Transformers bugs spanning imports, model configs, processor loading, training, and attention backends. The soft matches are mostly broad subsystem similarity rather than the same underlying defect, so none should be merged as duplicates.", + "confidence": 0.96, + "canonical_issue_reason": "issue:43097 is the clearest representative of a user-facing v5 regression with explicit impact and inbound references, making it a strong anchor for triage of version-breakage reports.", "canonical_pr_reason": null, - "best_issue_reason": "issue:43097 has the clearest cross-cutting user impact and references among the listed issues, making it the best single issue to keep as the representative regression report.", + "best_issue_reason": "issue:43873 is the best standalone issue to prioritize: it is open, active, and describes a broadly impactful, concrete failure mode around offline/remote loading behavior.", "best_pr_reason": null, "soft_edge_verdicts": [ { - "left": "issue:43408", - "right": "issue:44368", - "accept": false, - "reason": "Both are warning reports, but they concern different model/config mismatches and different code paths." - }, - { - "left": "issue:43986", - "right": "issue:44484", - "accept": false, - "reason": "One is an AutoProcessor/torchvision crash; the other is about save_pretrained shard sizing." - }, - { - "left": "issue:43874", - "right": "issue:44263", + "left": "issue:45440", + "right": "issue:46172", "accept": false, - "reason": "Different GLM internals: missing image-patch counting vs torch.split behavior in an indexer." + "reason": "Different DeepSeek model families and different bug classes; not the same code path." }, { "left": "issue:43531", "right": "issue:45341", "accept": false, - "reason": "A Qwen3-MoE sliding-window bug is unrelated to a testing_utils bug." - }, - { - "left": "issue:45440", - "right": "issue:46172", - "accept": false, - "reason": "Both touch DeepSeek models, but they describe different failures in different code paths." + "reason": "Sliding-window model behavior vs a testing utility bug; unrelated." }, { "left": "issue:43986", "right": "issue:45372", "accept": false, - "reason": "torchvision-missing video-model crash and a mistral_common import error are unrelated." + "reason": "Both involve processor loading, but the missing dependency/import and affected paths differ." }, { "left": "issue:38617", "right": "issue:42757", "accept": false, - "reason": "Both are import errors, but for different symbols and different modules." + "reason": "Separate ImportErrors from different modules and symbols." }, { "left": "issue:44261", "right": "issue:44485", "accept": false, - "reason": "Layernorm epsilon precision drift and RoPE implementation issues are distinct bugs." + "reason": "Precision warning in MLA layer norm vs GLM-5 RoPE implementation; unrelated defects." }, { "left": "issue:43064", "right": "issue:44315", "accept": false, - "reason": "FSDP2/PEFT optimizer-state corruption and Liger Kernel application are different training issues." + "reason": "Optimizer-state corruption under FSDP/PEFT vs Liger kernel not being applied; different problems." }, { "left": "issue:43097", "right": "issue:46050", "accept": false, - "reason": "A v5 API/config removal is not the same as broken quantized save/load behavior." + "reason": "Config removal regression vs quantized save/load corruption; not the same bug." }, { "left": "issue:43097", "right": "issue:45588", "accept": false, - "reason": "Config removal and a flash_attention.py AttributeError are different code paths." + "reason": "v5 config change issue vs flash-attention crash on sink-less models." }, { "left": "issue:44062", "right": "issue:44117", "accept": false, - "reason": "Tokenizer AddedToken kwargs conflict is unrelated to TOKENIZER_MAPPING_NAMES returning None." + "reason": "AddedToken constructor argument clash vs tokenizer-mapping None handling; different failure points." }, { "left": "issue:44410", "right": "issue:45375", "accept": false, - "reason": "Missing attention projections and a missing vision config field are different model bugs." + "reason": "Missing projections in a model layer vs a missing config field silently dropped." }, { "left": "issue:38617", "right": "issue:44351", "accept": false, - "reason": "Both are import failures, but they target different exported names and causes." + "reason": "Different missing-import regressions involving different names and modules." }, { "left": "issue:44038", "right": "issue:44410", "accept": false, - "reason": "A Qwen3-VL-Moe issue and a qwen3next projection issue are not the same bug." + "reason": "Qwen3-VL-MoE bug vs Qwen3next projection bug; same broad family, different concrete issue." }, { "left": "issue:43502", "right": "issue:43992", "accept": false, - "reason": "Local-files-only network leakage and a missing embed_tokens weight on load are unrelated." + "reason": "Offline-download behavior vs missing embedding weights on load; unrelated." }, { "left": "issue:43097", "right": "issue:43873", "accept": false, - "reason": "A config removal issue is not the same as quantization offloading misbehavior." + "reason": "Config API removal vs quantization offloading behavior; different code paths." }, { "left": "issue:44704", "right": "issue:45103", "accept": false, - "reason": "Kwarg forwarding to cached_file and an annotation-related auto_docstring crash are unrelated." + "reason": "cached_file kwargs propagation vs a docstring/annotation crash; not the same defect." }, { "left": "issue:43986", "right": "issue:44117", "accept": false, - "reason": "A model-loading crash and a tokenizer mapping edge case are different problems." + "reason": "Processor/torchvision loading crash vs tokenizer mapping lookup bug." }, { "left": "issue:43066", "right": "issue:43906", "accept": false, - "reason": "The second is only an isolated reproduction reference; there is not enough evidence it is the same tokenizer bug." + "reason": "Wrong tokenizer decoder type vs an unrelated reproduced issue; no concrete overlap." }, { "left": "issue:43824", "right": "issue:43986", "accept": false, - "reason": "Missing Qwen2.5-VL import and a torchvision-dependent video model crash are different issues." + "reason": "Missing exported model class vs AutoProcessor/torchvision crash; separate import and runtime paths." }, { "left": "issue:43208", "right": "issue:43550", "accept": false, - "reason": "xLSTM training blockers and a Bamba torch.compile/SDPA failure are unrelated." + "reason": "xLSTM training blockers vs Bamba compile/SDPA failure; different models and failures." }, { "left": "issue:43825", "right": "issue:44655", "accept": false, - "reason": "Pipeline task messaging and pipeline serialization are different behaviors." + "reason": "Pipeline error message vs pipeline serialization bug; not the same underlying change." }, { "left": "issue:43992", "right": "issue:44484", "accept": false, - "reason": "A missing load-time weight and a save_pretrained shard-size question are not duplicates." + "reason": "Missing checkpoint weight vs max_shard_size default behavior; unrelated." }, { "left": "issue:43012", "right": "issue:43550", "accept": false, - "reason": "A precision warning and a compile-time SDPA failure are different issues." + "reason": "PyTorch warning during bfloat16 compilation vs a torch.compile SDPA model failure." }, { "left": "issue:43012", "right": "issue:44368", "accept": false, - "reason": "Both are warnings, but they stem from different model/config conditions." + "reason": "Different warnings with different root causes and model setups." }, { "left": "issue:43352", "right": "issue:44380", "accept": false, - "reason": "Unsupported FlashAttention for one model is not the same as GPT2 attention scaling being ignored." + "reason": "Unsupported Flash Attention 2.0 for one model vs GPT2 attention-scaling being ignored." }, { "left": "issue:42907", "right": "issue:45753", "accept": false, - "reason": "A dequantized save failure and an image-specific infinite loop are unrelated." + "reason": "Save/load dequantization bug vs an infinite-loop vision prompt bug." }, { "left": "issue:42907", "right": "issue:44655", "accept": false, - "reason": "Saving dequantized weights is different from saving Pipeline objects." + "reason": "Model save/pretrained quantization issue vs pipeline object serialization." }, { "left": "issue:44162", "right": "issue:46172", "accept": false, - "reason": "ESM2 breakage and a DeepSeek tensor transpose bug are different code paths." + "reason": "ESM2 breakage vs a DeepSeek V4 tensor transpose bug; unrelated." }, { "left": "issue:43792", "right": "issue:46123", "accept": false, - "reason": "Whisper model load failure and MaskGenerationPipeline batching logic are unrelated." + "reason": "Whisper model load failure vs partial-batch mask pipeline bookkeeping bug." }, { "left": "issue:43352", "right": "issue:43976", "accept": false, - "reason": "A model-specific attention backend limitation is not a Python version compatibility bug." + "reason": "Attention backend support issue vs Python version incompatibility." }, { "left": "issue:44380", "right": "issue:46132", "accept": false, - "reason": "GPT2 backend-specific scaling behavior and AttentionInterface.register semantics are distinct changes." + "reason": "GPT2 backend-specific scaling bug vs AttentionInterface.register changing callback behavior." + }, + { + "left": "issue:44380", + "right": "issue:45479", + "accept": false, + "reason": "Attention backend config bug vs zero-loss sequence-classification bug; different subsystems." + }, + { + "left": "issue:43064", + "right": "issue:44380", + "accept": false, + "reason": "Distributed optimizer-state corruption vs attention backend scaling behavior." + }, + { + "left": "issue:44483", + "right": "issue:44514", + "accept": false, + "reason": "Chat/completions API rejection vs batched processor apply_chat_template crash." + }, + { + "left": "issue:45440", + "right": "issue:46172", + "accept": false, + "reason": "Different DeepSeek versions and different concrete implementation bugs; not mergeable as one fix." } ] }, @@ -14048,7 +14037,7 @@ "component_index": 1, "component_count": 17, "review_unit_index": 45, - "review_unit_count": 50, + "review_unit_count": 49, "status": "reviewed", "reason": null, "source": "llm", @@ -14064,13 +14053,13 @@ "issue:42898", "issue:42994", "issue:43012", - "issue:43064", "issue:43066", "issue:43208", "issue:43299", "issue:43352", "issue:43475", "issue:43502", + "issue:43525", "issue:43526", "issue:43582", "issue:43611", @@ -14089,7 +14078,6 @@ "issue:44380", "issue:44483", "issue:44485", - "issue:44514", "issue:44625", "issue:44655", "issue:44683", @@ -14099,23 +14087,21 @@ "issue:45290", "issue:45381", "issue:45478", - "issue:45479", + "issue:45593", "issue:45698", "issue:45753", "issue:46077", + "issue:46082", "issue:46172" ], "soft_pairs": [ - "issue:44380|issue:45479", - "issue:43064|issue:44380", - "issue:44483|issue:44514", "issue:44222|issue:45478", "issue:36683|issue:43582", "issue:43208|issue:43526", "issue:43012|issue:44683", "issue:42490|issue:43012", - "issue:43618|issue:43761", "issue:42757|issue:43824", + "issue:43618|issue:43761", "issue:45698|issue:46172", "issue:43618|issue:44380", "issue:44038|issue:44230", @@ -14125,8 +14111,8 @@ "issue:43299|issue:43867", "issue:43756|issue:44485", "issue:36683|issue:43475", - "issue:45276|issue:46077", "issue:43901|issue:44655", + "issue:45276|issue:46077", "issue:43299|issue:45753", "issue:42994|issue:44655", "issue:42898|issue:43066", @@ -14137,228 +14123,232 @@ "issue:43352|issue:43937", "issue:43611|issue:44655", "issue:44625|issue:46077", - "issue:43976|issue:44938" + "issue:43976|issue:44938", + "issue:46077|issue:46082", + "issue:43352|issue:45593", + "issue:36683|issue:43525", + "issue:43618|issue:45216" ], - "prepared_review_unit_hash": "e3d2bfee4ccf675eb6f17c72d0642bd5e7504930237c22cdf99083139fd11e71", + "prepared_review_unit_hash": "bfd191d700ef37b91fea7b12a506836b3beb3b37108f7c0f917e4e55027c54dd", "component_budget": { "node_count": 426, "item_count": 426, - "soft_pair_count": 1572, - "serialized_chars": 814130, - "estimated_input_tokens": 203533, - "estimated_eval_tokens": 407322 + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 32, - "serialized_chars": 36939, - "estimated_input_tokens": 9235, - "estimated_eval_tokens": 18726 + "soft_pair_count": 33, + "serialized_chars": 37104, + "estimated_input_tokens": 9276, + "estimated_eval_tokens": 18808 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is heterogeneous: many issues are Transformers v5 regressions, but they span unrelated subsystems (imports, loading/saving, chat templates, model-specific configs, and backend-specific attention behavior). None of the soft pairs look like the same underlying bug or change.", - "confidence": 0.69, - "canonical_issue_reason": "Issue 44380 is the best single representative if one must be chosen: it is a concrete, code-path-specific v5 regression with a clear runtime symptom. The rest of the set is too diverse to share one true canonical issue.", + "summary": "This set is mostly a grab bag of distinct Transformers regressions, import errors, and model-specific bugs. I did not find any soft pair that looks like the same underlying change, so I would not merge any of them as duplicates.", + "confidence": 0.34, + "canonical_issue_reason": "issue:45698 is the broadest representative here: a core from_pretrained/save_pretrained loading/saving bug with active discussion and inbound references, unlike the many narrow model-specific issues.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 44380 is the strongest representative because it describes a focused behavioral regression rather than a broad compatibility or model-family report. It is still only a forced choice; the overall cluster does not have a clear unified issue.", + "best_issue_reason": "issue:45698 is the best global representative because it targets a central library API path (model loading/saving) and has comparatively high discussion activity.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:44380", - "right": "issue:45479", - "accept": false, - "reason": "Different bugs: GPT2 attention scaling ignored under specific backends vs degenerate zero loss in sequence classification." - }, - { - "left": "issue:43064", - "right": "issue:44380", - "accept": false, - "reason": "Training/distributed optimizer-state mismatch vs attention-backend scaling regression; unrelated code paths." - }, - { - "left": "issue:44483", - "right": "issue:44514", - "accept": false, - "reason": "Chat/completions request rejection vs batched chat template crash; different API layers and triggers." - }, { "left": "issue:44222", "right": "issue:45478", "accept": false, - "reason": "FP8 save_pretrained failure vs Qwen3.5 MoE from_pretrained error; save-time vs load-time and different failure modes." + "reason": "Both involve Qwen3.5/MoE serialization, but one is FP8 save_pretrained and the other is a from_pretrained load error; different failure modes." }, { "left": "issue:36683", "right": "issue:43582", "accept": false, - "reason": "Missing Gemma3Config attribute vs Apple Silicon allocator TypeError; no shared underlying bug." + "reason": "Gemma3 config attribute error vs Apple Silicon allocator TypeError; unrelated bugs." }, { "left": "issue:43208", "right": "issue:43526", "accept": false, - "reason": "xLSTM training blockers vs BeitImageProcessorFast label reduction bug; different models and components." + "reason": "xLSTM training bugs are unrelated to a BEiT image-processor label reduction bug." }, { "left": "issue:43012", "right": "issue:44683", "accept": false, - "reason": "PyTorch precision warning during compile vs flex_attention failing on torch>=2.9; both compile-related but not the same bug." + "reason": "One is a warning during bfloat16 compilation, the other is a hard failure in compiled flex_attention on torch>=2.9; not the same issue." }, { "left": "issue:42490", "right": "issue:43012", "accept": false, - "reason": "Trailing-slash local model path warning vs bfloat16 compile warning; unrelated behaviors." + "reason": "Local-path slash import warning and bfloat16 compile warning are different code paths." }, { - "left": "issue:43618", - "right": "issue:43761", + "left": "issue:42757", + "right": "issue:43824", "accept": false, - "reason": "Both CLIP regressions, but one drops attentions and the other drops hidden_states; different outputs and code paths." + "reason": "Different missing imports from different packages/symbols; not a duplicate." }, { - "left": "issue:42757", - "right": "issue:43824", + "left": "issue:43618", + "right": "issue:43761", "accept": false, - "reason": "Missing huggingface_hub import vs missing Transformers model export; both import errors, but for different symbols and causes." + "reason": "Both are CLIP v5 regressions, but one drops attentions and the other drops hidden_states; no clear shared concrete bug." }, { "left": "issue:45698", "right": "issue:46172", "accept": false, - "reason": "Wrong custom module loaded after save_pretrained vs deepseek tensor transpose bug; unrelated serialization vs tensor math." + "reason": "Generic load/save module-resolution bug vs DeepSeek dtype/transpose bug; unrelated." }, { "left": "issue:43618", "right": "issue:44380", "accept": false, - "reason": "CLIP attentions regression vs GPT2 attention scaling regression; different models and symptoms." + "reason": "CLIP output-field regression and GPT2 attention-scaling backend behavior are separate problems." }, { "left": "issue:44038", "right": "issue:44230", "accept": false, - "reason": "Qwen3-VL-Moe bug report vs fp8 support request for Qwen3-VL/Qwen3.5 MoE; same family, not the same concrete change." + "reason": "Same family (Qwen3-VL/MoE) but different issues: generic loading breakage vs fp8 inference support." }, { "left": "issue:38617", "right": "issue:43502", "accept": false, - "reason": "Missing configuration_utils import vs local_files_only still making API requests; import bug vs network behavior." + "reason": "ImportError for a missing utility vs network requests despite local_files_only=True; unrelated." }, { "left": "issue:44483", "right": "issue:45290", "accept": false, - "reason": "Endpoint request validation failure vs chat template crash on tool-call messages; related area, different bug." + "reason": "OpenAI chat-completions request validation and chat-template tokenization crash are different paths." }, { "left": "issue:44483", "right": "issue:45381", "accept": false, - "reason": "API acceptance regression vs wrong vision_position_ids in Qwen2.5-VL; unrelated failures." + "reason": "API request rejection and Qwen2.5-VL video position-id bug are unrelated." }, { "left": "issue:43299", "right": "issue:43867", "accept": false, - "reason": "Qwen3VL MoE loading break vs sorted state_dict load error; different loading mechanisms and symptoms." + "reason": "Both are loading errors, but one is a Qwen3-VL/MoE regression and the other is a state_dict-order issue; not the same root cause." }, { "left": "issue:43756", "right": "issue:44485", "accept": false, - "reason": "Concrete Smollm3 RoPE-layer loss vs broad GLM-5 RoPE discussion; not the same bug." + "reason": "Smollm3 RoPE-layer mismatch and GLM-5 RoPE discussion are different model-specific bugs." }, { "left": "issue:36683", "right": "issue:43475", "accept": false, - "reason": "Gemma3Config missing vocab_size vs SAM3 output missing fpn_position_embeddings; both missing-attribute issues, but on different objects/models." - }, - { - "left": "issue:45276", - "right": "issue:46077", - "accept": false, - "reason": "Gemma4 embedding resize propagation bug vs Gemma4 attention-mask bug; same model family, different subsystems and fixes." + "reason": "Different models and different missing attributes; no shared code path." }, { "left": "issue:43901", "right": "issue:44655", "accept": false, - "reason": "Documentation mismatch for return_all_scores vs inability to save Pipeline objects; docs issue vs serialization bug." + "reason": "Docs/behavior mismatch for a pipeline arg is unrelated to pipeline save_pretrained failure." }, { - "left": "issue:43299", - "right": "issue:45753", + "left": "issue:45276", + "right": "issue:46077", "accept": false, - "reason": "Qwen3VL loading regression vs infinite-loop generation on a specific image; different symptoms and triggers." + "reason": "Gemma4 embedding resize propagation bug vs Gemma4 attention-mask bug; separate issues." }, { "left": "issue:42994", "right": "issue:44655", "accept": false, - "reason": "Quantized model saving failure vs pipeline save_pretrained failure; both save-related, but not the same code path." + "reason": "Quantized model saving failure is not the same as saving Pipeline objects." }, { "left": "issue:42898", "right": "issue:43066", "accept": false, - "reason": "Tokenization cleanup behavior change vs wrong tokenizer decoder type; distinct tokenizer regressions." + "reason": "Tokenization cleanup behavior change and wrong tokenizer decoder type are distinct tokenizer regressions." }, { "left": "issue:36010", "right": "issue:42757", "accept": false, - "reason": "Missing GenerationMixin import vs missing is_offline_mode import; separate import breakages." + "reason": "Both are import errors, but for different symbols/modules and likely unrelated fixes." }, { "left": "issue:44077", "right": "issue:44380", "accept": false, - "reason": "patchtsmixer post_init validation vs GPT2 attention scaling under SDPA/FlashAttention; unrelated." + "reason": "patchtsmixer post_init validation is unrelated to GPT2 attention scaling under SDPA/FlashAttention." }, { "left": "issue:36683", "right": "issue:44077", "accept": false, - "reason": "Gemma3Config attribute bug vs patchtsmixer post_init policy bug; no shared underlying issue." + "reason": "Gemma3 config attribute bug and patchtsmixer post_init policy change are unrelated." }, { "left": "issue:43867", "right": "issue:45216", "accept": false, - "reason": "state_dict ordering load error vs Qwen3.5 save_pretrained regression; load-time vs save-time problems." + "reason": "Sorted-state_dict load failure and incorrect Qwen3.5 checkpoint save are different serialization bugs." }, { "left": "issue:43352", "right": "issue:43937", "accept": false, - "reason": "Nemotron Flash Attention 2 support gap vs GLM-5 GenerationConfig invalid; different model setup issues." + "reason": "Flash Attention support warning and invalid GenerationConfig error are unrelated." }, { "left": "issue:43611", "right": "issue:44655", "accept": false, - "reason": "base_model_prefix loading regression vs pipeline save_pretrained failure; different persistence paths." + "reason": "Base-model-prefix loading regression and pipeline save_pretrained failure are different code paths." }, { "left": "issue:44625", "right": "issue:46077", "accept": false, - "reason": "Qwen3.5 num_labels propagation bug vs Gemma4 bidirectional attention-mask bug; unrelated model-specific regressions." + "reason": "Qwen3.5 num_labels propagation and Gemma4 attention-mask construction are unrelated." }, { "left": "issue:43976", "right": "issue:44938", "accept": false, - "reason": "Python version compatibility bug vs Python 3.14 load failure; both compatibility-related, but not the same underlying issue." + "reason": "Both are Python-version compatibility reports, but for different versions and likely different causes." + }, + { + "left": "issue:46077", + "right": "issue:46082", + "accept": false, + "reason": "Gemma4 attention-mask bug vs LlamaConfig head_dim validation are unrelated." + }, + { + "left": "issue:43352", + "right": "issue:45593", + "accept": false, + "reason": "Unsupported Flash Attention 2.0 and D-FINE auxiliary-loss behavior are unrelated." + }, + { + "left": "issue:36683", + "right": "issue:43525", + "accept": false, + "reason": "Different model configs and different missing attributes." + }, + { + "left": "issue:43618", + "right": "issue:45216", + "accept": false, + "reason": "CLIP output regression and Qwen3.5 save_pretrained regression are unrelated." } ] }, @@ -14374,7 +14364,7 @@ "component_index": 1, "component_count": 17, "review_unit_index": 46, - "review_unit_count": 50, + "review_unit_count": 49, "status": "reviewed", "reason": null, "source": "llm", @@ -14393,9 +14383,10 @@ "issue:43064", "issue:43352", "issue:43502", - "issue:43525", "issue:43575", + "issue:43576", "issue:43618", + "issue:43723", "issue:43867", "issue:43931", "issue:43976", @@ -14408,34 +14399,29 @@ "issue:44246", "issue:44355", "issue:44360", + "issue:44387", "issue:44484", "issue:44492", + "issue:44556", "issue:44855", "issue:44857", "issue:44938", "issue:44964", "issue:44977", "issue:45071", - "issue:45216", "issue:45278", "issue:45310", "issue:45357", "issue:45375", "issue:45381", "issue:45440", + "issue:45478", "issue:45561", - "issue:45593", "issue:45600", "issue:45850", - "issue:46002", - "issue:46077", - "issue:46082" + "issue:46132" ], "soft_pairs": [ - "issue:46077|issue:46082", - "issue:43352|issue:45593", - "issue:36683|issue:43525", - "issue:43618|issue:45216", "issue:36683|issue:44242", "issue:43352|issue:43931", "issue:43064|issue:44492", @@ -14452,8 +14438,8 @@ "issue:44938|issue:45381", "issue:43010|issue:43976", "issue:42490|issue:44075", + "issue:43992|issue:45478", "issue:43618|issue:44077", - "issue:43992|issue:46002", "issue:36296|issue:43575", "issue:28282|issue:33453", "issue:43867|issue:44484", @@ -14463,234 +14449,234 @@ "issue:42890|issue:44964", "issue:42890|issue:44112", "issue:42890|issue:45600", - "issue:43502|issue:44062" + "issue:43502|issue:44062", + "issue:43723|issue:44556", + "issue:43576|issue:46132", + "issue:44387|issue:44556" ], - "prepared_review_unit_hash": "34f818cc6c60d782ab02a4ec906a99ef497fe709d18330e0e7a7375b554e7390", + "prepared_review_unit_hash": "75aafdbfe773c959501ea31618bbac392a786b912b8de88ba12744b0ed89c37d", "component_budget": { "node_count": 426, "item_count": 426, - "soft_pair_count": 1572, - "serialized_chars": 814130, - "estimated_input_tokens": 203533, - "estimated_eval_tokens": 407322 + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 32, - "serialized_chars": 36843, - "estimated_input_tokens": 9211, - "estimated_eval_tokens": 18678 + "soft_pair_count": 31, + "serialized_chars": 36499, + "estimated_input_tokens": 9125, + "estimated_eval_tokens": 18506 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a noisy mix of unrelated Transformers bug reports: import/load regressions, config attribute errors, model-specific generation/loading failures, and a few training/test issues. None of the soft pairs look like true duplicates or mergeable PR counterparts.", - "confidence": 0.46, - "canonical_issue_reason": "issue:45278 is the best umbrella representative: it\u2019s a concrete, broad regression report about many import errors after upgrading, which matches the cluster\u2019s general theme of release-related breakage better than the more model-specific reports.", + "summary": "The cluster is still best treated as a loose collection of Transformers v5 regressions, model-loading bugs, and CI/test issues. The broad v5 import/loading report is the best anchor, but the soft edges are mostly false positives; in particular the UMT5Encoder loading bug should not be merged with the Qwen3.5 MoE from_pretrained reports.", + "confidence": 0.71, + "canonical_issue_reason": "issue:45278 is the best broad representative of the recurring v5 upgrade/import-regression theme, and it is concrete enough to anchor the cluster without collapsing distinct model-specific failures.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45278 is the most suitable global representative because it is broad, user-facing, and regression-oriented; it\u2019s the closest thing here to a cluster-level anchor, even though it is not a duplicate of the other issues.", + "best_issue_reason": "issue:45278 is the strongest single issue to represent the cluster because it covers the dominant upgrade/load regression pattern seen across several items, while remaining more specific than the generic complaints.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:46077", - "right": "issue:46082", - "accept": false, - "reason": "Different config validation bugs for different model families: Gemma4 attention masking vs LlamaConfig head_dim handling." - }, - { - "left": "issue:43352", - "right": "issue:45593", - "accept": false, - "reason": "Unrelated model behaviors: Flash Attention support error vs D-FINE auxiliary-loss logic." - }, - { - "left": "issue:36683", - "right": "issue:43525", - "accept": false, - "reason": "Both are missing-config-attribute symptoms, but for different models and fields (vocab_size vs pad_token_id), so not the same bug." - }, - { - "left": "issue:43618", - "right": "issue:45216", - "accept": false, - "reason": "Different subsystems: CLIP output fields vs Qwen3.5 save_pretrained checkpoint corruption." - }, { "left": "issue:36683", "right": "issue:44242", "accept": false, - "reason": "One is a config attribute error, the other is a MoE load-balancing loss computation bug." + "reason": "Different bugs: Gemma3Config vocab_size access vs missing load-balancing loss when router logits are disabled." }, { "left": "issue:43352", "right": "issue:43931", "accept": false, - "reason": "Different concrete problems: unsupported Flash Attention path vs Qwen3-VL weight-shape mismatch while loading." + "reason": "Unrelated failure modes: Flash Attention unsupported error vs Qwen3-VL weight-shape mismatch." }, { "left": "issue:43064", "right": "issue:44492", "accept": false, - "reason": "Training/optimizer-state bug and a cache-strategy typo are unrelated." + "reason": "Different subsystems and symptoms: distributed optimizer-state corruption vs a cache-strategy typo." }, { "left": "issue:43010", "right": "issue:45278", "accept": false, - "reason": "Cache update no-grad behavior and import-time regressions are different code paths and symptoms." + "reason": "A no_grad decoration request for cache/layer updates is not the same concrete bug as a broad v5 import/load regression." }, { "left": "issue:43992", "right": "issue:45310", "accept": false, - "reason": "Different from_pretrained failures on different Qwen variants; not the same underlying loading bug." + "reason": "Not the same bug: UMT5Encoder missing embed_tokens.weight is a different model/path from Qwen3.5 MoE from_pretrained errors." }, { "left": "issue:44355", "right": "issue:44855", "accept": false, - "reason": "Both involve import/runtime failures, but one is compiled-Python-file errors and the other is a Python 3.13 torch.jit parsing/IndentationError issue." + "reason": "Compiled-Python import failures and a Python 3.13 torch.jit parsing issue are different code paths." }, { "left": "issue:43867", "right": "issue:45375", "accept": false, - "reason": "Generic state_dict/load ordering trouble vs a missing config field dropped by @strict; different root causes." + "reason": "State_dict sorting load error and missing deepstack_visual_indexes in config are different model-loading problems." }, { "left": "issue:41669", "right": "issue:44246", "accept": false, - "reason": "Related to import performance, but one is a proposed star-import cleanup and the other is intermittent slowness when importing transformers." + "reason": "Both mention import performance, but one is a specific import-star cleanup proposal and the other is an intermittent slowdown report." }, { "left": "issue:42890", "right": "issue:45561", "accept": false, - "reason": "Both are test flakiness reports, but one is missing set_seed in a model integration test and the other is a pytest-xdist file race." + "reason": "Test flakiness from missing set_seed is unrelated to the xdist captured_info.txt race." }, { "left": "issue:36683", "right": "issue:43064", "accept": false, - "reason": "Config attribute error and distributed training optimizer-state corruption are unrelated." + "reason": "Config attribute error and distributed optimizer-state corruption are unrelated failures." }, { "left": "issue:44938", "right": "issue:45071", "accept": false, - "reason": "Python 3.14 import failure and PretrainedConfig type-checking breakage after v5.4.0 are different issues." + "reason": "Python-version load failure and PretrainedConfig type-checking breakage are different regressions." }, { "left": "issue:43867", "right": "issue:45357", "accept": false, - "reason": "Different save/load regressions: state_dict sorting vs incorrect visual encoder keys in Qwen3.5 save_pretrained." + "reason": "Different save/load bugs: sorted state_dict handling vs incorrect visual encoder keys on save_pretrained." }, { "left": "issue:43010", "right": "issue:45850", "accept": false, - "reason": "A specific cache-method no-grad issue is not the same as the vague broad regression complaint about too many models breaking." + "reason": "A cache update decoration issue is not the same bug as a broad 'v5 breaks too many models' complaint." }, { "left": "issue:44938", "right": "issue:45381", "accept": false, - "reason": "Python-version load failure and Qwen2.5-VL video position-id bug are unrelated." + "reason": "Generic Python 3.14 import failure is unrelated to Qwen2.5-VL video position-id behavior." }, { "left": "issue:43010", "right": "issue:43976", "accept": false, - "reason": "Cache update behavior and Python-version compatibility breakage are not the same bug." + "reason": "Cache update decoration and Python version compatibility are unrelated issues." }, { "left": "issue:42490", "right": "issue:44075", "accept": false, - "reason": "Import warning on local path handling vs optimizer SGD arguments not being used are unrelated." + "reason": "Local-files-only path handling is unrelated to SGD argument handling." }, { - "left": "issue:43618", - "right": "issue:44077", + "left": "issue:43992", + "right": "issue:45478", "accept": false, - "reason": "CLIP attentions field regression and patchtsmixer post_init policy are separate changes." + "reason": "Not the same bug: UMT5Encoder embed_tokens.weight omission is a different failure than Qwen3.5 MoE from_pretrained errors." }, { - "left": "issue:43992", - "right": "issue:46002", + "left": "issue:43618", + "right": "issue:44077", "accept": false, - "reason": "Different symptoms and code paths: missing encoder weights vs NaNs after checkout/init in classification heads." + "reason": "CLIP attentions regression and patchtsmixer post_init eligibility are different model-specific bugs." }, { "left": "issue:36296", "right": "issue:43575", "accept": false, - "reason": "Both mention tensor parallelism, but one is a training bug while the other is an OOM on model loading; not the same underlying issue." + "reason": "Tensor-parallel training bug and tp-related OOM on model load are different failures." }, { "left": "issue:28282", "right": "issue:33453", "accept": false, - "reason": "ImportError about missing PyTorch and a tokenizer-loading regression are unrelated." + "reason": "AutoModel missing PyTorch and tokenizer loading regression are unrelated issues." }, { "left": "issue:43867", "right": "issue:44484", "accept": false, - "reason": "Load-model failure from state_dict ordering is unrelated to the max_shard_size question." + "reason": "Sorted state_dict load error is unrelated to max_shard_size default behavior." }, { "left": "issue:43618", "right": "issue:44857", "accept": false, - "reason": "CLIP attentions regression and LwDetrImageLoss AMP/CUDA crash are different bugs." + "reason": "Attention-output regression and float16 AMP crash are different bugs." }, { "left": "issue:44360", "right": "issue:45440", "accept": false, - "reason": "DSA indexer activation discussion and DeepseekV3MoE divergence are unrelated." + "reason": "DSA indexer ReLU discussion is unrelated to DeepSeekV3MoE implementation divergence." }, { "left": "issue:43618", "right": "issue:44977", "accept": false, - "reason": "Different model-specific issues: CLIP output field regression vs Qwen3.5 flash-attention generation bug." + "reason": "CLIP attentions regression does not match Qwen3.5 flash-attention generation failure." }, { "left": "issue:42890", "right": "issue:44964", "accept": false, - "reason": "A flaky CI test issue is not the same as a model-loading regression for Phi-4 multimodal." + "reason": "SamHQ test seeding flakiness is unrelated to Phi-4 multimodal model-loading failure." }, { "left": "issue:42890", "right": "issue:44112", "accept": false, - "reason": "Both are CI/test failures, but they affect different tests with different causes." + "reason": "CI flakiness in a model test and a stale device-override test are different problems." }, { "left": "issue:42890", "right": "issue:45600", "accept": false, - "reason": "Different CI failures: seed-related instability vs stale auto_mappings references." + "reason": "Captured-info race and removed-config references are distinct CI issues." }, { "left": "issue:43502", "right": "issue:44062", "accept": false, - "reason": "Local-files-only network leakage and AddedToken special-arg duplication are unrelated loading/tokenizer bugs." + "reason": "Local-files-only API requests and AddedToken argument duplication are unrelated." + }, + { + "left": "issue:43723", + "right": "issue:44556", + "accept": false, + "reason": "Tokenizer loading and checkpoint reload after upgrade are both load-related, but the packet does not show the same concrete bug." + }, + { + "left": "issue:43576", + "right": "issue:46132", + "accept": false, + "reason": "Broken env CLI behavior and AttentionInterface.register behavior change are different features." + }, + { + "left": "issue:44387", + "right": "issue:44556", + "accept": false, + "reason": "Int4 CUDA memory regression is unrelated to checkpoint reload failures after upgrade." } ] }, - "evaluator_result": null, - "evaluator_used": false, - "retried": false, + "evaluator_result": { + "accept": false, + "feedback": "Reject: the accepted soft-edge merges are not grounded in the packet. issue:43992 is \"UMT5Encoder.from_pretrained misses embed_tokens.weight,\" while issue:45310 and issue:45478 are Qwen3.5 MoE from_pretrained errors; the packet does not support treating 43992 as the same bug as either of those. The rest of the verdicts are generally conservative, and the cluster summary is plausible, but these accepted pairings overstate evidence." + }, + "evaluator_used": true, + "retried": true, "accepted_nontrivial_soft_edge": false, "error_kind": null, "error_message": null @@ -14700,7 +14686,7 @@ "component_index": 1, "component_count": 17, "review_unit_index": 47, - "review_unit_count": 50, + "review_unit_count": 49, "status": "reviewed", "reason": null, "source": "llm", @@ -14728,7 +14714,6 @@ "issue:43526", "issue:43576", "issue:43673", - "issue:43723", "issue:43824", "issue:43827", "issue:43874", @@ -14739,7 +14724,6 @@ "issue:44351", "issue:44355", "issue:44360", - "issue:44387", "issue:44485", "issue:44492", "issue:44556", @@ -14750,20 +14734,18 @@ "issue:45071", "issue:45125", "issue:45341", + "issue:45362", "issue:45375", "issue:45468", "issue:45478", - "issue:46132", + "issue:45850", "issue:46143", "issue:46144" ], "soft_pairs": [ - "issue:44387|issue:44556", - "issue:43576|issue:46132", - "issue:43723|issue:44556", "issue:42890|issue:45341", - "issue:44222|issue:45125", "issue:42548|issue:44351", + "issue:44222|issue:45125", "issue:43010|issue:43065", "issue:44355|issue:44749", "issue:43502|issue:44485", @@ -14778,8 +14760,8 @@ "issue:44315|issue:44360", "issue:43502|issue:44961", "issue:30333|issue:36296", - "issue:43827|issue:44961", "issue:44355|issue:45071", + "issue:43827|issue:44961", "issue:44355|issue:44655", "issue:33453|issue:34689", "issue:42757|issue:43576", @@ -14793,238 +14775,263 @@ "issue:38617|issue:43576", "issue:46143|issue:46144", "issue:44908|issue:44961", - "issue:43673|issue:44492", "issue:44485|issue:44961", + "issue:43673|issue:44492", "issue:36296|issue:43901", "issue:41628|issue:44355", - "issue:36296|issue:43526" + "issue:36296|issue:43526", + "issue:45362|issue:45850" ], - "prepared_review_unit_hash": "368ed1501997938eca69fdba49e86073ef6fe4dd138f6471219c53281eed9808", + "prepared_review_unit_hash": "d77ad5e89078983e727462da05c7e855d10ef4e411d1999ca4abc0412ea6571b", "component_budget": { "node_count": 426, "item_count": 426, - "soft_pair_count": 1572, - "serialized_chars": 814130, - "estimated_input_tokens": 203533, - "estimated_eval_tokens": 407322 + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 }, "budget": { - "node_count": 48, - "item_count": 48, - "soft_pair_count": 40, - "serialized_chars": 39055, - "estimated_input_tokens": 9764, - "estimated_eval_tokens": 19784 + "node_count": 47, + "item_count": 47, + "soft_pair_count": 38, + "serialized_chars": 37818, + "estimated_input_tokens": 9455, + "estimated_eval_tokens": 19166 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "The cluster is mostly a grab bag of unrelated Transformers issues, but several are centered on v5 loading/import regressions. The broadest and most representative issue is the checkpoint reload regression after upgrading from v4.57 to v5.x.", - "confidence": 0.58, - "canonical_issue_reason": "Issue 44556 is the clearest, broad user-facing v5 regression in the set: checkpoints trained on v4.57 fail to reload on v5.2/v5.3. It is more central than the narrower model-, tokenizer-, or docs-specific reports.", + "summary": "This cluster is mostly a grab bag of unrelated Transformers regressions, import errors, model-loading issues, and docs/test bugs. I do not see any soft pair that is convincingly the same underlying defect, so I would not merge any of these issues as duplicates.", + "confidence": 0.93, + "canonical_issue_reason": "If a single anchor is needed, #45850 is the broadest umbrella issue for the v5 breakage theme, but the cluster is too heterogeneous for a true duplicate canonical.", "canonical_pr_reason": null, - "best_issue_reason": "44556 is the best overall issue to anchor the cluster because it describes a concrete, high-impact version-upgrade loading failure with good duplicate potential across nearby regressions.", + "best_issue_reason": "#45850 is the most general and central issue in the set, so it is the best umbrella candidate even though it is not a strong duplicate match for the rest.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:44387", - "right": "issue:44556", - "accept": false, - "reason": "Different failures: int4 quantization OOM/memory pressure vs checkpoint reload breakage after version upgrade." - }, - { - "left": "issue:43576", - "right": "issue:46132", - "accept": false, - "reason": "Unrelated bugs: CLI env command regression vs AttentionInterface.register changing function behavior." - }, - { - "left": "issue:43723", - "right": "issue:44556", - "accept": false, - "reason": "Both are loading-related, but tokenizer loading and checkpoint reload are different code paths and likely different bugs." - }, { "left": "issue:42890", "right": "issue:45341", "accept": false, - "reason": "Both mention testing, but one is a flaky integration test and the other is a generic testing_utils bug with no clear shared underlying defect." + "reason": "Different bugs: seed instability in a SAM HQ test vs a utility bug in testing_utils.py." }, { - "left": "issue:44222", - "right": "issue:45125", + "left": "issue:42548", + "right": "issue:44351", "accept": false, - "reason": "Different MoE issues: FP8 save_pretrained behavior vs missing tensor-parallel plan for a specific model class." + "reason": "Different missing symbols/import paths: PreTrainedModel vs HybridCache." }, { - "left": "issue:42548", - "right": "issue:44351", + "left": "issue:44222", + "right": "issue:45125", "accept": false, - "reason": "Both are import errors, but for different exported symbols and likely different missing-export regressions." + "reason": "Different FP8/serialization bug vs tensor-parallel `_tp_plan` loading bug." }, { "left": "issue:43010", "right": "issue:43065", "accept": false, - "reason": "Unrelated: no_grad/cache update semantics vs a dummy Conv2d in Sam3PixelDecoder." + "reason": "Cache `update()` no_grad concern is unrelated to the dummy Conv2d in Sam3PixelDecoder." }, { "left": "issue:44355", "right": "issue:44749", "accept": false, - "reason": "Different symptoms and subsystems: compiled Python file execution errors vs a performance regression during filtering." + "reason": "Compiled Python file errors and a data-filtering slowdown are unrelated." }, { "left": "issue:43502", "right": "issue:44485", "accept": false, - "reason": "No shared bug: offline/local-file request handling vs GLM-5 RoPE implementation." + "reason": "Offline/local-files-only request leakage is unrelated to GLM-5 RoPE implementation." }, { "left": "issue:45125", "right": "issue:45478", "accept": false, - "reason": "Same model family, but one is a missing _tp_plan/tensor-parallel issue and the other is a generic from_pretrained error; not enough to treat as the same bug." + "reason": "Both mention Qwen3.5 MoE, but one is tensor-parallel `_tp_plan` and the other is a broader from_pretrained failure; not clearly the same defect." }, { "left": "issue:30333", "right": "issue:33453", "accept": false, - "reason": "Training-job MLflow reporting vs tokenizer loading regression are unrelated." + "reason": "MLFlow failure reporting and tokenizer loading regression are different code paths." }, { "left": "issue:43010", "right": "issue:43232", "accept": false, - "reason": "Different generation/cache issues: no_grad decoration for cache updates vs _update_model_kwargs_for_generation after sync_gpus." + "reason": "`no_grad` decoration for cache update is not the same as the generation/sync_gpus kwargs issue." }, { "left": "issue:45125", "right": "issue:45375", "accept": false, - "reason": "Both are Qwen3_5Moe-related, but one is tensor parallelism and the other is a missing config field silently dropped by strict parsing." + "reason": "Tensor-parallel plan bug and missing vision config field are different model/config issues." }, { "left": "issue:30990", "right": "issue:31795", "accept": false, - "reason": "Unrelated: model loading hangs vs documentation confusion for forward arguments." + "reason": "Loading hang in Sentence Transformers is unrelated to a documentation wording issue." }, { "left": "issue:43502", "right": "issue:44222", "accept": false, - "reason": "Different code paths: offline API suppression vs FP8 MoE save_pretrained behavior." + "reason": "Local-files-only network leakage is unrelated to FP8 save_pretrained on MoE models." }, { "left": "issue:43299", "right": "issue:44556", "accept": false, - "reason": "Both are v5 loading regressions, but they target different models and likely different failure causes." + "reason": "Both are loading-related, but they concern different versions/models and different failure modes." }, { "left": "issue:42831", "right": "issue:43874", "accept": false, - "reason": "Different bugs: FP8 accuracy issue vs missing image patch-count method in an image processor." + "reason": "FP8 accuracy drift and a missing image-patch helper are separate problems." }, { "left": "issue:44315", "right": "issue:44360", "accept": false, - "reason": "Unrelated: Liger Kernel application issue vs a DSA indexer ReLU discussion." + "reason": "Liger Kernel/model_init behavior is unrelated to the DSA ReLU discussion." }, { "left": "issue:43502", "right": "issue:44961", "accept": false, - "reason": "No substantive overlap; local_files_only behavior is unrelated to the placeholder racoon issue." + "reason": "Local-files-only API leakage and a one-word issue are unrelated." }, { "left": "issue:30333", "right": "issue:36296", "accept": false, - "reason": "Training failure reporting vs tensor-parallel training bug are different problems." + "reason": "MLFlow reporting on failed jobs is unrelated to tensor-parallel training." }, { - "left": "issue:43827", - "right": "issue:44961", + "left": "issue:44355", + "right": "issue:45071", "accept": false, - "reason": "Docs mentioning pipeline() after removals is unrelated to the racoon issue." + "reason": "Compiled Python file execution errors and PretrainedConfig type checking are unrelated." }, { - "left": "issue:44355", - "right": "issue:45071", + "left": "issue:43827", + "right": "issue:44961", "accept": false, - "reason": "Compiled-file execution errors and PretrainedConfig type checking are unrelated regressions." + "reason": "Docs still referencing pipeline() and a placeholder issue are unrelated." }, { "left": "issue:44355", "right": "issue:44655", "accept": false, - "reason": "Different issues: compiled Python file handling vs inability to save Pipeline objects." + "reason": "Compiled-file runtime errors and inability to save Pipeline objects are different issues." }, { "left": "issue:33453", "right": "issue:34689", "accept": false, - "reason": "Both are loading regressions, but one is tokenizer loading and the other is model loading for a specific Llama vision model." + "reason": "Tokenizer-loading regression and Llama 3.2 vision model loading breakage are distinct." }, { "left": "issue:42757", "right": "issue:43576", "accept": false, - "reason": "Different breakages: huggingface_hub export/import issue vs v5 env command regression." + "reason": "Importing `is_offline_mode` and the `transformers env` CLI failure are related at most by version churn, not the same concrete bug." }, { - "left": "issue:43502", + "left": "issue:44222", "right": "issue:44230", "accept": false, - "reason": "Offline request suppression is unrelated to fp8 inference support for Qwen3-VL/Qwen3.5 MoE." + "reason": "FP8 MoE save_pretrained and FP8 inference support are different requests/failures." }, { "left": "issue:43576", "right": "issue:45071", "accept": false, - "reason": "Both are v5 breakages, but one is the env command and the other is PretrainedConfig typing; not the same bug." + "reason": "Broken CLI env command and PretrainedConfig type checking are unrelated." }, { - "left": "issue:43576", - "right": "issue:43824", + "left": "issue:43824", + "right": "issue:44961", "accept": false, - "reason": "Different failures: broken env command vs missing Qwen2_5_VLForConditionalGeneration export." + "reason": "Qwen2.5-VL import failure and the placeholder issue are unrelated." }, { "left": "issue:41628", "right": "issue:42548", "accept": false, - "reason": "Different missing exports: AutoImageProcessor vs PreTrainedModel." + "reason": "Different missing imports (`AutoImageProcessor` vs `PreTrainedModel`)." }, { "left": "issue:31795", "right": "issue:38617", "accept": false, - "reason": "Docs confusion about forward args is unrelated to an ImportError for layer_type_validation." + "reason": "Docs confusion and an import error for `layer_type_validation` are not the same bug." }, { "left": "issue:44961", "right": "issue:45468", "accept": false, - "reason": "No shared underlying bug; racoon is unrelated to Gemma-4 audio positional encoding." + "reason": "Placeholder issue and Gemma-4 audio positional encoding bug are unrelated." }, { "left": "issue:38617", "right": "issue:43576", "accept": false, - "reason": "ImportError for a utility symbol is not the same as the broken env command." + "reason": "Import error for `layer_type_validation` is not the same as the broken env command." }, { "left": "issue:46143", "right": "issue:46144", "accept": false, - "reason": "Related RoFormer area, but one is kwargs propagation and the other is attention-interface usage; not clearly the same concrete bug." + "reason": "Both are RoFormer-related, but one is kwargs forwarding and the other is attention-interface usage; not the same concrete defect." + }, + { + "left": "issue:44908", + "right": "issue:44961", + "accept": false, + "reason": "Scheduler kwargs handling and a placeholder issue are unrelated." + }, + { + "left": "issue:44485", + "right": "issue:44961", + "accept": false, + "reason": "GLM-5 RoPE implementation and a placeholder issue are unrelated." + }, + { + "left": "issue:43673", + "right": "issue:44492", + "accept": false, + "reason": "GenerationMixin cache-missing behavior and a cache-strategy typo are different issues." + }, + { + "left": "issue:36296", + "right": "issue:43901", + "accept": false, + "reason": "Tensor-parallel training bug and TextClassificationPipeline docs mismatch are unrelated." + }, + { + "left": "issue:41628", + "right": "issue:44355", + "accept": false, + "reason": "Importing `AutoImageProcessor` and compiled-file execution errors are unrelated." + }, + { + "left": "issue:36296", + "right": "issue:43526", + "accept": false, + "reason": "Tensor-parallel training and BeitImageProcessorFast label reduction are different code paths." + }, + { + "left": "issue:45362", + "right": "issue:45850", + "accept": false, + "reason": "Qwen3.5 chat crash is a specific model bug; #45850 is a broad meta issue about many v5 breakages, not the same defect." } ] }, @@ -15040,7 +15047,7 @@ "component_index": 1, "component_count": 17, "review_unit_index": 48, - "review_unit_count": 50, + "review_unit_count": 49, "status": "reviewed", "reason": null, "source": "llm", @@ -15088,8 +15095,8 @@ "issue:45125", "issue:45230", "issue:45278", + "issue:45310", "issue:45341", - "issue:45362", "issue:45468", "issue:45478", "issue:45542", @@ -15098,14 +15105,13 @@ "issue:46095" ], "soft_pairs": [ - "issue:45362|issue:45850", "issue:44623|issue:45230", "issue:44273|issue:45542", "issue:44273|issue:45003", "issue:44038|issue:45125", "issue:44393|issue:45071", - "issue:44393|issue:45753", "issue:44230|issue:46095", + "issue:44393|issue:45753", "issue:43576|issue:44556", "issue:43525|issue:45850", "issue:41669|issue:43673", @@ -15136,228 +15142,265 @@ "issue:44315|issue:44829", "issue:43576|issue:44273", "issue:44360|issue:45468", - "issue:44655|issue:44861" + "issue:44655|issue:44861", + "issue:45230|issue:45310" ], - "prepared_review_unit_hash": "ce64b5014b463eadf06df366b01c7c716c56a10b6e3ded32d2ec58aceb8db9f1", + "prepared_review_unit_hash": "a5464ac8b45fa8ec5eacbec5038d45724344def9fcabc9fba2ca76df419c82ce", "component_budget": { "node_count": 426, "item_count": 426, - "soft_pair_count": 1572, - "serialized_chars": 814130, - "estimated_input_tokens": 203533, - "estimated_eval_tokens": 407322 + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 }, "budget": { "node_count": 48, "item_count": 48, "soft_pair_count": 39, - "serialized_chars": 38833, - "estimated_input_tokens": 9709, - "estimated_eval_tokens": 19674 + "serialized_chars": 38845, + "estimated_input_tokens": 9712, + "estimated_eval_tokens": 19680 }, "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 regression reports: import/lazy-loading breakages, model-specific config/load errors, cache/generation issues, and a few docs/test-only items. They share a broad release-regression theme, but none of the listed soft pairs look like the same underlying bug/change.", - "confidence": 0.9, - "canonical_issue_reason": "issue:45278 is the best umbrella issue: it captures the broad post-upgrade breakage theme ('many import errors after update from 4.57.0 to 5.5.0') and is more concrete than the very generic 'breaks too many models' report.", + "summary": "This cluster is mostly a grab bag of unrelated closed issues, with only a loose common theme of Transformers v5 regressions and loading/import breakages. I found no pair that clearly describes the exact same underlying bug/change.", + "confidence": 0.77, + "canonical_issue_reason": "issue:45278 is the best umbrella representative: it is a broad, concrete regression report about upgrading from 4.57.0 to 5.5.0 and getting many import errors, which matches the cluster\u2019s overall v5 breakage theme better than the more generic or highly specific issues.", "canonical_pr_reason": null, - "best_issue_reason": "issue:45278 is the strongest global issue candidate because it is broad enough to represent the cluster while still describing a concrete, user-visible regression after a version upgrade.", + "best_issue_reason": "issue:45278 has the clearest high-level scope and is the most representative of the repeated upgrade/regression complaints in this set.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:45362", - "right": "issue:45850", - "accept": false, - "reason": "Both mention Qwen/model breakage, but one is a specific chat crash and the other is a generic complaint about v5. They are not the same underlying bug." - }, { "left": "issue:44623", "right": "issue:45230", "accept": false, - "reason": "One is about processor.save_pretrained missing files; the other is an empty 'Bug report' with no concrete matching failure. Not enough evidence of the same bug." + "reason": "44623 is a specific processor.save_pretrained missing-files bug; 45230 is just a generic bug report with no evidence it targets the same failure." }, { "left": "issue:44273", "right": "issue:45542", "accept": false, - "reason": "Lazy loading/import mechanics vs a TensorFlow backend error from tensorboard-only installs are different failures." + "reason": "One is a lazy-loading correctness problem; the other is a TensorFlow/tensorboard backend error. Different code paths and symptoms." }, { "left": "issue:44273", "right": "issue:45003", "accept": false, - "reason": "Both touch import/loading behavior, but one reports lazy loading not working and the other unsafe sys.modules access; related area, not the same bug." + "reason": "Both touch import/loading behavior, but 44273 is about lazy loading not working while 45003 is about unsafe sys.modules access; not the same concrete bug." }, { "left": "issue:44038", "right": "issue:45125", "accept": false, - "reason": "Different models and different symptoms: Qwen3-VL-Moe bug vs missing _tp_plan for tensor parallelism." + "reason": "Same general model family, but one is a broad Qwen3-VL-Moe bug and the other is a missing tensor-parallel plan attribute. Different fixes." }, { "left": "issue:44393", "right": "issue:45071", "accept": false, - "reason": "Qwen3-VL bounding-box output error is unrelated to PretrainedConfig type-checking regressions." + "reason": "Qwen3-VL bbox/output issues are unrelated to a PretrainedConfig type-checking regression." }, { - "left": "issue:44393", - "right": "issue:45753", + "left": "issue:44230", + "right": "issue:46095", "accept": false, - "reason": "Both are Qwen3-VL-family issues, but one is bbox output corruption and the other is an infinite loop on a specific image; different concrete bugs." + "reason": "FP8 inference support for Qwen models is unrelated to a DeepSeekV4 weight-conversion question." }, { - "left": "issue:44230", - "right": "issue:46095", + "left": "issue:44393", + "right": "issue:45753", "accept": false, - "reason": "FP8 inference support for Qwen models and a DeepSeek weight-conversion question are unrelated." + "reason": "Both are Qwen3-related, but one is bbox output corruption and the other is an infinite loop on a specific image; not the same underlying bug." }, { "left": "issue:43576", "right": "issue:44556", "accept": false, - "reason": "The env command regression and checkpoint reload incompatibility are separate v5 issues." + "reason": "Both are v5 regressions, but one breaks the transformers env command and the other breaks checkpoint reload after upgrade. Different failures." }, { "left": "issue:43525", "right": "issue:45850", "accept": false, - "reason": "A missing pad_token_id on Llama4Config is a specific config bug, not the same as a generic 'many models break' report." + "reason": "A missing pad_token_id on Llama4Config is a specific config bug, not the same as the broad 'too many models break in v5' complaint." }, { "left": "issue:41669", "right": "issue:43673", "accept": false, - "reason": "Import-star cleanup/performance complaint and a missing GenerationMixin cache in chunked prefill are different problems." + "reason": "Import-star slowdowns and a missing GenerationMixin cache in chunked_prefill are unrelated issues." }, { "left": "issue:44556", "right": "issue:45092", "accept": false, - "reason": "Both are upgrade/load failures, but one is checkpoint reload compatibility and the other is remote-code/meta-init incompatibility; not the same issue." + "reason": "Both involve v5 compatibility, but one is generic checkpoint reload failure and the other is remote-code/meta-init incompatibility for InternVL2. Not the same bug." }, { "left": "issue:43572", "right": "issue:45850", "accept": false, - "reason": "Missing pad_token_idx in StableLmConfig is a model-specific config regression, not the same as a broad v5-breakage umbrella." + "reason": "StableLmConfig pad_token_idx loss is a model-specific config regression, not the same as a broad v5 breakage report." }, { "left": "issue:44492", "right": "issue:45850", "accept": false, - "reason": "A typo in cache strategies is too narrow and unrelated to the broad 'breaks too many models' complaint." - }, - { - "left": "issue:42371", - "right": "issue:44492", - "accept": false, - "reason": "TF32 API guidance and a cache-strategy typo are unrelated." + "reason": "A typo in cache strategies is too specific and unrelated to the broad model-breakage complaint." }, { "left": "issue:45278", "right": "issue:45850", "accept": false, - "reason": "Both are broad v5 regressions, but one is specifically about import errors after upgrade while the other is a generic complaint about many models. They overlap in theme, not the same bug." + "reason": "Both are broad upgrade complaints, but one is specifically about import errors and the other about too many models breaking; similar theme, not the same underlying issue." }, { "left": "issue:44246", "right": "issue:44273", "accept": false, - "reason": "Slow imports and broken lazy loading are related symptoms, but not the same underlying defect." + "reason": "Import slowness and lazy loading malfunction are related only at a high level; they are not the same concrete defect." }, { "left": "issue:42890", "right": "issue:44908", "accept": false, - "reason": "A missing set_seed in tests and an inverse_sqrt scheduler kwarg bug are unrelated." + "reason": "Missing set_seed in tests and ignored lr_scheduler_kwargs are unrelated." }, { "left": "issue:44704", "right": "issue:44908", "accept": false, - "reason": "AutoProcessor.from_pretrained kwarg propagation and scheduler timescale handling are different code paths." + "reason": "AutoProcessor kwargs propagation and scheduler kwargs handling are different API surfaces and bugs." }, { "left": "issue:43408", "right": "issue:44315", "accept": false, - "reason": "A model-type mismatch warning and Liger kernel application with model_init are unrelated." + "reason": "A model-type warning for sam3_* and Liger Kernel not applied with model_init are unrelated." }, { "left": "issue:43976", "right": "issue:44393", "accept": false, - "reason": "Python version support problems and a Qwen3-VL output bug are not the same issue." + "reason": "Python version compatibility and Qwen3-VL bbox output errors are unrelated." }, { "left": "issue:43352", "right": "issue:44393", "accept": false, - "reason": "Flash Attention 2 support gating for Nemotron is unrelated to Qwen3-VL bounding-box failures." + "reason": "Flash Attention 2 support in NemotronH vs Qwen3-VL output issues are different code paths." }, { "left": "issue:44297", "right": "issue:45230", "accept": false, - "reason": "Tokenizer class mismatch on save_pretrained is unrelated to a generic bug report placeholder." + "reason": "Tokenizer save_pretrained mismatch is a concrete serialization bug; 45230 is too generic to match." }, { "left": "issue:43576", "right": "issue:44908", "accept": false, - "reason": "A broken env command and a scheduler kwarg bug are different failures." + "reason": "transformers env CLI breakage and inverse_sqrt scheduler kwargs handling are unrelated." + }, + { + "left": "issue:44295", + "right": "issue:45230", + "accept": false, + "reason": "Position_ids buffer access failure does not match the generic bug report." + }, + { + "left": "issue:44908", + "right": "issue:45071", + "accept": false, + "reason": "Scheduler kwargs handling and PretrainedConfig type checking are unrelated." }, { "left": "issue:41669", "right": "issue:44351", "accept": false, - "reason": "Import-star cleanup and missing HybridCache import are both import-related, but not the same bug." + "reason": "Import-star performance regression and missing HybridCache export are different import-related bugs." }, { "left": "issue:41669", "right": "issue:45341", "accept": false, - "reason": "Import-star usage in models and a testing_utils bug are unrelated." + "reason": "A slow import regression and a small testing_utils bug are not the same issue." }, { "left": "issue:44492", "right": "issue:44945", "accept": false, - "reason": "A cache-strategy typo and incorrect pipeline-parallel output are different code paths." + "reason": "Cache strategy typo and pipeline-parallelism output errors are unrelated." }, { "left": "issue:43874", "right": "issue:45850", "accept": false, - "reason": "Missing image-processor method for GLM46V is a specific multimodal bug, not the same as a broad v5 regression complaint." + "reason": "A missing image-patch helper in one processor is too specific to be the same as a broad v5 breakage report." }, { "left": "issue:44393", "right": "issue:45478", "accept": false, - "reason": "Different model families and different failures: Qwen3-VL bbox output vs Qwen3.5 Moe from_pretrained error." + "reason": "Qwen3-VL bounding-box hallucination/error and Qwen3.5-Moe from_pretrained failure are different bugs." }, { "left": "issue:44077", "right": "issue:44861", "accept": false, - "reason": "Optional post_init validation and _tied_weights_keys handling are unrelated." + "reason": "Optional post_init policy and tied-weight-key AttributeError are separate model-loading issues." }, { "left": "issue:43827", "right": "issue:44908", "accept": false, - "reason": "Docs still referencing pipeline() and a scheduler kwargs bug do not share the same underlying defect." + "reason": "Docs still referencing pipeline() and scheduler kwargs handling are unrelated." }, { "left": "issue:41669", "right": "issue:44908", "accept": false, - "reason": "Import-star cleanup/performance and inverse_sqrt scheduler kwargs are unrelated." + "reason": "Import-star slowdown and scheduler kwargs handling do not describe the same bug." + }, + { + "left": "issue:31795", + "right": "issue:33453", + "accept": false, + "reason": "Documentation confusion around forward args and tokenizer loading regression are unrelated." + }, + { + "left": "issue:44315", + "right": "issue:44829", + "accept": false, + "reason": "Liger Kernel not applied with model_init and degenerate training under flash_attention_3 are different concrete problems." + }, + { + "left": "issue:43576", + "right": "issue:44273", + "accept": false, + "reason": "A broken env command and lazy-loading malfunction are different symptoms and likely different fixes." + }, + { + "left": "issue:44360", + "right": "issue:45468", + "accept": false, + "reason": "DSA indexer discussion and Gemma-4 audio positional encoding are unrelated." + }, + { + "left": "issue:44655", + "right": "issue:44861", + "accept": false, + "reason": "Pipeline save_pretrained failure and tied-weight-key AttributeError are different serialization/model-loading bugs." + }, + { + "left": "issue:45230", + "right": "issue:45310", + "accept": false, + "reason": "Generic bug report versus a specific Qwen3.5-Moe from_pretrained regression; no evidence they are the same issue." } ] }, @@ -15373,7 +15416,7 @@ "component_index": 1, "component_count": 17, "review_unit_index": 49, - "review_unit_count": 50, + "review_unit_count": 49, "status": "reviewed", "reason": null, "source": "llm", @@ -15388,6 +15431,7 @@ "issue:34689", "issue:36296", "issue:37428", + "issue:37934", "issue:38617", "issue:41669", "issue:42371", @@ -15419,7 +15463,6 @@ "issue:44908", "issue:44998", "issue:45092", - "issue:45230", "issue:45290", "issue:45310", "issue:45341", @@ -15431,7 +15474,6 @@ "issue:46172" ], "soft_pairs": [ - "issue:45230|issue:45310", "issue:46018|issue:46172", "issue:44273|issue:44908", "issue:43519|issue:44485", @@ -15476,383 +15518,320 @@ "issue:42936|issue:46144", "issue:42936|issue:43813", "issue:43240|issue:43668", - "issue:43668|issue:43813" + "issue:43668|issue:43813", + "issue:37934|issue:43668", + "issue:42936|issue:43668" ], - "prepared_review_unit_hash": "6012be437a33d4bc8529a323a59a1ff16efdea83bad23fc6900382031d6fea02", + "prepared_review_unit_hash": "e851ffd5df1e8179b16c83ac8695695fc89847d6112ada97b8b9e8f892c2d876", "component_budget": { "node_count": 426, "item_count": 426, - "soft_pair_count": 1572, - "serialized_chars": 814130, - "estimated_input_tokens": 203533, - "estimated_eval_tokens": 407322 + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 }, "budget": { "node_count": 48, "item_count": 48, - "soft_pair_count": 46, - "serialized_chars": 40652, - "estimated_input_tokens": 10163, - "estimated_eval_tokens": 20582 + "soft_pair_count": 47, + "serialized_chars": 40969, + "estimated_input_tokens": 10243, + "estimated_eval_tokens": 20742 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a loose set of Transformers v5 regression reports, especially model-loading/tokenizer/config breakages, but the individual issues are still distinct. No pull requests are present.", - "confidence": 0.51, - "canonical_issue_reason": "Issue 45850 is the best umbrella for the cluster because it explicitly describes broad Transformers 5.0 model breakage, matching the shared regression theme better than the narrower, model-specific reports.", + "summary": "This is a noisy similarity cluster, mostly mixing assorted Transformers v5/model-loading regressions with unrelated tokenizer, scheduler, vision, and docs issues. The broadest shared theme is version-5 compatibility breakage, but none of the soft pairs are true duplicates or mergeable PR-equivalents.", + "confidence": 0.66, + "canonical_issue_reason": "Issue 45850 is the best umbrella issue because it is the only broadly scoped report capturing the shared theme of Transformers 5.0 breaking many models, while the others are narrower, model-specific, or unrelated subsystem bugs.", "canonical_pr_reason": null, - "best_issue_reason": "45850 is the strongest anchor for duplicate triage: open, broad, and central to the repeated v5 compatibility complaints across the cluster.", + "best_issue_reason": "45850 is the most representative cluster anchor: it generalizes the common v5 regression pattern better than the specific loading/config/tokenizer bugs elsewhere in the set.", "best_pr_reason": null, "soft_edge_verdicts": [ - { - "left": "issue:45230", - "right": "issue:45310", - "accept": false, - "reason": "Too generic to treat as the same bug; 45310 is a specific Qwen3.5 MoE from_pretrained failure, while 45230 has no concrete matching code-path evidence." - }, { "left": "issue:46018", "right": "issue:46172", "accept": false, - "reason": "Both mention DeepSeek-V4, but one is a gating logic complaint and the other is a dtype/transpose bug; different underlying problems." + "reason": "Different DeepSeek-v4 symptoms: shared expert gating vs a tensor transpose bug." }, { "left": "issue:44273", "right": "issue:44908", "accept": false, - "reason": "Lazy-loading behavior and inverse_sqrt scheduler kwargs are unrelated subsystems." + "reason": "Lazy-loading behavior and inverse_sqrt scheduler kwargs are unrelated." }, { "left": "issue:43519", "right": "issue:44485", "accept": false, - "reason": "Qwen3VL timestamp math and GLM-5 RoPE implementation are different model-specific bugs." + "reason": "Qwen3VL timestamp parsing and GLM-5 RoPE implementation are different bugs." }, { "left": "issue:43519", "right": "issue:44908", "accept": false, - "reason": "Processor timestamp calculation is unrelated to scheduler kwargs handling." + "reason": "Processor timestamp math is unrelated to scheduler kwargs handling." }, { "left": "issue:31795", "right": "issue:34689", "accept": false, - "reason": "Documentation confusion and a model-loading regression are not the same concrete bug." + "reason": "Documentation confusion and Llama 3.2 model-loading regression are not the same issue." }, { "left": "issue:41669", "right": "issue:43704", "accept": false, - "reason": "Import-star cleanup/performance is a codebase-wide refactor; VRAM leakage in Qwen3ForCausalLM is a separate runtime bug." + "reason": "Import-star performance cleanup is unrelated to a VRAM leak in multi-threaded inference." }, { "left": "issue:31795", "right": "issue:37428", "accept": false, - "reason": "Docs clarification for model.forward is unrelated to a flash-attention import error." + "reason": "Docs wording for forward args and a missing internal flash-attn import are different problems." }, { "left": "issue:44315", "right": "issue:45092", "accept": false, - "reason": "Both concern v5-era initialization flows, but one is about model_init not applying Liger Kernel and the other about remote-code checkpoint/meta-init incompatibility." + "reason": "Liger Kernel application with model_init and InternVL2 meta-initialization incompatibility are separate integration bugs." }, { "left": "issue:44485", "right": "issue:45468", "accept": false, - "reason": "GLM-5 RoPE math and Gemma-4 audio positional encoding are different model implementations." + "reason": "GLM-5 RoPE and Gemma-4 audio positional encoding affect different models and code paths." }, { "left": "issue:43704", "right": "issue:44485", "accept": false, - "reason": "VRAM leakage across dataloader threads is not the same as RoPE implementation correctness." + "reason": "VRAM leak in Qwen3ForCausalLM is unrelated to GLM-5 RoPE math." }, { "left": "issue:43298", "right": "issue:46095", "accept": false, - "reason": "Continuous batching support for audio models and a DeepSeek-V4 weight-conversion question are unrelated." + "reason": "Continuous batching for audio models and a DeepSeek-V4 weight-conversion question are not duplicates." }, { "left": "issue:43502", "right": "issue:43519", "accept": false, - "reason": "local_files_only network leakage and timestamp calculation are different failure modes." + "reason": "local_files_only network leakage and timestamp calculation in a processor are different failures." }, { "left": "issue:42371", "right": "issue:43704", "accept": false, - "reason": "TF32 API guidance is unrelated to Qwen3ForCausalLM VRAM leakage." + "reason": "TF32 behavior settings and a Qwen3 memory leak do not match." }, { "left": "issue:43519", "right": "issue:44623", "accept": false, - "reason": "Timestamp calculation and processor.save_pretrained missing files are separate processor issues." + "reason": "Timestamp calculation and processor.save_pretrained missing files are different processor bugs." }, { "left": "issue:43519", "right": "issue:43704", "accept": false, - "reason": "Processor timestamp math does not match the multi-thread VRAM leak." + "reason": "Processor timestamp math and VRAM leakage are unrelated." }, { "left": "issue:43606", "right": "issue:45468", "accept": false, - "reason": "CPU offload device mismatch for suno/bark-small is different from Gemma4AudioRelPositionalEncoding." + "reason": "CPU-offload device mismatch and Gemma-4 audio positional encoding are different model paths." }, { "left": "issue:44263", "right": "issue:45468", "accept": false, - "reason": "torch.split return handling in GlmMoeDsaIndexer is unrelated to Gemma-4 audio encoding." + "reason": "torch.split return handling and Gemma-4 audio positional encoding are unrelated." }, { "left": "issue:44261", "right": "issue:45468", "accept": false, - "reason": "Missing rms_norm_eps in MLA q_a_layernorm is a precision/config bug, not the same as Gemma-4 audio positional encoding." + "reason": "MLA RMS-eps precision discussion is not the same as Gemma-4 audio encoding." }, { "left": "issue:42913", "right": "issue:43066", "accept": false, - "reason": "Both are tokenizer v5 complaints, but one is a behavior regression and the other is a decoder type mismatch; not the same bug." + "reason": "Tokenizer v4/v5 behavior differences and wrong decoder type are both tokenizer-related but distinct bugs with different fixes." }, { "left": "issue:44908", "right": "issue:44998", "accept": false, - "reason": "Scheduler kwargs handling is unrelated to the unrelated 'Unemployment' issue." + "reason": "Scheduler kwargs handling and an unrelated placeholder issue are not duplicates." }, { "left": "issue:44492", "right": "issue:44998", "accept": false, - "reason": "Cache-strategy typo and an unrelated placeholder issue are not duplicates." + "reason": "Cache-strategy typo and an unrelated placeholder issue are not the same bug." }, { "left": "issue:44861", "right": "issue:45290", "accept": false, - "reason": "Tie-weight-key crash and chat-template tool-call crash affect different code paths." + "reason": "Tied-weights helper crash and chat template tool-call crash affect different code paths." }, { "left": "issue:43618", "right": "issue:44998", "accept": false, - "reason": "CLIPOutput attentions regression is unrelated to the placeholder issue." + "reason": "CLIPOutput attentions regression and an unrelated placeholder issue are not duplicates." }, { "left": "issue:44998", "right": "issue:45850", "accept": false, - "reason": "The placeholder issue does not match the broad v5 model-breakage report." + "reason": "Placeholder issue text is not a real report about Transformers 5.0 breakages." }, { "left": "issue:44998", "right": "issue:45341", "accept": false, - "reason": "The placeholder issue is unrelated to a testing_utils bug." + "reason": "Placeholder issue and a testing_utils bug are unrelated." }, { "left": "issue:43525", "right": "issue:44861", "accept": false, - "reason": "Missing pad_token_id in Llama4Config and list-handling in tied weights are separate config/model-init bugs." + "reason": "Missing pad_token_id in Llama4Config and tied-weight-key crashes are different configuration/model bugs." }, { "left": "issue:43572", "right": "issue:44861", "accept": false, - "reason": "StableLmConfig pad_token_idx absence is unrelated to _get_tied_weight_keys list handling." + "reason": "StableLmConfig pad_token_idx regression and tied-weight-key crashes are unrelated." }, { "left": "issue:31515", "right": "issue:38617", "accept": false, - "reason": "Slow from_pretrained checkpoint loading is a performance issue; layer_type_validation import failure is a missing symbol bug." + "reason": "Slow checkpoint loading and a missing configuration_utils import are not the same bug." }, { "left": "issue:44230", "right": "issue:44623", "accept": false, - "reason": "fp8 support request and processor.save_pretrained missing files are different feature/bug areas." + "reason": "FP8 support request and missing processor save files are unrelated." }, { "left": "issue:30333", "right": "issue:31515", "accept": false, - "reason": "MLFlow job-state reporting and slow checkpoint loading are unrelated training/loading issues." + "reason": "MLflow job status reporting and slow from_pretrained checkpoint loading are different issues." }, { "left": "issue:31515", "right": "issue:36296", "accept": false, - "reason": "Slow checkpoint loading is unrelated to tensor-parallel training behavior." + "reason": "Checkpoint load slowness and tensor-parallel training bugs are separate topics." }, { "left": "issue:31515", "right": "issue:31795", "accept": false, - "reason": "Checkpoint load performance is not the same as documentation confusion in model.forward." + "reason": "Model load performance and forward-signature documentation do not describe the same defect." }, { "left": "issue:44230", "right": "issue:45310", "accept": false, - "reason": "fp8 support request and Qwen3.5 MoE from_pretrained failure are different topics." + "reason": "FP8 inference support and Qwen3.5 MoE from_pretrained error are not the same change." }, { "left": "issue:31515", "right": "issue:33453", "accept": false, - "reason": "A load-speed complaint is not the same as a tokenizer-loading regression." + "reason": "Slow checkpoint loading and tokenizer-loading regression are different load-path problems." }, { "left": "issue:43298", "right": "issue:44230", "accept": false, - "reason": "Continuous batching for audio models and fp8 inference support are unrelated." + "reason": "Audio continuous batching support and FP8 inference support are unrelated feature areas." }, { "left": "issue:43717", "right": "issue:46144", "accept": false, - "reason": "Mamba init_weights usage and RoFormer attention interface usage are distinct architecture issues." + "reason": "Mamba init_weights behavior and RoFormer attention interface usage are unrelated model implementations." }, { "left": "issue:42936", "right": "issue:46018", "accept": false, - "reason": "Mask2former ignore_value and DeepSeek-V4 gating are unrelated model bugs." + "reason": "Mask2Former ignore_value handling and DeepSeek-V4 expert gating are different model bugs." }, { "left": "issue:43668", "right": "issue:44031", "accept": false, - "reason": "Incorrect type hint for norm_eps is unrelated to tokenizer regex warning behavior." + "reason": "Config type-hint correctness and tokenizer regex warning behavior are unrelated." }, { "left": "issue:42936", "right": "issue:43240", "accept": false, - "reason": "Mask2former ignore_value is unrelated to kwargs not being passed into loss calculation." - }, - { - "left": "issue:43240", - "right": "issue:46144", - "accept": false, - "reason": "Loss-function kwargs forwarding and RoFormer attention interface usage are different code paths." - }, - { - "left": "issue:42936", - "right": "issue:44031", - "accept": false, - "reason": "Mask2former ignore_value and tokenizer regex warnings are unrelated." - }, - { - "left": "issue:42936", - "right": "issue:46144", - "accept": false, - "reason": "Mask2former ignore_value bug is unrelated to RoFormer attention implementation details." - }, - { - "left": "issue:42936", - "right": "issue:43813", - "accept": false, - "reason": "Mask2former ignore_value and a typo in quantization operation naming are not the same bug." - }, - { - "left": "issue:43240", - "right": "issue:43668", - "accept": false, - "reason": "Loss kwargs forwarding and a type-hint typo are different issues." - }, - { - "left": "issue:43668", - "right": "issue:43813", - "accept": false, - "reason": "Incorrect type hint and quantization-operation typo 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": 1, - "component_count": 17, - "review_unit_index": 50, - "review_unit_count": 50, - "status": "reviewed", - "reason": null, - "source": "llm", - "cache_hit": false, - "model": "gpt-5.4-mini?service_tier=flex", - "cluster_id": "cluster-10000-426", - "nodes": [ - "issue:37934", - "issue:42936", - "issue:43668" - ], - "soft_pairs": [ - "issue:37934|issue:43668", - "issue:42936|issue:43668" - ], - "prepared_review_unit_hash": "3ad2b0b93371467d4b0b03de9c7a5572ecae628f88f10067e0952b2bc5d5b16d", - "component_budget": { - "node_count": 426, - "item_count": 426, - "soft_pair_count": 1572, - "serialized_chars": 814130, - "estimated_input_tokens": 203533, - "estimated_eval_tokens": 407322 - }, - "budget": { - "node_count": 3, - "item_count": 3, - "soft_pair_count": 2, - "serialized_chars": 2313, - "estimated_input_tokens": 579, - "estimated_eval_tokens": 1414 - }, - "overflow_policy": "truncate_then_skip", - "trimmed": false, - "aggressively_trimmed": false, - "split": true, - "analyst_result": { - "summary": "These three closed issues are about different model/configuration topics and do not look like duplicates of one underlying bug.", - "confidence": 0.95, - "canonical_issue_reason": "Issue 43668 is the most self-contained and concrete report: a narrow type-hint defect in ModernBERTConfig, with a clear, isolated fix path.", - "canonical_pr_reason": null, - "best_issue_reason": "43668 is the cleanest representative of the cluster because it is specific, actionable, and least ambiguous.", - "best_pr_reason": null, - "soft_edge_verdicts": [ + "reason": "Mask2Former ignore_value usage and loss kwargs propagation are different training bugs." + }, + { + "left": "issue:43240", + "right": "issue:46144", + "accept": false, + "reason": "Loss-function kwargs forwarding and RoFormer attention interface usage are unrelated." + }, + { + "left": "issue:42936", + "right": "issue:44031", + "accept": false, + "reason": "Mask2Former ignore_value handling is unrelated to tokenizer regex warnings." + }, + { + "left": "issue:42936", + "right": "issue:46144", + "accept": false, + "reason": "Mask2Former ignore_value and RoFormer attention interface bugs do not share a code path." + }, + { + "left": "issue:42936", + "right": "issue:43813", + "accept": false, + "reason": "Mask2Former ignore_value handling is unrelated to the quantization typo." + }, + { + "left": "issue:43240", + "right": "issue:43668", + "accept": false, + "reason": "Loss kwargs forwarding and config type-hint correction are not the same issue." + }, + { + "left": "issue:43668", + "right": "issue:43813", + "accept": false, + "reason": "Incorrect type hint and quantization typo are different, unrelated fixes." + }, { "left": "issue:37934", "right": "issue:43668", "accept": false, - "reason": "Different components and different bug classes: Llama4TextL2Norm behavior/RMS-norm question versus ModernBERTConfig type-hint correctness. They do not describe the same underlying change." + "reason": "Llama4 norm question and ModernBERT type-hint issue are unrelated model-specific concerns." }, { "left": "issue:42936", "right": "issue:43668", "accept": false, - "reason": "Mask2former ignore_value being unused is a runtime/model-logic issue, while 43668 is a config annotation/type-hint issue. No shared concrete code-path or duplicate symptom." + "reason": "Mask2Former ignore_value bug and ModernBERT type-hint issue are unrelated." } ] }, @@ -15885,52 +15864,52 @@ "issue:10519|issue:11005", "issue:11005|issue:11307" ], - "prepared_review_unit_hash": "a98b6e4a56839f83b9481cabcd71029c820fe9ad2c5b6527a5098934d8ff710c", + "prepared_review_unit_hash": "ecc5c441706e5a9bac00c2a9e2e6ed5acb96f40a24c1e5a3ed8adcc419812bc4", "component_budget": { "node_count": 3, "item_count": 3, "soft_pair_count": 3, - "serialized_chars": 2876, - "estimated_input_tokens": 719, - "estimated_eval_tokens": 1694 + "serialized_chars": 2877, + "estimated_input_tokens": 720, + "estimated_eval_tokens": 1696 }, "budget": { "node_count": 3, "item_count": 3, "soft_pair_count": 3, - "serialized_chars": 2876, - "estimated_input_tokens": 719, - "estimated_eval_tokens": 1694 + "serialized_chars": 2877, + "estimated_input_tokens": 720, + "estimated_eval_tokens": 1696 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These three issues are not duplicates of each other: one is about tokenizer truncation behavior, one about ReduceLROnPlateau-like training logic, and one about Wav2Vec2 word boundary timestamps. The cluster appears to be a false positive based only on loose textual similarity.", - "confidence": 0.98, + "summary": "All three issues are about different features/topics: token truncation behavior, ReduceLROnPlateau-like training logic, and Wav2Vec2 word time offsets. They are not duplicates of the same underlying bug or change.", + "confidence": 0.99, "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": null, + "best_issue_reason": "Issue #10519 is the clearest self-contained feature request, but it is not a duplicate/canonical target for the others.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:10519", "right": "issue:11307", "accept": false, - "reason": "Different problem domains: truncation strategy for token sequences vs. extracting word start/end offsets from Wav2Vec2 outputs." + "reason": "Different domains and requests: truncation strategy vs. Wav2Vec2 word boundary offsets. No shared code path or concrete overlap." }, { "left": "issue:10519", "right": "issue:11005", "accept": false, - "reason": "No shared underlying bug or feature request; tokenizer truncation behavior is unrelated to ReduceLROnPlateau-like scheduler functionality." + "reason": "Unrelated functionality: sequence truncation behavior vs. scheduler-like training control. Similar wording only, not the same bug/change." }, { "left": "issue:11005", "right": "issue:11307", "accept": false, - "reason": "These describe unrelated requests in different subsystems: learning-rate scheduling vs. audio model timestamp extraction." + "reason": "Completely different features: learning-rate scheduler behavior vs. speech model time offsets. No plausible duplicate relationship." } ] }, @@ -15962,7 +15941,7 @@ "issue:13244|issue:16618", "issue:15980|issue:16618" ], - "prepared_review_unit_hash": "cff74c3abda48a8211e2ce4bedc36b9a8f1ae32dad7a14b1cf94560386382172", + "prepared_review_unit_hash": "c844e731f4dd665315b8f75e9aa9c463f0707f96b01778ce85d5826e4f407878", "component_budget": { "node_count": 3, "item_count": 3, @@ -15984,24 +15963,24 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "This cluster does not look like a true duplicate set overall: one issue is about Tapas tokenization, while the other two concern model loading / download errors. Of the remaining pair, the titles suggest related Hugging Face model-loading problems but not the same concrete bug.", - "confidence": 0.63, - "canonical_issue_reason": "issue:16618 is the broadest and most user-facing model-loading report, so it is the best representative for the cluster if one issue must be chosen.", + "summary": "One issue is unrelated tokenization behavior, while the other two are both model-loading failures but appear to target different failure modes rather than the same concrete bug.", + "confidence": 0.79, + "canonical_issue_reason": "Issue #16618 is the broadest and most central model-loading report in the cluster, so it is the best representative of the loading-failure theme.", "canonical_pr_reason": null, - "best_issue_reason": "issue:16618 has the most general, high-impact framing ('Can't load the model'), making it the best cluster anchor among these issues.", + "best_issue_reason": "#16618 best captures the shared model-loading symptom; #15980 is narrower and focused on a bad auth/error-message path, and #13244 is unrelated.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:13244", "right": "issue:16618", "accept": false, - "reason": "Tapas tokenization vs. BERT model loading are different subsystems and different bugs; the titles do not indicate the same underlying issue." + "reason": "Tokenization differences in TAPAS vs. a model-load failure are different bugs in different code paths." }, { "left": "issue:15980", "right": "issue:16618", "accept": false, - "reason": "Both involve model download/loading, but one is specifically about a bad auth error message for private models while the other is a generic failure to load a public model; not enough evidence of the same concrete bug." + "reason": "Both involve model loading, but #15980 is specifically about private-model login/error handling, while #16618 is a generic inability to load 'bert-base-uncased'; not clearly the same underlying bug." } ] }, @@ -16031,40 +16010,40 @@ "soft_pairs": [ "issue:26350|issue:33260" ], - "prepared_review_unit_hash": "6c79f34f4f8407a9a3bede626f70878dc985d723fa61857ca09f8f0ba2acf6dd", + "prepared_review_unit_hash": "d7f96afad994a04dedda35e173755ada1dd5f2e09a0fedd7a875873cd12cfffe", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1589, - "estimated_input_tokens": 398, - "estimated_eval_tokens": 1052 + "serialized_chars": 1588, + "estimated_input_tokens": 397, + "estimated_eval_tokens": 1050 }, "budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1589, - "estimated_input_tokens": 398, - "estimated_eval_tokens": 1052 + "serialized_chars": 1588, + "estimated_input_tokens": 397, + "estimated_eval_tokens": 1050 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are two separate community feature requests: one for Flash Attention 2 support across more architectures and one for GGUF support across more architectures. They share only a broad \"community contribution\" framing, not the same underlying change.", + "summary": "These are two separate community contribution tracking issues for different features: Flash Attention 2 support vs GGUF support. They are both broad architecture-support requests, but not the same underlying bug or change.", "confidence": 0.98, - "canonical_issue_reason": "No canonical issue: the requests target different feature areas and different code paths (Flash Attention 2 vs GGUF).", + "canonical_issue_reason": "Issue #26350 is the better canonical issue because it is older and appears to be the earlier open community-tracking thread, but it is only canonical for its own feature area, not for the pair.", "canonical_pr_reason": null, - "best_issue_reason": "Neither issue is a duplicate of the other; both are standalone feature requests, so there is no single best issue to represent a merged duplicate cluster.", + "best_issue_reason": "#26350 is the stronger issue candidate by age and discussion activity, but #33260 is equally valid as a separate tracker for a different feature request.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:26350", "right": "issue:33260", "accept": false, - "reason": "Different enhancement requests for unrelated functionality; they do not appear to describe the same bug or the same concrete change." + "reason": "Different requested features and code paths: Flash Attention 2 support for more architectures versus GGUF support for more architectures. Similar framing as community contributions is not enough to treat them as duplicates." } ] }, @@ -16112,124 +16091,124 @@ "issue:43834|issue:44365", "issue:44365|issue:44370" ], - "prepared_review_unit_hash": "8f8dfc1ff62a06ee509004a4389c066a90a654f2edd53a0c1dd655f9fd50429c", + "prepared_review_unit_hash": "f16fbed2cd573279a698a7c35051a0b971b06d85153c38f871de40c5aeadf51f", "component_budget": { "node_count": 6, "item_count": 6, "soft_pair_count": 15, - "serialized_chars": 8793, - "estimated_input_tokens": 2199, - "estimated_eval_tokens": 4654 + "serialized_chars": 8791, + "estimated_input_tokens": 2198, + "estimated_eval_tokens": 4652 }, "budget": { "node_count": 6, "item_count": 6, "soft_pair_count": 15, - "serialized_chars": 8793, - "estimated_input_tokens": 2199, - "estimated_eval_tokens": 4654 + "serialized_chars": 8791, + "estimated_input_tokens": 2198, + "estimated_eval_tokens": 4652 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are separate i18n documentation-translation requests, not duplicates. The high textual similarity is mostly from a shared issue template; the named language-specific issues are clearly different targets, and the placeholder-title issues are not safely mergeable into one underlying request.", - "confidence": 0.95, - "canonical_issue_reason": "issue:43834 is the most template-like and central-looking item, so it is the best representative if one must be chosen, but it does not subsume the other language-specific translation requests.", + "summary": "These are translation-request issues for different languages plus near-identical template-placeholder issues. They share a common documentation-translation pattern, but they are not duplicates of the same underlying bug/change.", + "confidence": 0.98, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "issue:43834 is the best cluster representative because it is generic/template-shaped and sits near the center of the similarity graph, though the cluster is not truly a duplicate set.", + "best_issue_reason": "issue 43834 is the best representative of the shared template because it is the generic placeholder form of the docs-translation request, but it is still not a duplicate of the language-specific issues.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:27088", "right": "issue:44370", "accept": false, - "reason": "Different translation requests; no evidence they target the same language or same underlying change." + "reason": "Different translation request instances; one is Turkish and the other is a generic placeholder template, so they are not the same underlying issue." }, { "left": "issue:27088", "right": "issue:44365", "accept": false, - "reason": "Different translation requests; similarity is from shared i18n template text, not the same issue." + "reason": "Different language-specific request vs. generic placeholder; only the template is shared." }, { "left": "issue:27088", "right": "issue:43834", "accept": false, - "reason": "Turkish translation request vs a generic placeholder/template issue; not the same underlying task." + "reason": "Turkish docs translation request and a generic placeholder issue are not the same concrete issue." }, { "left": "issue:27088", "right": "issue:38435", "accept": false, - "reason": "Turkish vs Romanian translation request; distinct language targets." + "reason": "Different target languages (Turkish vs Romanian), so these are separate translation requests, not duplicates." }, { "left": "issue:27088", "right": "issue:28936", "accept": false, - "reason": "Turkish vs Spanish translation request; clearly separate issues." + "reason": "Different target languages (Turkish vs Spanish); same topic, but distinct issues." }, { "left": "issue:28936", "right": "issue:43834", "accept": false, - "reason": "Spanish translation request vs placeholder/template issue; similarity is superficial." + "reason": "Spanish translation request and a placeholder template are not the same underlying issue." }, { "left": "issue:28936", "right": "issue:44365", "accept": false, - "reason": "Spanish translation request vs generic template-shaped issue; not the same underlying request." + "reason": "Spanish-specific request vs. generic placeholder; shared template only." }, { "left": "issue:28936", "right": "issue:38435", "accept": false, - "reason": "Spanish vs Romanian translation requests; different languages, not duplicates." + "reason": "Different target languages (Spanish vs Romanian), so not duplicates." }, { "left": "issue:28936", "right": "issue:44370", "accept": false, - "reason": "Spanish translation request vs generic template-shaped issue; no same-code-path change." + "reason": "Spanish translation request and a generic placeholder issue are distinct items." }, { "left": "issue:38435", "right": "issue:44370", "accept": false, - "reason": "Romanian translation request vs generic template-shaped issue; not the same request." + "reason": "Romanian-specific request vs. placeholder template; not the same issue." }, { "left": "issue:38435", "right": "issue:44365", "accept": false, - "reason": "Romanian translation request vs generic template-shaped issue; similarity is from shared boilerplate." + "reason": "Romanian translation request and a generic placeholder template are not duplicates." }, { "left": "issue:38435", "right": "issue:43834", "accept": false, - "reason": "Romanian translation request vs placeholder/template issue; distinct underlying task." + "reason": "Romanian-specific request vs. generic placeholder; only the translation-request template matches." }, { "left": "issue:43834", "right": "issue:44370", "accept": false, - "reason": "Both are template-shaped, but there is no proof they are the same language-specific request; too ambiguous to merge." + "reason": "Both are placeholder-style translation issues, but they are separate issue instances rather than one concrete bug/change." }, { "left": "issue:43834", "right": "issue:44365", "accept": false, - "reason": "Both are placeholder/template issues, but similarity is likely boilerplate rather than one shared translation request." + "reason": "Near-identical placeholder template, but not the same underlying request." }, { "left": "issue:44365", "right": "issue:44370", "accept": false, - "reason": "Both are placeholder/template issues; high similarity alone is not enough to treat them as the same underlying request." + "reason": "Two placeholder template issues; same form, but not a single duplicate report." } ] }, @@ -16308,250 +16287,250 @@ "issue:43837|issue:45177", "issue:45322|issue:46133" ], - "prepared_review_unit_hash": "41b40713c57d9bd0a9efacd0ae79791b291c440e140cae30882d307ce1a393e0", + "prepared_review_unit_hash": "121b167e370e3a0fe54c4c2373a23a6cf72bc855aa51d94d41c2eb239b63991a", "component_budget": { "node_count": 16, "item_count": 16, "soft_pair_count": 36, - "serialized_chars": 21660, - "estimated_input_tokens": 5415, - "estimated_eval_tokens": 11086 + "serialized_chars": 21655, + "estimated_input_tokens": 5414, + "estimated_eval_tokens": 11084 }, "budget": { "node_count": 16, "item_count": 16, "soft_pair_count": 36, - "serialized_chars": 21660, - "estimated_input_tokens": 5415, - "estimated_eval_tokens": 11086 + "serialized_chars": 21655, + "estimated_input_tokens": 5414, + "estimated_eval_tokens": 11084 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "This cluster is a loose topical group of model-addition requests, but the artifacts are for different models and often different modalities (vision, audio, OCR, ASR/TTS, BERT variants). They are not duplicates of one underlying bug/change, so no soft edges are accepted.", - "confidence": 0.95, + "summary": "These items are all broad model-request issues, but they do not describe the same underlying bug or change. The similarities are mostly topical (\"add model support\"), not duplicates.", + "confidence": 0.94, "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": null, + "best_issue_reason": "issue:43671 is the strongest representative because it is a concrete, active request with notable discussion activity, but it should still be treated as distinct from the other model requests.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:36817", "right": "issue:41211", "accept": false, - "reason": "EuroBert config vs DEIMv2 are unrelated model requests; only generic add-model similarity." + "reason": "Different model requests: EuroBert vs DEIMv2. Same broad theme, but not the same model or code change." }, { "left": "issue:36817", "right": "issue:43671", "accept": false, - "reason": "EuroBert model config and Qwen3-TTS support are different model families and modalities." + "reason": "EuroBert config support and Qwen3-TTS support are unrelated model additions with different architectures and tasks." }, { "left": "issue:36817", "right": "issue:36879", "accept": false, - "reason": "Both are add-model requests, but for different architectures (EuroBert vs RF-DETR), not the same change." + "reason": "EuroBert and RF-DETR are completely different models and use cases." }, { "left": "issue:36879", "right": "issue:41211", "accept": false, - "reason": "RF-DETR and DEIMv2 are distinct vision models with different implementation paths." + "reason": "RF-DETR and DEIMv2 are different vision models; not the same underlying request." }, { "left": "issue:36879", "right": "issue:43909", "accept": false, - "reason": "RF-DETR is a detection model; LFM2.5 Audio is an audio model request." + "reason": "RF-DETR is a detection model, while LFM2.5 Audio 1.5B is an audio model request." }, { "left": "issue:36879", "right": "issue:44121", "accept": false, - "reason": "RF-DETR and the weight-sparse transformer are different architectures and code paths." + "reason": "RF-DETR and OpenAI Weight-Sparse Transformer target different model families and code paths." }, { "left": "issue:36879", "right": "issue:43837", "accept": false, - "reason": "RF-DETR vs Qwen3-ASR are different model types and do not look mergeable into one fix." + "reason": "RF-DETR and Qwen3-ASR are unrelated; one is object detection, the other speech recognition." }, { "left": "issue:36879", "right": "issue:43671", "accept": false, - "reason": "RF-DETR and Qwen3-TTS are unrelated detection vs speech-model requests." + "reason": "RF-DETR and Qwen3-TTS are different tasks and model implementations." }, { "left": "issue:37015", "right": "issue:43837", "accept": false, - "reason": "NeoBERT and Qwen3-ASR are unrelated model families and modalities." + "reason": "NeoBERT and Qwen3-ASR are unrelated model requests; only the same generic \"add support\" pattern is shared." }, { "left": "issue:37015", "right": "issue:44121", "accept": false, - "reason": "NeoBERT and the OpenAI weight-sparse transformer are different concrete model implementations." + "reason": "NeoBERT and the weight-sparse transformer request are different architectures with different integration work." }, { "left": "issue:37015", "right": "issue:43909", "accept": false, - "reason": "NeoBERT vs LFM2.5 Audio are different model classes, not the same bug/change." + "reason": "NeoBERT is a text model request, while LFM2.5 Audio 1.5B is an audio model request." }, { "left": "issue:36879", "right": "issue:37015", "accept": false, - "reason": "RF-DETR and NeoBERT are distinct model additions with separate support code." + "reason": "RF-DETR and NeoBERT are unrelated model additions with no shared concrete bug or patch." }, { "left": "issue:37015", "right": "issue:42503", "accept": false, - "reason": "NeoBERT and ModernVBERT are both BERT-like, but they are different models and integration work." + "reason": "NeoBERT and ModernVBERT are distinct BERT-style model requests, not the same model or implementation." }, { "left": "issue:41115", "right": "issue:43837", "accept": false, - "reason": "MiniCPM3 and Qwen3-ASR are different model requests; one is general VLM, the other speech ASR." + "reason": "MiniCPM3 architecture support and Qwen3-ASR support are different model integrations." }, { "left": "issue:41115", "right": "issue:44121", "accept": false, - "reason": "MiniCPM3 and the sparse transformer request are unrelated architectures." + "reason": "MiniCPM3 and the weight-sparse transformer request are different architectures and not the same code-path change." }, { "left": "issue:41115", "right": "issue:42503", "accept": false, - "reason": "MiniCPM3 and ModernVBERT are different model additions, not one shared fix." + "reason": "MiniCPM3 and ModernVBERT are separate model requests, not duplicates." }, { "left": "issue:41115", "right": "issue:45027", "accept": false, - "reason": "MiniCPM3 and Voxtral TTS target different modalities and code paths." + "reason": "MiniCPM3 and Voxtral TTS target different model families and modalities." }, { "left": "issue:41211", "right": "issue:43671", "accept": false, - "reason": "DEIMv2 and Qwen3-TTS are unrelated vision vs speech requests." + "reason": "DEIMv2 and Qwen3-TTS are unrelated model additions." }, { "left": "issue:42503", "right": "issue:43837", "accept": false, - "reason": "ModernVBERT and Qwen3-ASR are different models and modalities." + "reason": "ModernVBERT and Qwen3-ASR are different model families and tasks." }, { "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 change." }, { "left": "issue:42503", "right": "issue:42738", "accept": false, - "reason": "ModernVBERT and BERT-like models with RoPE overlap only broadly, but they are not the same concrete request." + "reason": "ModernVBERT and the generic BERT-like RoPE request are related only by theme, not by a specific duplicate fix." }, { "left": "issue:42738", "right": "issue:45027", "accept": false, - "reason": "BERT-like RoPE support and Voxtral TTS are unrelated model features." + "reason": "A generic BERT-like RoPE request and Voxtral TTS support are unrelated." }, { "left": "issue:43671", "right": "issue:43837", "accept": false, - "reason": "Qwen3-TTS and Qwen3-ASR share a model prefix, but they are different speech tasks and likely separate implementations." + "reason": "Qwen3-TTS vs Qwen3-ASR are different modalities with separate implementations; same family, not same change." }, { "left": "issue:43671", "right": "issue:45027", "accept": false, - "reason": "Qwen3-TTS and Voxtral TTS are different TTS models, not one duplicate request." + "reason": "Qwen3-TTS support and Voxtral-4B-TTS support are different model requests, even though both are speech-related." }, { "left": "issue:43671", "right": "issue:45177", "accept": false, - "reason": "Qwen3-TTS and DeepSeek-OCR2 are different modalities and support areas." + "reason": "Qwen3-TTS and DeepSeek-OCR2 are unrelated capabilities and model code paths." }, { "left": "issue:43671", "right": "issue:44121", "accept": false, - "reason": "Qwen3-TTS and the sparse transformer request are unrelated model families." + "reason": "Qwen3-TTS and the weight-sparse transformer request are unrelated model additions." }, { "left": "issue:43671", "right": "issue:43909", "accept": false, - "reason": "Qwen3-TTS and LFM2.5 Audio are both audio-related, but they are distinct model additions." + "reason": "Qwen3-TTS and LFM2.5 Audio are both audio-related, but they are different model requests and not the same fix." }, { "left": "issue:43837", - "right": "issue:43909", + "right": "issue:44205", "accept": false, - "reason": "Qwen3-ASR and LFM2.5 Audio are different speech/audio model requests." + "reason": "Qwen3-ASR and SAM3-LiteText are different tasks and model integrations." }, { - "left": "issue:43837", + "left": "issue:44121", "right": "issue:45322", "accept": false, - "reason": "Qwen3-ASR and EUPE are unrelated speech vs perception-encoder requests." + "reason": "Weight-sparse transformer support and EUPE are distinct model requests with no shared concrete patch." }, { "left": "issue:43837", - "right": "issue:44121", + "right": "issue:45322", "accept": false, - "reason": "Qwen3-ASR and the sparse transformer request are different code paths and model types." + "reason": "Qwen3-ASR and EUPE are unrelated model families and use cases." }, { - "left": "issue:43909", - "right": "issue:44205", + "left": "issue:45322", + "right": "issue:46133", "accept": false, - "reason": "LFM2.5 Audio and SAM3-LiteText are different model families and modalities." + "reason": "EUPE and TIPSv2 are both vision-language related, but they are different architectures/papers and not the same change." }, { - "left": "issue:44121", - "right": "issue:45322", + "left": "issue:43837", + "right": "issue:43909", "accept": false, - "reason": "Sparse transformer support and EUPE are different architectures, not one combined fix." + "reason": "Qwen3-ASR and LFM2.5 Audio are distinct audio-model support requests, not duplicates." }, { "left": "issue:43837", - "right": "issue:45027", + "right": "issue:44121", "accept": false, - "reason": "Qwen3-ASR and Voxtral TTS are different speech models and tasks." + "reason": "Speech recognition support and a sparse-transformer model request are unrelated." }, { - "left": "issue:37015", - "right": "issue:45027", + "left": "issue:43837", + "right": "issue:45177", "accept": false, - "reason": "NeoBERT and Voxtral TTS are unrelated model support requests." + "reason": "Qwen3-ASR and DeepSeek-OCR2 are separate modalities and different model integrations." }, { - "left": "issue:43837", - "right": "issue:45177", + "left": "issue:37015", + "right": "issue:45027", "accept": false, - "reason": "Qwen3-ASR and DeepSeek-OCR2 are different modalities and implementation areas." + "reason": "NeoBERT and Voxtral TTS are different model families and tasks." }, { - "left": "issue:45322", - "right": "issue:46133", + "left": "issue:43909", + "right": "issue:44205", "accept": false, - "reason": "EUPE and TIPSv2 are separate vision-language model requests from different projects." + "reason": "LFM2.5 Audio is an audio model request, while SAM3-LiteText is a segmentation/text-related request." } ] }, @@ -16581,7 +16560,7 @@ "soft_pairs": [ "issue:41084|issue:43037" ], - "prepared_review_unit_hash": "a75cd6297bc7944188b8dc40d670423272fa34b1cb67949d0a294cbdece59104", + "prepared_review_unit_hash": "1e07463b0c45a1ac2a897eb2282b801eb6ea33f06e698fb4a0431d57e876b4fe", "component_budget": { "node_count": 2, "item_count": 2, @@ -16603,18 +16582,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These issues are about different feature requests/problems: one is about block decoding configuration, the other is about adding DeepSeek v3.2 support. They do not appear to describe the same underlying bug or change.", - "confidence": 0.96, - "canonical_issue_reason": "Issue #41084 is the more specific technical request and is unrelated to model-version support; it is a clearer standalone issue for block decoding behavior.", + "summary": "The two issues are unrelated: one is about block decoding behavior, the other requests support for DeepSeek v3.2. No duplicate cluster here.", + "confidence": 0.97, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "Issue #41084 is the best representative only because it is more concrete and narrowly scoped, but it does not duplicate #43037.", + "best_issue_reason": null, "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:41084", "right": "issue:43037", "accept": false, - "reason": "Different requests: block decoding configuration versus DeepSeek v3.2 model support. They do not describe the same code path or change." + "reason": "The titles and implied scopes do not match: block decoding is a low-level decoding behavior issue, while DeepSeek v3.2 support is a model-compatibility request. They do not appear to describe the same underlying bug or change." } ] }, @@ -16644,12 +16623,12 @@ "soft_pairs": [ "issue:43598|issue:43600" ], - "prepared_review_unit_hash": "19e0b0d4e8f88aef9ea619b16844b57c920d4c7220ff55cbe4957c5a209f8c11", + "prepared_review_unit_hash": "6121320bf6993239ff9918d84eadfc7e9162ee12c9e1425158cc5031a10677cd", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1583, + "serialized_chars": 1584, "estimated_input_tokens": 396, "estimated_eval_tokens": 1048 }, @@ -16657,7 +16636,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1583, + "serialized_chars": 1584, "estimated_input_tokens": 396, "estimated_eval_tokens": 1048 }, @@ -16666,18 +16645,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The two issues are thematically related to custom trainers in Trainer, but they describe different problems: one is about `remove_unused_column` customizability, the other about FLOPS estimation. They do not look like duplicates of the same underlying bug.", - "confidence": 0.92, - "canonical_issue_reason": "Issue 43598 is the more concrete and broadly representative Trainer customization request, centered on a specific behavior knob (`remove_unused_column`).", + "summary": "Two closed issues about Trainer extensibility/custom trainers, but they target different behaviors: column removal customizability vs FLOPS estimation. They are related in theme, not duplicates.", + "confidence": 0.89, + "canonical_issue_reason": "Issue #43598 is the broader, more central Trainer customizability request and is the better representative of this cluster.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 43598 is the best representative issue for the cluster because it clearly frames a Trainer behavior that could affect custom trainers, while 43600 is a separate FLOPS-estimation concern.", + "best_issue_reason": "#43598 is the most representative issue because it focuses on the core Trainer customization concern rather than a narrower FLOPS-estimation symptom.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43598", "right": "issue:43600", "accept": false, - "reason": "Both mention custom trainers, but they target different code paths and fixes: one is about `remove_unused_column` behavior, the other about FLOPS counting. They are related topics, not the same bug or a single mergeable change." + "reason": "Both mention custom trainers, but they concern different code paths and outcomes: `remove_unused_column` behavior versus FLOPS estimation. They are related, not the same underlying bug or change." } ] }, @@ -16707,7 +16686,7 @@ "soft_pairs": [ "issue:43684|issue:44405" ], - "prepared_review_unit_hash": "9f707a1f44a3e1d756dd4ef7f0a913164753c8d087dd913a870bb423e01d4045", + "prepared_review_unit_hash": "429e69995ba19e34da2f7c2541faf825dab3ae8a62b0e797d6013b2b216a9427", "component_budget": { "node_count": 2, "item_count": 2, @@ -16729,18 +16708,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The two issues are about different Qwen model families and different auto-class capabilities, so they do not look like duplicates of the same underlying request.", - "confidence": 0.94, - "canonical_issue_reason": "Issue 43684 is the broader registration request, covering AutoModel and AutoModelForConditionalGeneration for Qwen3-Omni.", + "summary": "These look like two separate feature requests, not duplicates: one asks for Qwen3-Omni registration in AutoModel/AutoModelForConditionalGeneration, the other asks for AutoModelForSequenceClassification support for Qwen3.5. Different model families and different auto-class targets.", + "confidence": 0.96, + "canonical_issue_reason": "Issue #43684 is the broader request and names explicit auto-class registration changes, so it is the better representative issue for the cluster even though the two issues are not duplicates.", "canonical_pr_reason": null, - "best_issue_reason": "Issue 43684 is the better representative because it states a clearer, broader auto-class registration gap rather than a narrow class-specific support request.", + "best_issue_reason": "#43684 is the most general/central item in the set; #44405 is a distinct, narrower support request for a different model family and task head.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43684", "right": "issue:44405", "accept": false, - "reason": "Different targets: Qwen3-Omni model registration for AutoModel/AutoModelForConditionalGeneration versus Qwen3.5 support for AutoModelForSequenceClassification. They are not the same bug/change." + "reason": "Reject: they target different functionality. #43684 is about Qwen3-Omni registration for AutoModel/AutoModelForConditionalGeneration, while #44405 is about AutoModelForSequenceClassification support for Qwen3.5 (Qwen3_5Config)." } ] }, @@ -16777,77 +16756,80 @@ "issue:43845|issue:43859", "issue:43846|issue:43859" ], - "prepared_review_unit_hash": "c20ee06d3ad920eed0f2e3c118b5a5a696e304d54bea27941fa5b6a4bdac4a4b", + "prepared_review_unit_hash": "ad48316e08176e03b2e29562203b8dc389155c714c49f10a077890c035e789d1", "component_budget": { "node_count": 4, "item_count": 4, "soft_pair_count": 6, - "serialized_chars": 4263, - "estimated_input_tokens": 1066, - "estimated_eval_tokens": 2388 + "serialized_chars": 4267, + "estimated_input_tokens": 1067, + "estimated_eval_tokens": 2390 }, "budget": { "node_count": 4, "item_count": 4, "soft_pair_count": 6, - "serialized_chars": 4263, - "estimated_input_tokens": 1066, - "estimated_eval_tokens": 2388 + "serialized_chars": 4267, + "estimated_input_tokens": 1067, + "estimated_eval_tokens": 2390 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The cluster looks like low-information duplicate issue noise: one unrelated title (\u201cFett\u201d) and three near-identical \u201chuggingface \u201d placeholders. There isn\u2019t enough evidence to merge any pair as the same underlying bug report.", - "confidence": 0.31, - "canonical_issue_reason": "Issue 43859 is the best representative of the repeated placeholder \u201chuggingface \u201d reports because it is one of the matching titles and has the only discussion activity, but all candidates are too underspecified to be a strong canonical.", + "summary": "43835 is unrelated, 43845 and 43846 look like the strongest duplicate pair, and 43859 is only a weaker title match because its body is noticeably shorter. I would anchor the cluster on 43845, but only accept the 43845\u219443846 soft edge.", + "confidence": 0.66, + "canonical_issue_reason": "issue:43845 is the earliest strong match in the cluster and pairs cleanly with issue:43846, while issue:43859 is less convincing on body evidence.", "canonical_pr_reason": null, - "best_issue_reason": "43859 is the most suitable issue for dedupe tracking among the placeholders since it matches the repeated title and has the most activity, but it is still not a high-quality canonical issue.", + "best_issue_reason": "issue:43845 is the best issue anchor because it is earliest among the near-duplicate huggingface reports and has the strongest duplicate signal with issue:43846.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:43835", "right": "issue:43845", "accept": false, - "reason": "Different titles and no description-level evidence of the same bug; only weak text similarity." + "reason": "'Fett' appears unrelated to the huggingface reports and does not share enough evidence of the same underlying issue." }, { "left": "issue:43835", "right": "issue:43846", "accept": false, - "reason": "Different titles and no concrete shared issue details, so this looks like unrelated placeholder noise." + "reason": "'Fett' is textually unrelated to the huggingface issues and does not look like the same report." }, { "left": "issue:43835", "right": "issue:43859", "accept": false, - "reason": "No shared bug description or explicit linkage; similarity is too shallow to treat as a duplicate." + "reason": "'Fett' does not appear to describe the same underlying issue as the huggingface reports." }, { "left": "issue:43845", "right": "issue:43846", - "accept": false, - "reason": "They share the same placeholder title, but there is no evidence they report the same underlying problem." + "accept": true, + "reason": "Same title and matching body length make these look like the same report." }, { "left": "issue:43845", "right": "issue:43859", "accept": false, - "reason": "Same generic title, but no specific symptom, target, or code-path evidence to justify a duplicate merge." + "reason": "The title matches, but the body is much shorter here, so the evidence is not strong enough to treat this as the same issue." }, { "left": "issue:43846", "right": "issue:43859", "accept": false, - "reason": "The reports are only superficially similar; nothing indicates the same underlying bug or change request." + "reason": "They share a title, but the shorter body on 43859 makes this too weak to confidently merge as a duplicate." } ] }, - "evaluator_result": null, - "evaluator_used": false, - "retried": false, - "accepted_nontrivial_soft_edge": false, + "evaluator_result": { + "accept": false, + "feedback": "The cluster judgment is mostly plausible, but the writeup overstates the evidence. In the packet, 43845 and 43846 share the same title and body length, but 43859 has a much shorter body length (46 vs 84), so the repeated claims about \u201csame body characteristics\u201d / \u201csimilar body length\u201d are not grounded. The canonical-earliest choice for 43845 is fine, but the soft-edge rationales for 43845\u219443859 and 43846\u219443859 should be phrased more conservatively. The 43835 \u2018Fett\u2019 rejection is reasonable, but overall this is not conservative enough to accept as written." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null }, @@ -16881,88 +16863,88 @@ "issue:44572|issue:44937", "issue:44573|issue:44937" ], - "prepared_review_unit_hash": "405844ac650d22d818a97976a121277d1b04ad35d113c683331bf4c5e02030b4", + "prepared_review_unit_hash": "f17b9dcb542b461d01da14fe5df908e93606c85f96dd0bca0ba4ba155d8588e8", "component_budget": { "node_count": 5, "item_count": 5, "soft_pair_count": 9, - "serialized_chars": 5791, - "estimated_input_tokens": 1448, - "estimated_eval_tokens": 3152 + "serialized_chars": 5788, + "estimated_input_tokens": 1447, + "estimated_eval_tokens": 3150 }, "budget": { "node_count": 5, "item_count": 5, "soft_pair_count": 9, - "serialized_chars": 5791, - "estimated_input_tokens": 1448, - "estimated_eval_tokens": 3152 + "serialized_chars": 5788, + "estimated_input_tokens": 1447, + "estimated_eval_tokens": 3150 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "This cluster appears to be a set of unrelated low-information/spam issues, with no concrete shared bug or feature description to justify deduping beyond generic similarity.", - "confidence": 0.92, - "canonical_issue_reason": "issue:44371 is the earliest generic spam-placeholder issue in the set, so it is the best representative of this noise cluster.", + "summary": "All items are low-signal spam/generic placeholders with no substantive problem statement, and they do not appear to describe the same underlying issue. I would not merge any of the soft pairs.", + "confidence": 0.96, + "canonical_issue_reason": null, "canonical_pr_reason": null, - "best_issue_reason": "issue:44371 is the most suitable cluster representative because it is a minimal, content-free spam issue and predates the other similar placeholders.", + "best_issue_reason": null, "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:44371", "right": "issue:44572", "accept": false, - "reason": "Both are spam-placeholder titles, but there is no evidence they refer to the same underlying report or change." + "reason": "Same placeholder title, but there is no substantive description to confirm a shared underlying bug/change; treat as separate spam artifacts." }, { "left": "issue:44371", "right": "issue:44573", "accept": false, - "reason": "Identical-looking spam titles are not enough to confirm the same issue; no concrete shared problem is described." + "reason": "Both are generic spam placeholders, but there is no concrete evidence they refer to the same issue." }, { "left": "issue:44371", "right": "issue:44418", "accept": false, - "reason": "One is a spam placeholder and the other is a documentation suggestion; these do not look like the same underlying issue." + "reason": "One is spam and the other is a vague documentation suggestion; too little concrete overlap to call them duplicates." }, { "left": "issue:44371", "right": "issue:44937", "accept": false, - "reason": "A spam placeholder and a specific phrase-based issue are not the same concrete problem." + "reason": "Different titles and no meaningful body detail; not enough evidence of the same underlying issue." }, { "left": "issue:44418", "right": "issue:44572", "accept": false, - "reason": "Documentation enhancement versus spam placeholder; no sign of a shared bug or request." + "reason": "A generic docs suggestion and a spam placeholder are not the same concrete issue." }, { "left": "issue:44418", "right": "issue:44573", "accept": false, - "reason": "Documentation enhancement versus spam placeholder; the similarity is only superficial." + "reason": "No shared bug/change is described; these look like unrelated low-information issues." }, { "left": "issue:44572", "right": "issue:44573", "accept": false, - "reason": "Both are spam placeholders, but there is no substantive evidence they describe one underlying issue rather than separate spam posts." + "reason": "Matching spam-style titles alone is not enough; there is no evidence they describe one concrete issue." }, { "left": "issue:44572", "right": "issue:44937", "accept": false, - "reason": "A spam placeholder and a specific quoted-name issue are unrelated on the available evidence." + "reason": "Generic spam placeholder versus unrelated named-title issue; no clear duplicate relationship." }, { "left": "issue:44573", "right": "issue:44937", "accept": false, - "reason": "A spam placeholder and a specific quoted-name issue are unrelated on the available evidence." + "reason": "Different titles and no substantive content to support merging as duplicates." } ] }, @@ -16992,7 +16974,7 @@ "soft_pairs": [ "issue:45145|issue:45151" ], - "prepared_review_unit_hash": "6da72f3c52adac3498e870da75a3625d86c066be055806df50057c4514d10ed6", + "prepared_review_unit_hash": "22c27b177c03f29cdfffe2026c6bb2e657f7fcd695f13379907153b50ab3b495", "component_budget": { "node_count": 2, "item_count": 2, @@ -17014,24 +16996,24 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "#45145 is still the best canonical issue because it is earlier and already has an inbound reference. However, the pair should not be auto-accepted as a duplicate from the evidence shown alone: the packet only exposes identical titles, similar body lengths, and a soft similarity score, without body text or explicit linkage.", - "confidence": 0.84, - "canonical_issue_reason": "#45145 is earlier and already has an inbound reference, so it is the stronger canonical copy for this report.", + "summary": "#45145 is the best canonical issue anchor because it was opened first and already has an inbound reference. The pair looks related, but the evidence here is only identical titles plus a soft similarity signal, so I would treat the duplicate relationship as tentative rather than confirmed.", + "confidence": 0.79, + "canonical_issue_reason": "#45145 was created earlier and has one inbound reference, so it is the strongest anchor for the cluster.", "canonical_pr_reason": null, - "best_issue_reason": "#45145 is the best issue to keep as canonical because it predates #45151 and has outside attention.", + "best_issue_reason": "#45145 is the best representative issue because it is the earliest instance and already has slightly more observed attention.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45145", "right": "issue:45151", "accept": false, - "reason": "The titles match and the similarity signal is strong, but the packet does not expose the bodies or any explicit linkage. That is enough to suspect a duplicate, but not enough to confidently accept it as the same underlying issue." + "reason": "The pair shares an identical title and a moderate soft-similarity signal, but the packet does not provide enough direct evidence to conclude they are definitely the same underlying report." } ] }, "evaluator_result": { "accept": false, - "feedback": "The canonical-issue choice is grounded: #45145 is earlier and has an inbound reference. However, the duplicate verdict is a bit too strong for the evidence provided. The packet only shows identical titles, similar body lengths, and a soft-similarity signal with Jaccard 0.63; it does not expose the bodies or any explicit linkage. This supports a likely duplicate, but not the confident \u2018same underlying issue\u2019 framing." + "feedback": "The canonical-choice rationale is grounded (earlier creation time and one inbound reference), but the duplicate verdict is a bit overstated. The packet only supports same title plus a soft-similarity signal (Jaccard 0.63); it does not directly show the bodies are \"very similar\" or that the issues are definitively the same underlying report. Please keep the soft-edge language more conservative." }, "evaluator_used": true, "retried": true, @@ -17112,40 +17094,40 @@ "soft_pairs": [ "issue:45672|issue:45674" ], - "prepared_review_unit_hash": "4f4a08854d233215f84ec137240966c5dc7c43b59bf479e09fba70d23558c4ad", + "prepared_review_unit_hash": "fb1b74ec895b58e48e3988f7988f784318fadee4c5f485eb4329ec1ed3c01c5f", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1668, - "estimated_input_tokens": 417, - "estimated_eval_tokens": 1090 + "serialized_chars": 1669, + "estimated_input_tokens": 418, + "estimated_eval_tokens": 1092 }, "budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1668, - "estimated_input_tokens": 417, - "estimated_eval_tokens": 1090 + "serialized_chars": 1669, + "estimated_input_tokens": 418, + "estimated_eval_tokens": 1092 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are two distinct 4-bit inference bugs, not duplicates: one is a Gemma4/audio-module torch.finfo() TypeError on uint8 weights, the other is a BitsAndBytesConfig regression where llm_int8_skip_modules drops the default lm_head exclusion and triggers an AssertionError.", - "confidence": 0.94, - "canonical_issue_reason": "Issue #45674 is the cleaner cluster anchor: it describes a specific, configuration-driven regression with a clear failing path, while #45672 is a separate model/module-specific bug.", + "summary": "These are two distinct open issues in the 4-bit inference/BitsAndBytes area, but they describe different failures: one is a Gemma4 audio-module `torch.finfo()` TypeError on `uint8` weights, the other is a `BitsAndBytesConfig.llm_int8_skip_modules` regression that drops the default `lm_head` exclusion and triggers an AssertionError. They should not be merged as duplicates.", + "confidence": 0.95, + "canonical_issue_reason": "Issue #45674 is the better canonical representative because it states a more concrete configuration-level regression and has a cleaner, actionable failure mode.", "canonical_pr_reason": null, - "best_issue_reason": "#45674 is the best representative issue for the cluster because it is tightly scoped and easier to map to a concrete fix, even though it is not a duplicate of #45672.", + "best_issue_reason": "Issue #45674 appears to be the best issue in the cluster: it has a precise title, a specific root-cause hint, and is likely easier to triage and fix than the narrower Gemma4 audio-module report.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45672", "right": "issue:45674", "accept": false, - "reason": "Different failure modes and code paths: Gemma4 audio module torch.finfo() on uint8 weights versus BitsAndBytesConfig skipping the default lm_head exclusion and causing an AssertionError. They share only broad 4-bit inference context." + "reason": "Same broad 4-bit inference area, but different code paths and symptoms: Gemma4 audio modules hit `torch.finfo()` on `uint8`, while the other is a BitsAndBytesConfig `lm_head` exclusion regression causing an AssertionError. Not the same underlying bug." } ] }, @@ -17175,40 +17157,40 @@ "soft_pairs": [ "issue:45762|issue:45763" ], - "prepared_review_unit_hash": "319ecca25a3df64330644d4387fc07c4062ca8e85f66c228423af59c8d5b8081", + "prepared_review_unit_hash": "2c726ebeeb7bb79fdf99689aabddaf190cc53b9ebd6679ebeeabfb6e8a1e6277", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1476, - "estimated_input_tokens": 369, - "estimated_eval_tokens": 994 + "serialized_chars": 1477, + "estimated_input_tokens": 370, + "estimated_eval_tokens": 996 }, "budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 1476, - "estimated_input_tokens": 369, - "estimated_eval_tokens": 994 + "serialized_chars": 1477, + "estimated_input_tokens": 370, + "estimated_eval_tokens": 996 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both items look like unrelated low-signal/spam issues containing personal names and dates, not a real duplicate bug cluster.", + "summary": "Two sparse closed issues with different personal-name titles and dates; there is no evidence they describe the same underlying bug or request.", "confidence": 0.96, - "canonical_issue_reason": "Issue #45762 is the earlier of the two and can serve as the cluster anchor, but both issues are equally non-actionable personal-data spam.", + "canonical_issue_reason": "No canonical issue: the issues have different titles, no body, no shared targets, and no shared bug context.", "canonical_pr_reason": null, - "best_issue_reason": "No issue in this cluster is a good canonical bug report; both are closed spam-like entries with no technical content.", + "best_issue_reason": "Neither issue is a good canonical choice; if forced, 45762 is only slightly earlier, but the pair does not look like duplicates.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45762", "right": "issue:45763", "accept": false, - "reason": "They only share a vague spam-like pattern; the titles name different people and dates, with no indication of the same underlying issue." + "reason": "Reject: the titles refer to different people and dates, with no supporting discussion or linkage indicating the same underlying issue." } ] }, @@ -17238,7 +17220,7 @@ "soft_pairs": [ "issue:45764|issue:45765" ], - "prepared_review_unit_hash": "131e9b53aed97a99ee290a1f0d70bf55a9bee708ed0e337bb32044ada1b46850", + "prepared_review_unit_hash": "5d0004029bf5f12fd9ff7021273cce9580bb120c45c3e9a3fc928795a69e395f", "component_budget": { "node_count": 2, "item_count": 2, @@ -17260,18 +17242,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The two closed issues are just personal-name/date titles with no bodies or linked context, and they do not describe the same underlying problem. The soft similarity signal is not enough to treat them as duplicates.", + "summary": "These are two unrelated, content-free issues with only personal-name/date titles; there is no evidence they describe the same underlying bug or request.", "confidence": 0.97, - "canonical_issue_reason": "No canonical issue: neither issue contains actionable problem description, and the titles indicate unrelated placeholder-like entries rather than a shared bug report.", + "canonical_issue_reason": "If a representative must be chosen, issue 45764 is the earliest, but neither issue has substantive content to justify a true canonical selection.", "canonical_pr_reason": null, - "best_issue_reason": "Neither issue is a good representative target for deduping because both lack substantive content and do not appear related.", + "best_issue_reason": "Issue 45764 is only marginally better as a cluster representative because it appears first; otherwise it is no more informative than 45765.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45764", "right": "issue:45765", "accept": false, - "reason": "Only superficial title similarity; the names/dates differ and there is no body, target, or code-path evidence showing the same underlying issue." + "reason": "The titles reference different names and dates and contain no shared problem statement, so they do not look like duplicates." } ] }, @@ -17311,91 +17293,91 @@ "issue:45871|issue:45873", "issue:45872|issue:45873" ], - "prepared_review_unit_hash": "31728d4160dbe4fb33cd95a7aa20d6e68b5f875b917754f44ecb005ac9f8e553", + "prepared_review_unit_hash": "4afbd60b1863eeaaf60e38b49f402f472c08832d4b036ef02601a26fb8e2762f", "component_budget": { "node_count": 5, "item_count": 5, "soft_pair_count": 8, - "serialized_chars": 5721, - "estimated_input_tokens": 1431, - "estimated_eval_tokens": 3118 + "serialized_chars": 5726, + "estimated_input_tokens": 1432, + "estimated_eval_tokens": 3120 }, "budget": { "node_count": 5, "item_count": 5, "soft_pair_count": 8, - "serialized_chars": 5721, - "estimated_input_tokens": 1431, - "estimated_eval_tokens": 3118 + "serialized_chars": 5726, + "estimated_input_tokens": 1432, + "estimated_eval_tokens": 3120 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The packet suggests two highly similar issue pairs: 45870/45871 and 45872/45873. The evidence is metadata-only, so I\u2019d treat them as likely duplicates rather than proven exact duplicates. 45869 looks topic-adjacent but distinct.", - "confidence": 0.78, - "canonical_issue_reason": "45870 is the best canonical issue here because it is the earlier and slightly more active member of the clearest duplicate-like pair (45870/45871). 45872/45873 looks like a separate similar pair, and 45869 is adjacent but not the same issue.", + "summary": "This cluster actually contains two separate duplicate pairs: 45870/45871 are exact title/body duplicates, and 45872/45873 are exact title/body duplicates. Issue 45869 is a different financial dataset item and should not be merged with either pair.", + "confidence": 0.98, + "canonical_issue_reason": "No single canonical issue cleanly represents the whole cluster because it mixes two distinct duplicate pairs plus one unrelated issue; within the \"Example: Financial time series prediction with DataForge\" pair, 45870 is the better canonical representative because it is slightly earlier and has more discussion activity.", "canonical_pr_reason": null, - "best_issue_reason": "45870 is the strongest overall representative among these issues: it anchors the most clearly repeated title/body pattern, is earlier than 45871, and has a bit more discussion activity. The other pair (45872/45873) appears separate, and 45869 is too different to serve as the main representative.", + "best_issue_reason": "45870 is the best overall issue representative only among the exact duplicate pair it belongs to, but the cluster is not a single coherent duplicate set.", "best_pr_reason": null, "soft_edge_verdicts": [ { "left": "issue:45869", "right": "issue:45870", "accept": false, - "reason": "Both mention finance/data themes, but the titles describe different concrete things: a dataset bundle vs an example about financial time series prediction." + "reason": "Different subjects: a financial data bundle listing vs an example/tutorial issue." }, { "left": "issue:45869", "right": "issue:45871", "accept": false, - "reason": "Both mention finance/data themes, but the titles describe different concrete things: a dataset bundle vs an example about financial time series prediction." + "reason": "Different subjects: a financial data bundle listing vs an example/tutorial issue." }, { "left": "issue:45870", "right": "issue:45871", "accept": true, - "reason": "Identical titles and matching body length with no differentiating metadata make these likely the same underlying issue." + "reason": "Same title and matching content pattern; these look like exact duplicate issues." }, { "left": "issue:45870", "right": "issue:45872", "accept": false, - "reason": "They share the broad DataForge/finance theme, but the titles point to different requests: prediction example vs scraping service." + "reason": "Related finance/DataForge theme, but different requested change: example/tutorial vs scraping-service integration." }, { "left": "issue:45870", "right": "issue:45873", "accept": false, - "reason": "They share the broad DataForge/finance theme, but the titles point to different requests: prediction example vs scraping service." + "reason": "Related finance/DataForge theme, but different requested change: example/tutorial vs scraping-service integration." }, { "left": "issue:45871", "right": "issue:45872", "accept": false, - "reason": "They share the broad DataForge/finance theme, but the titles point to different requests: prediction example vs scraping service." + "reason": "Related finance/DataForge theme, but different requested change: example/tutorial vs scraping-service integration." }, { "left": "issue:45871", "right": "issue:45873", "accept": false, - "reason": "They share the broad DataForge/finance theme, but the titles point to different requests: prediction example vs scraping service." + "reason": "Related finance/DataForge theme, but different requested change: example/tutorial vs scraping-service integration." }, { "left": "issue:45872", "right": "issue:45873", "accept": true, - "reason": "Identical titles and matching body length with no differentiating metadata make these likely the same underlying issue." + "reason": "Same title and matching content pattern; these look like exact duplicate issues." } ] }, "evaluator_result": { - "accept": false, - "feedback": "The output is a bit too strong for the evidence in the packet. The titles and body lengths support that 45870/45871 and 45872/45873 are highly similar, but the packet does not provide body text or explicit duplicate markers, so calling them \"exact duplicates\" is overstated. Similarly, 45869 is only topic-adjacent from the metadata available; that part is fine, but the overall summary and canonical reasoning should stay more conservative. 45870 as the earlier, slightly more active representative is reasonable, but the confidence and duplicate wording should be softened." + "accept": true, + "feedback": "The summary is conservative and grounded in the packet: 45870/45871 and 45872/45873 have identical titles and max similarity, while 45869 is clearly a different dataset issue. The soft-edge verdicts do not overstate evidence, and the canonical choice for 45870 over 45871 is supported by earlier creation time and higher discussion activity." }, "evaluator_used": true, - "retried": true, + "retried": false, "accepted_nontrivial_soft_edge": true, "error_kind": null, "error_message": null @@ -17452,172 +17434,172 @@ "pull_request:43823|pull_request:45213", "pull_request:43251|pull_request:43254" ], - "prepared_review_unit_hash": "753f0da67ef2d400714e8a0703123e49b5dad79b7b4fa11e0d3dfed0065e3166", + "prepared_review_unit_hash": "34711673bc32d80f93c4a3866901d04d2a4b9abe06a94df4ff932e8fa8deba11", "component_budget": { "node_count": 14, "item_count": 14, "soft_pair_count": 22, - "serialized_chars": 38364, - "estimated_input_tokens": 9591, - "estimated_eval_tokens": 19438 + "serialized_chars": 38369, + "estimated_input_tokens": 9593, + "estimated_eval_tokens": 19442 }, "budget": { "node_count": 14, "item_count": 14, "soft_pair_count": 22, - "serialized_chars": 38364, - "estimated_input_tokens": 9591, - "estimated_eval_tokens": 19438 + "serialized_chars": 38369, + "estimated_input_tokens": 9593, + "estimated_eval_tokens": 19442 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "The cluster is mostly unrelated PRs; the only true duplicate pair is the two cross-entropy kwargs fixes (#43251 and #43254).", - "confidence": 0.95, + "summary": "This cluster is mostly heterogeneous: model additions, tokenizer/API refactors, doc/test-only updates, and unrelated loading/conversion fixes. The only true near-duplicate pair is the fixed_cross_entropy kwargs change, where one PR is effectively a narrower variant of the other. ", + "confidence": 0.96, "canonical_issue_reason": null, - "canonical_pr_reason": "#43251 is the better canonical PR: it has the explicit issue link, clearer intent, and a slightly more complete implementation of the same fixed_cross_entropy change.", + "canonical_pr_reason": "PR 43251 is the better canonical representative: it has the explicit issue target (43240) and implements the concrete cross-entropy kwargs fix in one place. PR 43254 is a close variant of the same change.", "best_issue_reason": null, - "best_pr_reason": "#43251 is the best representative PR for the duplicate pair because it directly targets the tracked bug and is the most self-contained version of that fix.", + "best_pr_reason": "PR 43251 is the strongest cluster anchor because it directly addresses the underlying bug, has the explicit issue link, and is the only pair here that cleanly collapses into one duplicate fix.", "soft_edge_verdicts": [ { "left": "pull_request:39895", "right": "pull_request:43424", "accept": false, - "reason": "Unrelated model-doc/new-model PR versus an executorch dynamic-shape test change; no shared bug or code path." + "reason": "Both are PRs, but one adds VideoPrism and the other adds a dynamic-shape export test; unrelated code paths and no plausible single merged fix." }, { "left": "pull_request:41855", "right": "pull_request:43823", "accept": false, - "reason": "Mistral tokenizer API additions and a new MobileLLM model are different changes with no common defect." + "reason": "Tokenizer missing methods for MistralCommon is unrelated to adding MobileLLM; different models and different API surfaces." }, { "left": "pull_request:41855", "right": "pull_request:43424", "accept": false, - "reason": "Tokenizer method additions are unrelated to executorch export dynamic-shape testing." + "reason": "Mistral tokenizer method additions and ExecuTorch dynamic-shape tests are unrelated changes." }, { "left": "pull_request:42668", "right": "pull_request:43823", "accept": false, - "reason": "Processor/pretrained-loading robustness work is unrelated to adding MobileLLM." + "reason": "Processor auto-loading robustness and adding a new model are not the same underlying bug/change." }, { "left": "pull_request:43251", "right": "pull_request:46067", "accept": false, - "reason": "Both touch infrastructure, but they fix different problems: cross-entropy kwargs versus base_model_prefix handling in conversions." + "reason": "Cross-entropy kwargs support is unrelated to conversion/base_model_prefix loading fixes." }, { "left": "pull_request:43251", "right": "pull_request:43823", "accept": false, - "reason": "Loss-function kwargs fix and new MobileLLM model addition are unrelated." + "reason": "A loss-function fix and a new model addition do not address the same code-path problem." }, { "left": "pull_request:43251", "right": "pull_request:44827", "accept": false, - "reason": "Cross-entropy kwargs handling is unrelated to Mistral4 test and MoE fixes." + "reason": "Cross-entropy kwargs support is unrelated to Mistral4 test fixes and MOE contiguity changes." }, { "left": "pull_request:43424", "right": "pull_request:46067", "accept": false, - "reason": "Executorch dynamic-shape tests and conversion base_model_prefix handling address different code paths." + "reason": "ExecuTorch export tests and loading/conversion prefix handling are unrelated." }, { "left": "pull_request:43424", "right": "pull_request:44827", "accept": false, - "reason": "Different subsystems and different bugs; no plausible single merged PR." + "reason": "Dynamic-shape export tests and Mistral4 test fixes are different issues." }, { "left": "pull_request:43424", "right": "pull_request:43823", "accept": false, - "reason": "Executorch testing change is unrelated to introducing MobileLLM." + "reason": "A test for exportability does not match adding a new model implementation." }, { "left": "pull_request:43823", "right": "pull_request:45655", "accept": false, - "reason": "New model addition and config resolution ordering bug are unrelated." + "reason": "New MobileLLM model support and config-resolution ordering in modeling_utils are unrelated." }, { "left": "pull_request:43823", "right": "pull_request:43995", "accept": false, - "reason": "Adding MobileLLM is unrelated to Falcon output-interface refactoring." + "reason": "MobileLLM addition and Falcon output-interface refactoring are different model-specific changes." }, { "left": "pull_request:43823", "right": "pull_request:46067", "accept": false, - "reason": "New model addition and conversion-prefix fixes are different changes." + "reason": "New model support and conversion prefix handling are not the same concrete fix." }, { "left": "pull_request:43995", "right": "pull_request:44827", "accept": false, - "reason": "Falcon output-interface refactor and Mistral4 test fixes are not the same underlying change." + "reason": "Falcon output collection refactor and Mistral4 test fixes do not target the same bug or change." }, { "left": "pull_request:43995", "right": "pull_request:45213", "accept": false, - "reason": "Falcon refactoring and the model-creation skill PR are unrelated." + "reason": "Falcon refactoring and the model-creation skill are unrelated." }, { "left": "pull_request:44827", "right": "pull_request:46067", "accept": false, - "reason": "Mistral4 tests/MoE fix is unrelated to conversion prefix handling." + "reason": "Mistral4 test fixes and loading-prefix conversion fixes are unrelated." }, { "left": "pull_request:44827", "right": "pull_request:45170", "accept": false, - "reason": "Mistral4 test fixes and layrnorm-to-layernorm renaming address different models and bugs." + "reason": "Mistral4 tests and layrnorm->layernorm renaming are different problems." }, { "left": "pull_request:44827", "right": "pull_request:46048", "accept": false, - "reason": "Mistral4 test fixes are unrelated to the TDT loss kernel addition." + "reason": "Mistral4 tests and TDT loss kernel integration are unrelated." }, { "left": "pull_request:44827", "right": "pull_request:45655", "accept": false, - "reason": "Mistral4 test fixes and config resolution ordering are unrelated." + "reason": "A model-specific test fix and a class-annotation resolution fix are not the same code-path problem." }, { "left": "pull_request:44827", "right": "pull_request:45213", "accept": false, - "reason": "A model test-fix PR and the model-creation skill PR are not the same underlying change." + "reason": "Mistral4 test fixes and the model creation skill are not mergeable as one duplicate change." }, { "left": "pull_request:43823", "right": "pull_request:45213", "accept": false, - "reason": "A model implementation PR and a repository skill/documentation PR are unrelated." + "reason": "Adding MobileLLM and adding an internal model-creation skill are unrelated." }, { "left": "pull_request:43251", "right": "pull_request:43254", "accept": true, - "reason": "Both PRs fix the same fixed_cross_entropy issue by passing supported nn.functional.cross_entropy kwargs in src/transformers/loss/loss_utils.py; they are mergeable into one change." + "reason": "These are two near-identical fixes for the same fixed_cross_entropy kwargs issue in the same file and same issue target; they could plausibly be merged into one PR." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is grounded in the packet and stays conservative: it identifies the only clearly duplicate pair as #43251/#43254, which is supported by the shared file and shared issue target. The soft-edge rejections are also reasonable and do not overstate evidence." + "feedback": "The cluster summary is grounded in the packet and stays conservative. The only accepted soft-edge pair, PR 43251 vs 43254, is clearly the same fixed_cross_entropy kwargs issue in the same file with the same issue target, so mergeability is plausible. The other rejections are appropriately cautious and supported by the diffs/titles." }, "evaluator_used": true, "retried": false, @@ -17644,40 +17626,40 @@ "soft_pairs": [ "pull_request:40546|pull_request:42781" ], - "prepared_review_unit_hash": "5f6f020c618e8b60d60b92cc6702a4ce1273d864df8717303d3ac73ed121f4ee", + "prepared_review_unit_hash": "fddbcc199c0d370c1e3a472884f26270f09a2fcc808b96c123c6e8c7418b33d9", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 7225, - "estimated_input_tokens": 1807, - "estimated_eval_tokens": 3870 + "serialized_chars": 7223, + "estimated_input_tokens": 1806, + "estimated_eval_tokens": 3868 }, "budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 7225, - "estimated_input_tokens": 1807, - "estimated_eval_tokens": 3870 + "serialized_chars": 7223, + "estimated_input_tokens": 1806, + "estimated_eval_tokens": 3868 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both PRs are about VibeVoice, but they target different scopes: #40546 adds the base VibeVoice integration, while #42781 adds a separate realtime/acoustic-tokenizer extension. They overlap on model-family plumbing, but they are not the same underlying change.", - "confidence": 0.89, + "summary": "Both PRs are related VibeVoice work, but they are not the same change: #40546 adds the core VibeVoice model/docs, while #42781 extends the stack with realtime support and a VibeVoice acoustic tokenizer. They share some scaffolding files, but the concrete scope differs enough that they should not be merged as duplicates.", + "confidence": 0.91, "canonical_issue_reason": null, - "canonical_pr_reason": "#40546 is the original reviewed PR for the core VibeVoice implementation; #42781 is a later, separate realtime-oriented extension that adds additional components rather than duplicating the base feature.", + "canonical_pr_reason": "#40546 is the cleaner baseline implementation of VibeVoice and is less specialized than #42781, which adds a separate realtime/tokenizer extension on top of that base work.", "best_issue_reason": null, - "best_pr_reason": "#40546 is the most representative and mature PR for the core VibeVoice feature set. #42781 is draft and appears to build on the base implementation instead of being a duplicate of it.", + "best_pr_reason": "#40546 is the best representative PR for the main VibeVoice integration because it focuses on the core model import path and documentation rather than the additional realtime-specific pieces.", "soft_edge_verdicts": [ { "left": "pull_request:40546", "right": "pull_request:42781", "accept": false, - "reason": "Reject: the shared files reflect common VibeVoice scaffolding, but #40546 implements the base model while #42781 adds realtime and acoustic-tokenizer work. Different concrete code paths and feature scope." + "reason": "Shared VibeVoice scaffolding is not enough here: #40546 implements the initial VibeVoice model, while #42781 adds realtime behavior plus an acoustic tokenizer and related generation changes. They are related, but not the same underlying bug/change and would not plausibly be one merged PR." } ] }, @@ -17707,12 +17689,12 @@ "soft_pairs": [ "pull_request:40755|pull_request:41224" ], - "prepared_review_unit_hash": "dda6230ad1835ebb5c207c6a41492556143ca14d35e09294b9aebe092d96fed5", + "prepared_review_unit_hash": "7c7c580b7abf5d0ac6dbf6cbe4b4ee641004624064b217a1095a17cacd224ec5", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4675, + "serialized_chars": 4676, "estimated_input_tokens": 1169, "estimated_eval_tokens": 2594 }, @@ -17720,7 +17702,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4675, + "serialized_chars": 4676, "estimated_input_tokens": 1169, "estimated_eval_tokens": 2594 }, @@ -17729,18 +17711,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These two pull requests are not duplicates: one adds TimesFM covariate forecasting support, the other adds DINOv3 image classification support. They touch different model families, different code paths, and different user-facing features.", - "confidence": 0.98, + "summary": "These are two unrelated feature PRs: one adds TimesFM covariate forecasting, the other adds DINOv3 image-classification support. They share only broad similarity as model feature work, not the same underlying change.", + "confidence": 0.96, "canonical_issue_reason": null, - "canonical_pr_reason": "No canonical PR for the cluster: the PRs implement unrelated features in different subsystems, so they should not be deduplicated against each other.", + "canonical_pr_reason": "PR 41224 is the cleaner canonical representative: it is a focused, concrete feature addition with explicit model, auto-class, docs, and tests changes.", "best_issue_reason": null, - "best_pr_reason": "PR #41224 is the better standalone representative because it is a narrower, concrete feature addition with a clearly scoped implementation and docs/tests, whereas PR #40755 is a larger cross-file feature expansion but still unrelated to the other PR.", + "best_pr_reason": "PR 41224 is the strongest standalone PR in the cluster because it targets one specific code path (DINOv3 image classification support) and appears well-scoped for mergeability.", "soft_edge_verdicts": [ { "left": "pull_request:40755", "right": "pull_request:41224", "accept": false, - "reason": "Different models and different functionality: TimesFM covariate forecasting vs DINOv3 image classification. They do not appear to fix the same underlying bug or add the same change." + "reason": "Reject: these PRs implement different features in different model families. They do not appear to fix the same bug or deliver the same change, and would not plausibly be merged into one PR." } ] }, @@ -17835,7 +17817,7 @@ "pull_request:41797|pull_request:45186", "pull_request:45186|pull_request:45630" ], - "prepared_review_unit_hash": "1a77cfd8dbcbaf2c68561baddc3ea8071dba4c9767d528f30b274d27e12be1fe", + "prepared_review_unit_hash": "6890fd3ec6a5a4ecde2b95032cdc9bbf0d5c780675cae266ca0a974ed9062bbf", "component_budget": { "node_count": 3, "item_count": 3, @@ -17857,24 +17839,24 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "All three PRs are model-introduction docs/code additions for different models, not the same underlying change. They share generic Transformers scaffolding, but each targets a distinct model family and codepath, so the soft links should be rejected.", - "confidence": 0.96, + "summary": "These are three separate model-addition pull requests (DeepSeek OCR, Isaac, Kimi2-6). They share repository scaffolding for adding new models, but they target different architectures and code paths, so they should not be deduplicated.", + "confidence": 0.97, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #45630 is the strongest representative of the cluster because it is the newest, has the highest review activity, and appears to be the most fully developed model-addition change; however, it is still distinct from the other PRs.", + "canonical_pr_reason": "PR 45630 is the broadest and most complete of the three, with the largest integration surface and highest review activity, so it is the best representative PR for this cluster if one must be chosen.", "best_issue_reason": null, - "best_pr_reason": "PR #45630 is a reasonable cluster representative for a new-model addition, with broad integration across auto-mappings, model files, and tests, and substantially more discussion than the others.", + "best_pr_reason": "PR 45630 is the strongest standalone example of a full model integration PR: it adds docs, auto mappings, model code, processors, and tests for a distinct model family.", "soft_edge_verdicts": [ { "left": "pull_request:41797", "right": "pull_request:45186", "accept": false, - "reason": "Different model additions: DeepSeek OCR vs Isaac. Shared Transformers scaffolding and docs/auto-registry updates are generic and do not indicate the same bug or mergeable code-path fix." + "reason": "Different models and different implementations; overlap is only generic new-model scaffolding in docs/auto-registration, not the same bug or change." }, { "left": "pull_request:45186", "right": "pull_request:45630", "accept": false, - "reason": "Different model additions: Isaac vs Kimi2-6. They touch similar registry surfaces, but the underlying models, files, and implementation details are unrelated." + "reason": "Separate model families with distinct files, configs, and tests; they do not fix the same code-path problem and would not plausibly be merged into one PR." } ] }, @@ -17904,12 +17886,12 @@ "soft_pairs": [ "pull_request:42975|pull_request:42976" ], - "prepared_review_unit_hash": "af911583abe8598bb3fe2b3c84815b38aeee806e092c9d2a8aa37e256f31afcb", + "prepared_review_unit_hash": "0cfbf19903bf1e47cfe00559a5c0f1fa58a0197681698021756d675fa9f5d8bc", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 5934, + "serialized_chars": 5935, "estimated_input_tokens": 1484, "estimated_eval_tokens": 3224 }, @@ -17917,7 +17899,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 5934, + "serialized_chars": 5935, "estimated_input_tokens": 1484, "estimated_eval_tokens": 3224 }, @@ -17926,18 +17908,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are both broad workflow-maintenance PRs, but they target different GitHub Actions and different upgrade goals: Node 24 compatibility vs. docker/build-push-action version bumps. They should not be deduplicated.", - "confidence": 0.96, + "summary": "Both PRs are GitHub Actions maintenance updates, but they target different underlying changes: PR 42975 updates core JS-based workflow actions for Node 24 compatibility, while PR 42976 mainly bumps docker/build-push-action to v6. They overlap in a few workflow files, but they are not the same fix and should not be merged as duplicates.", + "confidence": 0.94, "canonical_issue_reason": null, - "canonical_pr_reason": "No single canonical PR for deduplication: 42975 is about upgrading common GitHub Actions for Node 24 compatibility, while 42976 is a separate build-push-action upgrade across Docker image workflows.", + "canonical_pr_reason": "PR 42975 is the broader, clearer representative of the cluster theme: it updates multiple GitHub Actions used across workflows specifically for Node 24 compatibility.", "best_issue_reason": null, - "best_pr_reason": "PR 42975 is the better representative of the cluster because it is the broader, more central Actions compatibility upgrade; 42976 is narrower and concerns a different action family.", + "best_pr_reason": "PR 42975 is the best cluster representative because it has the widest scope and the most directly stated purpose, even though it is not a duplicate of PR 42976.", "soft_edge_verdicts": [ { "left": "pull_request:42975", "right": "pull_request:42976", "accept": false, - "reason": "Reject: both touch workflow files, but the concrete changes are different. 42975 upgrades checkout/cache/upload-artifact for Node 24 compatibility; 42976 upgrades docker/build-push-action v5 to v6. They are separate maintenance changes and would not plausibly be merged into one PR as the same fix." + "reason": "Different concrete changes: 42975 upgrades several JS-based actions for Node 24 compatibility, while 42976 mainly upgrades docker/build-push-action to v6. Shared workflow filenames alone are not enough to treat them as the same PR." } ] }, @@ -17976,76 +17958,76 @@ "pull_request:43451|pull_request:45490", "pull_request:45490|pull_request:46162" ], - "prepared_review_unit_hash": "f083c0de25764dbd2637dc7789da90f8ef0e81c233de28a557b57f5b7eb8bcc9", + "prepared_review_unit_hash": "ed7a3768c62003d32cab6c8a255bee3074b662598d489724ba9cb94e21c4b669", "component_budget": { "node_count": 5, "item_count": 5, "soft_pair_count": 7, - "serialized_chars": 16484, - "estimated_input_tokens": 4121, - "estimated_eval_tokens": 8498 + "serialized_chars": 16470, + "estimated_input_tokens": 4118, + "estimated_eval_tokens": 8492 }, "budget": { "node_count": 5, "item_count": 5, "soft_pair_count": 7, - "serialized_chars": 16484, - "estimated_input_tokens": 4121, - "estimated_eval_tokens": 8498 + "serialized_chars": 16470, + "estimated_input_tokens": 4118, + "estimated_eval_tokens": 8492 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These PRs are not duplicates: they add different models (ViT NEPA, Molmo, Molmo2, CTSM) or make a small test-only XPU update. The shared files are mostly generic docs/auto-registry scaffolding, not the same underlying change.", - "confidence": 0.97, + "summary": "These PRs are only loosely similar as model-addition or test-change scaffolding, but they target different models/code paths and are not duplicates.", + "confidence": 0.95, "canonical_issue_reason": null, - "canonical_pr_reason": null, + "canonical_pr_reason": "No canonical PR: 42978 adds ViT NEPA, 43448 adds Molmo, 43451 adds Molmo2, 45490 adds CTSM, and 46162 is a small test-only XPU tweak; they share template files but not the same underlying change.", "best_issue_reason": null, - "best_pr_reason": "pull_request:43451 is the most substantial standalone PR: it introduces the full Molmo2 stack (config, modeling, processing, video processing, auto-mappings, docs, and tests), making it the clearest representative of a complete model addition.", + "best_pr_reason": "If one PR must represent the cluster, 43451 is the strongest standalone candidate because it is the largest and most complete model-addition effort, with the highest review activity; however, it is 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. The overlap is generic model-registration/docs boilerplate, not the same code path or bug." + "reason": "Different model additions (ViT NEPA vs Molmo); shared docs/auto-registry files are generic scaffolding, not the same bug or change." }, { "left": "pull_request:42978", "right": "pull_request:43451", "accept": false, - "reason": "Different model families: ViT NEPA vs Molmo2. Shared auto/docs files are expected for any new model and do not indicate a duplicate change." + "reason": "Different model additions (ViT NEPA vs Molmo2); overlap is only in shared registration/docs files." }, { "left": "pull_request:42978", "right": "pull_request:45490", "accept": false, - "reason": "ViT NEPA vision model addition and CTSM time-series model addition are unrelated changes; only generic registry/docs files overlap." + "reason": "Different model additions (ViT NEPA vs CTSM); same broad transformer repository plumbing, but unrelated concrete changes." }, { "left": "pull_request:43448", "right": "pull_request:43451", "accept": false, - "reason": "Molmo and Molmo2 are separate model additions with distinct source trees and tests; similar naming alone is not enough to treat them as the same change." + "reason": "Molmo and Molmo2 are distinct model ports with different files and functionality; too closely named, but not the same underlying PR." }, { "left": "pull_request:43448", "right": "pull_request:45490", "accept": false, - "reason": "Molmo (vision-language) and CTSM (time-series) are different model integrations; shared auto/docs edits are broad scaffolding, not one bug fix." + "reason": "Molmo vs CTSM are unrelated model additions; shared auto/model registry edits are not enough to merge them as duplicates." }, { "left": "pull_request:43451", "right": "pull_request:45490", "accept": false, - "reason": "Molmo2 model integration and CTSM model integration target different architectures and files; they are not plausibly mergeable into one duplicate PR." + "reason": "Molmo2 and CTSM are separate model implementations with different conversion/modeling/processor code paths." }, { "left": "pull_request:45490", "right": "pull_request:46162", "accept": false, - "reason": "CTSM is a new model-addition PR, while 46162 is a small test-only XPU expectation change for Florence2/LFM2-VL. They do not fix the same underlying issue." + "reason": "CTSM model integration and a small Florence2/LFM2_VL deterministic test tweak are unrelated changes and could not plausibly be merged as one PR." } ] }, @@ -18080,58 +18062,58 @@ "pull_request:43665|pull_request:46102", "pull_request:43665|pull_request:44815" ], - "prepared_review_unit_hash": "b676da0a4a5b1911fbcea6336f3994385cf74534901b2d36072ab3b61516a15c", + "prepared_review_unit_hash": "b80fc94beb90bd761466ca832ecb5f9ea59c818fbba9bcf703be777a77a71cb7", "component_budget": { "node_count": 4, "item_count": 4, "soft_pair_count": 4, - "serialized_chars": 9391, - "estimated_input_tokens": 2348, - "estimated_eval_tokens": 4952 + "serialized_chars": 9382, + "estimated_input_tokens": 2346, + "estimated_eval_tokens": 4948 }, "budget": { "node_count": 4, "item_count": 4, "soft_pair_count": 4, - "serialized_chars": 9391, - "estimated_input_tokens": 2348, - "estimated_eval_tokens": 4952 + "serialized_chars": 9382, + "estimated_input_tokens": 2346, + "estimated_eval_tokens": 4948 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These PRs are not duplicates: they touch different subsystems and fix different problems. The cluster is mostly connected by generic titles like \u201cfix\u201d or \u201cdequant fix,\u201d not by the underlying change.", - "confidence": 0.98, + "summary": "These are four unrelated pull requests that only share shallow signals like test edits or generic fix titles. None appear to be duplicate fixes for the same underlying bug, so no soft edges should be merged.", + "confidence": 0.91, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #44815 is the most substantive and concrete change in the set: it has a specific loading/dequantization fix spanning core model loading and fine-grained FP8 integration, with a clearer technical scope than the test-only or broadly named PRs.", + "canonical_pr_reason": "PR 44815 is the most substantive and clearly scoped fix: it updates core loading and FP8 dequant logic across multiple affected files, with a concrete code-path change rather than test-only edits.", "best_issue_reason": null, - "best_pr_reason": "PR #44815 is the best representative of the cluster because it contains an identifiable code-path fix and meaningful implementation detail, whereas #43665 is effectively test noise, #43291 is Whisper-specific, and #46102 is an FSDP-specific cleanup/fix.", + "best_pr_reason": "PR 44815 is the best representative fix because it contains the real implementation change and associated test coverage, whereas the others are either test-only, unrelated, or narrower in scope.", "soft_edge_verdicts": [ { "left": "pull_request:43291", "right": "pull_request:43665", "accept": false, - "reason": "Whisper tokenizer decoding/test fix vs unrelated CLIP/ViT test edits; no shared bug or code path." + "reason": "Both touch tests, but they target different components and different failure modes; no shared concrete bug or mergeable single fix is evident." }, { "left": "pull_request:43291", "right": "pull_request:44815", "accept": false, - "reason": "Whisper decoding fix vs FP8 dequant/loading fix in unrelated modules; not the same underlying issue." + "reason": "Whisper tokenizer/test changes are unrelated to the dequant/core-loading FP8 fix; they do not address the same code-path problem." }, { "left": "pull_request:43665", "right": "pull_request:46102", "accept": false, - "reason": "Test assertion changes in CLIP/ViT vs FSDP tied-weights cleanup; clearly different changes." + "reason": "These are separate test/code changes in different subsystems (vision model tests vs FSDP tied-weights handling); they are not the same bug." }, { "left": "pull_request:43665", "right": "pull_request:44815", "accept": false, - "reason": "Test-only edits in vision models vs dequantization/loading changes; no plausible single merged PR." + "reason": "Despite both mentioning tests, the underlying fixes are unrelated: CLIP/ViT test edits versus FP8 dequant and model-loading behavior." } ] }, @@ -18168,70 +18150,70 @@ "pull_request:45932|pull_request:46100", "pull_request:46100|pull_request:46114" ], - "prepared_review_unit_hash": "02cc4a059d7d3ecd6bcf4e2acb8e68a288ac5eaea2bc20ff2bcce1ee72872f1a", + "prepared_review_unit_hash": "2c6668ccf409a08ab4bb022c52fbc3b3841cbb81e37ec1bbb42117662256e736", "component_budget": { "node_count": 5, "item_count": 5, "soft_pair_count": 5, - "serialized_chars": 11340, - "estimated_input_tokens": 2835, - "estimated_eval_tokens": 5926 + "serialized_chars": 11341, + "estimated_input_tokens": 2836, + "estimated_eval_tokens": 5928 }, "budget": { "node_count": 5, "item_count": 5, "soft_pair_count": 5, - "serialized_chars": 11340, - "estimated_input_tokens": 2835, - "estimated_eval_tokens": 5926 + "serialized_chars": 11341, + "estimated_input_tokens": 2836, + "estimated_eval_tokens": 5928 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Mostly unrelated placeholder/test PRs. The only true duplicate pair is #46100 and #46114, which have the same ViT test-file edit; the other PRs only overlap superficially on broad test/workflow areas.", - "confidence": 0.93, + "summary": "Only one clear duplicate pair appears here: the two ViT test-migration PRs with the same exact failing asserts. The other PRs are all different bot/CI or broader instrumentation experiments and should not be merged as duplicates.", + "confidence": 0.84, "canonical_issue_reason": null, - "canonical_pr_reason": "#46114 is the best canonical PR for the duplicate pair because it has the same diff as #46100 but a clearer title/body indicating the forked test-migration/OTEL purpose.", + "canonical_pr_reason": "PR #46100 is the cleanest representative of the duplicate ViT test-migration change: it is a minimal, self-contained diff in one test file, while #46114 appears to be the forked copy of the same patch.", "best_issue_reason": null, - "best_pr_reason": "#46114 is the best representative of the cluster\u2019s only real duplicate set: it is effectively identical to #46100 while being more explicit about intent.", + "best_pr_reason": "PR #46100 is the best standalone PR for the cluster because it is the smallest and clearest instance of the duplicated change, with no extra unrelated workflow or instrumentation edits.", "soft_edge_verdicts": [ { "left": "pull_request:43488", - "right": "pull_request:43532", + "right": "pull_request:46114", "accept": false, - "reason": "Both are dummy \"don't merge\" PRs and both touch the Vit test file, but the actual changes are different (setup/version/deps vs workflow/notification/testing-utils edits). Not the same underlying change." + "reason": "Both mention the ViT test file, but #43488 is mainly a setup/version/dependency bot-check change while #46114 is a test-migration fork; they are not the same concrete fix." }, { "left": "pull_request:43488", - "right": "pull_request:46114", + "right": "pull_request:43532", "accept": false, - "reason": "Only superficial overlap on a Vit test file. #43488 is a setup/version/dependency bot check, while #46114 is a test-migration PR with failing assertions. Different code paths and goals." + "reason": "These are different repository-bot/CI experiments with different edited files and goals; they do not share the same underlying change." }, { "left": "pull_request:43532", "right": "pull_request:46114", "accept": false, - "reason": "Shared test-file name is not enough; #43532 is mainly CI/workflow infrastructure and helper changes, while #46114 is a tiny ViT test migration. Not mergeable as one PR." + "reason": "#43532 is a workflow/self-comment CI change, whereas #46114 is a ViT test migration patch; same broad area, but not the same code-path change." }, { "left": "pull_request:45932", "right": "pull_request:46100", "accept": false, - "reason": "#45932 is a broad telemetry/grafana-related PR across workflows and multiple model files; #46100 only edits the ViT test file to add failing assertions. Different underlying changes." + "reason": "#45932 is a much broader grafana/telemetry and model-file change set; despite both touching ViT tests, it does not look like the same concrete change as #46100." }, { "left": "pull_request:46100", "right": "pull_request:46114", "accept": true, - "reason": "These are effectively identical PRs: same ViT test-file diff adding the same failing assertions, with only title/body differences. They look like the same underlying change." + "reason": "These are effectively the same PR: same single-file ViT test diff, same added assertions, and the titles indicate duplicate forked test-migration efforts." } ] }, "evaluator_result": { "accept": true, - "feedback": "The conclusions are grounded in the packet and the soft-edge calls are conservative. The only clearly identical diff pair is #46100/#46114, while the other PR pairs share at most superficial filename or test/workflow overlap and do not look mergeable as one bugfix PR." + "feedback": "The review is grounded in the packet and stays conservative. The only clear duplicate pair is 46100/46114: they have the same single-file ViT test diff with identical added assertions, and the titles support the forked-copy interpretation. The other pairings are correctly rejected because they share only loose topical similarity, not the same concrete fix." }, "evaluator_used": true, "retried": false, @@ -18258,12 +18240,12 @@ "soft_pairs": [ "pull_request:43612|pull_request:43613" ], - "prepared_review_unit_hash": "46c0665b308c5c86f8be1e0f5d3a82f1e40ac1c9458f843a2f8c79f6fd305781", + "prepared_review_unit_hash": "e2f796eda585d6c510b729fd1c38b7213100a2344f318cabf238a3677f0c6b69", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 6083, + "serialized_chars": 6082, "estimated_input_tokens": 1521, "estimated_eval_tokens": 3298 }, @@ -18271,7 +18253,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 6083, + "serialized_chars": 6082, "estimated_input_tokens": 1521, "estimated_eval_tokens": 3298 }, @@ -18280,18 +18262,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are two separate PRs adding different promptable segmentation pipelines; their overlap is mostly shared scaffolding/docs, not the same underlying change.", - "confidence": 0.98, + "summary": "Two open PRs add similarly named but distinct pipelines: one for Promptable Concept Segmentation and one for Promptable Visual Segmentation. They share repository plumbing and docs, but the underlying feature scope, model integrations, and tests are different, so they should not be merged as duplicates.", + "confidence": 0.96, "canonical_issue_reason": null, - "canonical_pr_reason": null, + "canonical_pr_reason": "No single canonical PR for a duplicate cluster: these are separate feature additions with different tasks, model stacks, and test coverage.", "best_issue_reason": null, - "best_pr_reason": "PR 43613 is the broader, more complete pipeline addition (more model/test coverage and implementation files), so it is the better representative if one PR must be picked.", + "best_pr_reason": "PR 43613 is the better representative of the cluster only because it is the broader, more comprehensive pipeline addition; however, it is not a duplicate of 43612.", "soft_edge_verdicts": [ { "left": "pull_request:43612", "right": "pull_request:43613", "accept": false, - "reason": "Different features: promptable concept segmentation vs promptable visual segmentation. Shared docs/registry/pipeline boilerplate is expected and not enough to merge them as one PR." + "reason": "Different concrete features: concept segmentation vs visual segmentation, with different docs, model files, and tests. Shared infrastructure alone is not enough to treat them as the same change." } ] }, @@ -18321,40 +18303,40 @@ "soft_pairs": [ "pull_request:43651|pull_request:43663" ], - "prepared_review_unit_hash": "5e3995bf6ffde54256bca1688f1593ea86f1bc316969bdd838a123c9522a7a5a", + "prepared_review_unit_hash": "a7e226d0ba4b8d3c03e016adddcd7163523c471ee7dd8593843bfc7524e048be", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4292, - "estimated_input_tokens": 1073, - "estimated_eval_tokens": 2402 + "serialized_chars": 4293, + "estimated_input_tokens": 1074, + "estimated_eval_tokens": 2404 }, "budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4292, - "estimated_input_tokens": 1073, - "estimated_eval_tokens": 2402 + "serialized_chars": 4293, + "estimated_input_tokens": 1074, + "estimated_eval_tokens": 2404 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These two PRs both touch `trainer.py`, but they fix different Trainer extension points: one changes gradient-accumulation loss scaling, the other adds an overrideable signature-column helper for dataset filtering. They should not be clustered as duplicates.", + "summary": "Two open PRs in `trainer.py` touch trainer customization, but they address different code paths: gradient-accumulation loss scaling vs. signature-column filtering. They are not duplicates.", "confidence": 0.96, "canonical_issue_reason": null, - "canonical_pr_reason": "No single canonical PR: #43651 and #43663 address separate code paths and separate user-facing behaviors, so neither subsumes the other as a duplicate representative.", + "canonical_pr_reason": "PR 43651 is the more concrete behavior change: it updates `training_step` loss scaling semantics and adds a dedicated hook for custom trainers. If one PR must represent the cluster, this is the clearer canonical candidate.", "best_issue_reason": null, - "best_pr_reason": "#43651 is the cleaner cluster anchor if one must be chosen: it is a focused, concrete behavior change with a specific trainer loss-scaling path and a clearer standalone purpose.", + "best_pr_reason": "PR 43651 is the strongest standalone PR here because it fixes a specific training-path behavior with a direct code change, whereas PR 43663 is a separate refactor/hook for column filtering.", "soft_edge_verdicts": [ { "left": "pull_request:43651", "right": "pull_request:43663", "accept": false, - "reason": "Same file, but different underlying changes: gradient-accumulation loss scaling vs signature-column filtering. They are independent trainer improvements and could not plausibly be merged as one fix." + "reason": "Different trainer extensibility features in the same file, but not the same underlying bug or change: one concerns GA loss scaling, the other signature-column inspection. They are not plausibly one merged PR." } ] }, @@ -18384,12 +18366,12 @@ "soft_pairs": [ "pull_request:43656|pull_request:43842" ], - "prepared_review_unit_hash": "7cdef20d78d7967505ae84821948aea210b518ea19ac3a454ef278ea5ced89bb", + "prepared_review_unit_hash": "eb024b05c1c4f838ea3d6e833be36b0219d8574a00350eab1b603a7b554c137a", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 3663, + "serialized_chars": 3664, "estimated_input_tokens": 916, "estimated_eval_tokens": 2088 }, @@ -18397,7 +18379,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 3663, + "serialized_chars": 3664, "estimated_input_tokens": 916, "estimated_eval_tokens": 2088 }, @@ -18406,28 +18388,25 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both PRs target the same runtime TypeAdapter annotation failure in `src/transformers/cli/serve.py`, but `43842` is the narrower and more explicit fix; `43656` is a broader typing/annotations cleanup that also touches the same bug.", - "confidence": 0.84, + "summary": "Two open PRs touch the same CLI file and both address annotation-related NameErrors, but they fix different runtime causes: one is a broader typing/future-annotations cleanup, the other specifically guards pydantic/TypeAdapter for environments without pydantic installed.", + "confidence": 0.76, "canonical_issue_reason": null, - "canonical_pr_reason": "43842 most directly matches the concrete bug: it fixes the `TypeAdapter` NameError with a minimal, explicit `TYPE_CHECKING`/string-annotation change tied to the pydantic-absent case.", + "canonical_pr_reason": "PR 43842 is the tighter, more concrete fix: it directly addresses the TypeAdapter NameError when pydantic is absent and has a clear, narrowly scoped change set.", "best_issue_reason": null, - "best_pr_reason": "43842 is the best representative of the cluster because it targets the exact runtime failure with the least collateral change and the clearest bug description.", + "best_pr_reason": "PR 43842 is the best standalone candidate because it fixes a specific, reproducible runtime failure with minimal collateral changes and an explicit issue target.", "soft_edge_verdicts": [ { "left": "pull_request:43656", "right": "pull_request:43842", - "accept": true, - "reason": "Both fix the same underlying `TypeAdapter` NameError in the CLI serve path; they are alternative implementations of the same bug fix and could plausibly be consolidated into one PR." + "accept": false, + "reason": "Reject: despite similar titles and shared filename, these are not the same underlying bug. 43656 is a broader typing/future-annotations refactor around serve.py, while 43842 specifically fixes TypeAdapter import/name resolution when pydantic is not installed." } ] }, - "evaluator_result": { - "accept": true, - "feedback": "The summary is grounded in the packet: both PRs modify `src/transformers/cli/serve.py` and their titles/diffs point to the same `TypeAdapter` NameError class of bug. The soft-edge acceptance is conservative enough because the two fixes are compatible and could plausibly be folded into one PR, with `43842` being the narrower, more explicit variant." - }, - "evaluator_used": true, + "evaluator_result": null, + "evaluator_used": false, "retried": false, - "accepted_nontrivial_soft_edge": true, + "accepted_nontrivial_soft_edge": false, "error_kind": null, "error_message": null }, @@ -18577,8 +18556,8 @@ "pull_request:44071|pull_request:44086", "pull_request:44073|pull_request:44101", "pull_request:44017|pull_request:44071", - "pull_request:44056|pull_request:44101", "pull_request:44059|pull_request:44071", + "pull_request:44056|pull_request:44101", "pull_request:44073|pull_request:44161", "pull_request:44059|pull_request:44068", "pull_request:44018|pull_request:44059", @@ -18612,412 +18591,430 @@ "pull_request:44044|pull_request:44129", "pull_request:44044|pull_request:44154" ], - "prepared_review_unit_hash": "1166def1b3c415d6317bb6a15703929009acad7cb330eba4df53e583a1849cdf", + "prepared_review_unit_hash": "b410f3365977c6c09ed1b21c6b872882f6e649192146e6ac416ecf40cfda6f1c", "component_budget": { "node_count": 28, "item_count": 28, "soft_pair_count": 78, - "serialized_chars": 84900, - "estimated_input_tokens": 21225, - "estimated_eval_tokens": 42706 + "serialized_chars": 84902, + "estimated_input_tokens": 21226, + "estimated_eval_tokens": 42708 }, "budget": { "node_count": 27, "item_count": 27, "soft_pair_count": 72, - "serialized_chars": 75096, - "estimated_input_tokens": 18774, - "estimated_eval_tokens": 37804 + "serialized_chars": 75098, + "estimated_input_tokens": 18775, + "estimated_eval_tokens": 37806 }, "overflow_policy": "truncate_then_skip", "trimmed": false, "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "This cluster is mostly a set of parallel model-specific output-tracing refactors tied to the same umbrella tracking issue, not one duplicate bug. The only clear duplicates are within the GPT-Neo pair and the GPT-J/CodeGen pair(s).", - "confidence": 0.9, - "canonical_issue_reason": null, - "canonical_pr_reason": "PR 44066 is the clearest representative of the concrete GPT-J/CodeGen output-tracing refactor: it has an explicit issue link, touches the right code paths, and is the most complete multi-file instance of that same change family.", - "best_issue_reason": null, - "best_pr_reason": "PR 44066 is the best single representative of the deduplicated subset because it directly implements the standardized output-tracing change for GPT-J/CodeGen and is explicitly tied to the umbrella issue.", + "summary": "This cluster is centered on the umbrella standardized output-tracing refactor tracked by issue #43979, with multiple model-specific PRs. The clearest duplicate set is the GPT-J/GPT-Neo line; most other pairs are just unrelated models sharing the same refactor theme and should stay separate.", + "confidence": 0.82, + "canonical_issue_reason": "#43979 is the shared tracking issue for the standardized output-tracing work and is referenced by many of the PRs in this cluster.", + "canonical_pr_reason": "#44722 is the most complete representative of the GPT-J/CodeGen standardized output-tracing refactor and best subsumes the narrower GPT-J-only variants.", + "best_issue_reason": "#43979 is the only clear cluster-level issue anchor and matches the common refactor theme across the accepted duplicates.", + "best_pr_reason": "#44722 best represents the actual concrete change set among the duplicates because it covers the shared GPT-J path and also includes the related CodeGen update.", "soft_edge_verdicts": [ { "left": "pull_request:43996", "right": "pull_request:44085", "accept": false, - "reason": "Different model families and different code paths; same umbrella issue is not enough to merge." + "reason": "Both are output-tracing refactors, but they target different model families and code paths (CVT/FNet vs GPT-J)." }, { "left": "pull_request:43996", "right": "pull_request:44044", "accept": false, - "reason": "Unrelated model implementations; similar refactor theme only." + "reason": "Different models and implementations; only the umbrella refactor theme is shared, not the concrete bug/change." }, { "left": "pull_request:44066", "right": "pull_request:44085", "accept": true, - "reason": "Both are GPT-J output-tracing refactors on the same code path and could be merged into one PR." + "reason": "Both change the GPT-J output-tracing path in modeling_gptj.py; 44066 is a broader variant that also touches CodeGen, but the shared GPT-J fix is the same underlying change." }, { "left": "pull_request:44007", "right": "pull_request:44072", "accept": false, - "reason": "Different models; only the output-tracing pattern is shared." + "reason": "Same tracking issue, but the actual code paths are unrelated model families (ResNet/RegNet vs EfficientNet/RT-DETR ResNet)." }, { "left": "pull_request:44072", "right": "pull_request:44722", "accept": false, - "reason": "Different model families and code paths; shared issue target does not imply duplicate." + "reason": "Both are output-tracing refactors, but they apply to different models and are not the same concrete fix." }, { "left": "pull_request:44066", "right": "pull_request:44072", "accept": false, - "reason": "GPT-J/CodeGen versus EfficientNet; not the same underlying change." + "reason": "GPT-J/CodeGen versus EfficientNet; too different to merge as the same underlying change." }, { "left": "pull_request:44013", "right": "pull_request:44044", "accept": false, - "reason": "Different model subsystems; same refactor style only." + "reason": "Different model families and different forward/output plumbing; shared issue target is not enough." }, { "left": "pull_request:44066", "right": "pull_request:44086", "accept": false, - "reason": "GPT-J/CodeGen and MGP-STR are unrelated implementations." + "reason": "GPT-J/CodeGen refactor vs MGP-STR; same theme, but not the same code-path problem." }, { "left": "pull_request:44066", "right": "pull_request:44071", "accept": false, - "reason": "Different model families; shared output-tracing theme is too broad." + "reason": "GPT-J/CodeGen and MPT are separate implementations; not the same concrete bug/change." }, { "left": "pull_request:44018", "right": "pull_request:44068", "accept": true, - "reason": "Both target GPT-Neo output-tracing in the same file/code path and look like duplicate variants of the same fix." + "reason": "Both are GPT-Neo output-tracing refactors on the same model/file, so they look like the same underlying change." }, { "left": "pull_request:44066", "right": "pull_request:44068", "accept": false, - "reason": "GPT-J/CodeGen versus GPT-Neo; not the same concrete change." + "reason": "GPT-J/CodeGen versus GPT-Neo; these are separate model-specific refactors." }, { "left": "pull_request:44018", "right": "pull_request:44066", "accept": false, - "reason": "GPT-Neo and GPT-J/CodeGen are different model implementations." + "reason": "Both are output-tracing work, but for different model code paths (GPT-Neo vs GPT-J/CodeGen)." }, { "left": "pull_request:44019", "right": "pull_request:44722", "accept": false, - "reason": "ResNet versus GPT-J/CodeGen; same refactor pattern only." + "reason": "ResNet output tracing and GPT-J/CodeGen output tracing are unrelated model changes." }, { "left": "pull_request:44086", "right": "pull_request:44722", "accept": false, - "reason": "Different models and different code paths." + "reason": "MGP-STR and GPT-J/CodeGen are different model implementations; not mergeable as one fix." }, { "left": "pull_request:44019", "right": "pull_request:44085", "accept": false, - "reason": "ResNet and GPT-J are unrelated changes." + "reason": "Different models and different refactor scopes; only the generic output-tracing idea overlaps." }, { "left": "pull_request:44018", "right": "pull_request:44019", "accept": false, - "reason": "Different model families; not mergeable as one fix." + "reason": "GPT-Neo and ResNet are unrelated code paths." }, { "left": "pull_request:44019", "right": "pull_request:44071", "accept": false, - "reason": "ResNet versus MPT; same umbrella pattern only." + "reason": "ResNet versus MPT; same broad pattern, different concrete changes." }, { "left": "pull_request:44068", "right": "pull_request:44116", "accept": false, - "reason": "GPT-Neo and Flaubert are different code paths." + "reason": "GPT-Neo and Flaubert are unrelated model families." }, { "left": "pull_request:44068", "right": "pull_request:44722", "accept": false, - "reason": "GPT-Neo and GPT-J/CodeGen are separate model changes." + "reason": "GPT-Neo refactor vs GPT-J/CodeGen refactor; not the same underlying model change." }, { "left": "pull_request:44018", "right": "pull_request:44116", "accept": false, - "reason": "GPT-Neo versus Flaubert; unrelated implementations." + "reason": "GPT-Neo and Flaubert touch different model code paths." }, { "left": "pull_request:44071", "right": "pull_request:44722", "accept": false, - "reason": "MPT versus GPT-J/CodeGen; shared refactor style only." + "reason": "MPT and GPT-J/CodeGen are separate implementations, so these are not duplicates." }, { "left": "pull_request:44018", "right": "pull_request:44086", "accept": false, - "reason": "GPT-Neo and MGP-STR are unrelated model code paths." + "reason": "GPT-Neo versus MGP-STR; shared topic is too broad." }, { "left": "pull_request:44018", "right": "pull_request:44085", "accept": false, - "reason": "GPT-Neo versus GPT-J; not the same underlying change." + "reason": "GPT-Neo and GPT-J are different model-specific output-tracing updates." }, { "left": "pull_request:44068", "right": "pull_request:44086", "accept": false, - "reason": "Different model families and implementations." + "reason": "Different model families and code paths." }, { "left": "pull_request:44071", "right": "pull_request:44085", "accept": false, - "reason": "MPT and GPT-J are unrelated." + "reason": "MPT and GPT-J are not the same fix." }, { "left": "pull_request:44059", "right": "pull_request:44086", "accept": false, - "reason": "GPT-2 versus MGP-STR; same tracing refactor theme only." + "reason": "GPT-2 output tracing and MGP-STR output tracing are unrelated implementations." }, { "left": "pull_request:44085", "right": "pull_request:44086", "accept": false, - "reason": "Both are output-tracing refactors, but for different model code paths." + "reason": "GPT-J/CodeGen versus MGP-STR; only the refactor style overlaps." }, { "left": "pull_request:44098", "right": "pull_request:44101", "accept": false, - "reason": "ViLT and Flaubert/XLM are different model families." + "reason": "ViLT and Flaubert are different model families and do not share the same code-path change." }, { "left": "pull_request:44059", "right": "pull_request:44722", "accept": false, - "reason": "GPT-2 versus GPT-J/CodeGen; similar pattern, different code paths." + "reason": "GPT-2 and GPT-J/CodeGen are separate model refactors." }, { "left": "pull_request:44018", "right": "pull_request:44071", "accept": false, - "reason": "GPT-Neo and MPT are unrelated model implementations." + "reason": "GPT-Neo and MPT are distinct model implementations." }, { "left": "pull_request:44017", "right": "pull_request:44722", "accept": false, - "reason": "SegFormer and GPT-J/CodeGen are different subsystems." + "reason": "SegFormer and GPT-J/CodeGen are unrelated concrete changes." }, { "left": "pull_request:44017", "right": "pull_request:44161", "accept": false, - "reason": "SegFormer and LongT5 are unrelated code paths." + "reason": "SegFormer and LongT5 are different model code paths." }, { - "left": "pull_request:44073", - "right": "pull_request:44101", + "left": "pull_request:44017", + "right": "pull_request:44073", "accept": false, - "reason": "ViLT versus Flaubert/XLM; not the same concrete change." + "reason": "SegFormer and VisualBert are unrelated model-specific refactors." }, { - "left": "pull_request:43997", - "right": "pull_request:44073", + "left": "pull_request:44068", + "right": "pull_request:44071", "accept": false, - "reason": "RegNet versus ViLT; same output-tracing motif only." + "reason": "GPT-Neo and MPT are separate implementations." }, { "left": "pull_request:43997", - "right": "pull_request:44056", + "right": "pull_request:44073", "accept": false, - "reason": "RegNet versus MPNet; different models and code paths." + "reason": "RegNet and VisualBert are different models; same refactor theme only." }, { "left": "pull_request:44071", "right": "pull_request:44076", "accept": false, - "reason": "MPT versus ImageGPT; unrelated implementations." + "reason": "MPT and ImageGPT are unrelated model code paths." }, { "left": "pull_request:44074", "right": "pull_request:44101", "accept": false, - "reason": "TextNet versus Flaubert/XLM; not a duplicate change." + "reason": "TextNet and Flaubert are different underlying changes." }, { "left": "pull_request:44056", "right": "pull_request:44161", "accept": false, - "reason": "MPNet and LongT5 are unrelated model code paths." + "reason": "MPNet and LongT5 are not the same concrete bug/change." }, { "left": "pull_request:43997", - "right": "pull_request:44018", + "right": "pull_request:44017", + "accept": false, + "reason": "RegNet and SegFormer are unrelated model refactors." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44056", "accept": false, - "reason": "RegNet and GPT-Neo are different subsystems." + "reason": "RegNet and MPNet are separate model-specific updates." }, { "left": "pull_request:44071", "right": "pull_request:44129", "accept": false, - "reason": "MPT and SpeechT5 are unrelated implementations." + "reason": "MPT and SpeechT5 are different code paths." }, { "left": "pull_request:44010", "right": "pull_request:44071", "accept": false, - "reason": "SqueezeBert and MPT are different model code paths." + "reason": "SqueezeBert and MPT are unrelated model implementations." }, { "left": "pull_request:44010", "right": "pull_request:44073", "accept": false, - "reason": "SqueezeBert and ViLT are unrelated changes." + "reason": "SqueezeBert and VisualBert do not fix the same underlying issue." }, { "left": "pull_request:44018", - "right": "pull_request:44059", + "right": "pull_request:44129", "accept": false, - "reason": "GPT-Neo and GPT-2 are separate model implementations." + "reason": "GPT-Neo and SpeechT5 are different model families." }, { - "left": "pull_request:44010", - "right": "pull_request:44044", + "left": "pull_request:44074", + "right": "pull_request:44161", "accept": false, - "reason": "SqueezeBert and DeBERTa are unrelated." + "reason": "TextNet and LongT5 are unrelated concrete changes." }, { - "left": "pull_request:44044", - "right": "pull_request:44074", + "left": "pull_request:43997", + "right": "pull_request:44018", "accept": false, - "reason": "DeBERTa and TextNet are different model code paths." + "reason": "RegNet and GPT-Neo are different model-specific refactors." }, { - "left": "pull_request:44044", - "right": "pull_request:44129", + "left": "pull_request:43997", + "right": "pull_request:44161", + "accept": false, + "reason": "RegNet and LongT5 are unrelated code paths." + }, + { + "left": "pull_request:44074", + "right": "pull_request:44161", "accept": false, - "reason": "DeBERTa and SpeechT5 are unrelated." + "reason": "TextNet and LongT5 are different model implementations." }, { "left": "pull_request:44018", + "right": "pull_request:44059", + "accept": false, + "reason": "GPT-Neo and GPT-2 are separate model refactors." + }, + { + "left": "pull_request:44101", "right": "pull_request:44161", "accept": false, - "reason": "GPT-Neo and LongT5 are different model implementations." + "reason": "Flaubert and LongT5 are unrelated underlying changes." }, { "left": "pull_request:43997", "right": "pull_request:44074", "accept": false, - "reason": "RegNet versus TextNet; same refactor pattern only." + "reason": "RegNet and TextNet are different model code paths." }, { "left": "pull_request:43997", "right": "pull_request:44129", "accept": false, - "reason": "RegNet and SpeechT5 are different code paths." + "reason": "RegNet and SpeechT5 are unrelated model implementations." }, { "left": "pull_request:44071", - "right": "pull_request:44086", + "right": "pull_request:44129", "accept": false, - "reason": "MPT and MGP-STR are unrelated model implementations." + "reason": "MPT and SpeechT5 are different concrete changes." }, { "left": "pull_request:44010", "right": "pull_request:44018", "accept": false, - "reason": "SqueezeBert and GPT-Neo are different model families." + "reason": "SqueezeBert and GPT-Neo are separate model-specific refactors." }, { "left": "pull_request:43996", "right": "pull_request:44161", "accept": false, - "reason": "CVT/FNet and LongT5 are unrelated changes." + "reason": "CVT/FNet and LongT5 are different model families." }, { "left": "pull_request:43996", "right": "pull_request:44017", "accept": false, - "reason": "CVT/FNet and SegFormer are different model code paths." + "reason": "CVT/FNet and SegFormer do not share the same concrete code path." }, { "left": "pull_request:44010", "right": "pull_request:44044", "accept": false, - "reason": "SqueezeBert and DeBERTa are unrelated implementations." + "reason": "SqueezeBert and DeBERTa v2 are unrelated model-specific changes." }, { "left": "pull_request:44044", - "right": "pull_request:44154", + "right": "pull_request:44074", "accept": false, - "reason": "DeBERTa and VITS are different model families." + "reason": "DeBERTa v2 and TextNet are different underlying fixes." }, { - "left": "pull_request:44072", - "right": "pull_request:44154", - "accept": false, - "reason": "EfficientNet and VITS are unrelated code paths." + "left": "pull_request:44066", + "right": "pull_request:44722", + "accept": true, + "reason": "They address the same GPT-J output-tracing refactor; 44722 is a later, broader variant that still includes the shared GPT-J fix." }, { - "left": "pull_request:44101", - "right": "pull_request:44161", - "accept": false, - "reason": "Flaubert/XLM and LongT5 are unrelated model implementations." + "left": "pull_request:44085", + "right": "pull_request:44722", + "accept": true, + "reason": "Both touch the GPT-J output-tracing path in modeling_gptj.py, so they look like the same underlying bug/change despite 44722 also including CodeGen." }, { - "left": "pull_request:44073", - "right": "pull_request:44161", + "left": "pull_request:44085", + "right": "pull_request:44129", "accept": false, - "reason": "ViLT and LongT5 are different code paths." + "reason": "GPT-J and SpeechT5 are unrelated model code paths." }, { - "left": "pull_request:44059", - "right": "pull_request:44068", + "left": "pull_request:43996", + "right": "pull_request:44007", "accept": false, - "reason": "GPT-2 and GPT-Neo are separate model implementations." + "reason": "Both are output-tracing refactors, but for different model families and concrete implementations." }, { - "left": "pull_request:44059", - "right": "pull_request:44071", + "left": "pull_request:43996", + "right": "pull_request:44129", "accept": false, - "reason": "GPT-2 and MPT are unrelated." + "reason": "CVT/FNet and SpeechT5 are not the same fix." }, { - "left": "pull_request:44068", - "right": "pull_request:44161", + "left": "pull_request:44044", + "right": "pull_request:44129", "accept": false, - "reason": "GPT-Neo and LongT5 are different models." + "reason": "DeBERTa v2 and SpeechT5 are separate model-specific changes." }, { - "left": "pull_request:44068", - "right": "pull_request:44071", + "left": "pull_request:44044", + "right": "pull_request:44154", "accept": false, - "reason": "GPT-Neo and MPT are unrelated code paths." + "reason": "DeBERTa v2 and VITS are unrelated model implementations." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is well grounded in the packet and stays conservative: it treats the cluster as mostly parallel model-specific refactors, and only marks the GPT-Neo and GPT-J/CodeGen pairs as clear mergeable duplicates. The listed soft-edge rejections are appropriately cautious for different model families/code paths, and the accept verdicts are supported by shared file/code-path evidence." + "feedback": "The summary is grounded in the packet: #43979 is the clear shared issue anchor, and the chosen canonical PR #44722 is a reasonable broader representative of the GPT-J/CodeGen subset. The soft-edge verdicts are conservative overall, with accepts only where the diffs share the same model/file and likely the same bugfix, and rejects where only the high-level refactor theme overlaps." }, "evaluator_used": true, "retried": false, @@ -19055,14 +19052,14 @@ "pull_request:43997|pull_request:44010", "pull_request:44010|pull_request:44129" ], - "prepared_review_unit_hash": "f9be3b987b95733fb8b21b2459a14eb6983db4a6deafef9040a148ff9dc371ec", + "prepared_review_unit_hash": "91f4fcae2c6b189afeb35d562bbc343bcdc5a5db23bcf096c76932889c8500bc", "component_budget": { "node_count": 28, "item_count": 28, "soft_pair_count": 78, - "serialized_chars": 84900, - "estimated_input_tokens": 21225, - "estimated_eval_tokens": 42706 + "serialized_chars": 84902, + "estimated_input_tokens": 21226, + "estimated_eval_tokens": 42708 }, "budget": { "node_count": 8, @@ -19077,48 +19074,48 @@ "aggressively_trimmed": false, "split": true, "analyst_result": { - "summary": "All items are model-specific PRs in the same broad output-tracing migration wave, but they touch different architectures and code paths, so they are not duplicates of one another.", - "confidence": 0.95, + "summary": "These PRs are all model-specific refactors toward standardized output tracing/capture, but they target different architectures and code paths. They are related by pattern, not duplicates.", + "confidence": 0.96, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 44129 is the strongest anchor: it has an explicit issue target, a concrete SpeechT5-specific refactor, and is clearly scoped to the standardized output-capture migration pattern.", + "canonical_pr_reason": "pull_request:44073 is the clearest representative of the shared refactor pattern: it explicitly adds _can_record_outputs plus can_return_tuple/capture_outputs for a single model, with a clean, self-contained change.", "best_issue_reason": null, - "best_pr_reason": "PR 44129 is the best representative of the cluster because it is explicit about its tracking issue and presents a focused, model-local implementation of the output tracing refactor.", + "best_pr_reason": "pull_request:44073 is the best single representative PR for this cluster because it most clearly expresses the standardized output-tracing migration pattern, while still being narrowly scoped to one model.", "soft_edge_verdicts": [ { "left": "pull_request:44024", "right": "pull_request:44076", "accept": false, - "reason": "Both are standardized output-tracing refactors, but FocalNet and ImageGPT are different model implementations with different forward paths and hook targets; not the same change." + "reason": "Same broad output-tracing migration theme, but different models (FocalNet vs ImageGPT) and different code paths; not plausibly one merged fix." }, { "left": "pull_request:44073", "right": "pull_request:44074", "accept": false, - "reason": "VisualBert and TextNet are unrelated model families; these are separate per-model migrations, not one shared bug fix or mergeable PR." + "reason": "VisualBert and TextNet are separate model implementations; these are parallel refactors, not the same underlying bug/change." }, { "left": "pull_request:44056", "right": "pull_request:44073", "accept": false, - "reason": "MPNet and VisualBert change different encoder stacks and output-recording hooks; same migration style, but not the same concrete code-path problem." + "reason": "Both refactor output recording, but they touch different model stacks (MPNet vs VisualBert) and are not the same concrete change." }, { "left": "pull_request:44056", "right": "pull_request:44074", "accept": false, - "reason": "Different model families and different output-capture adaptations; too broad to treat as duplicates." + "reason": "Different models and different forward/output plumbing; too broad to treat as the same PR." }, { "left": "pull_request:43997", "right": "pull_request:44010", "accept": false, - "reason": "RegNet and SqueezeBert are unrelated architectures, so these are separate refactors rather than duplicate fixes." + "reason": "RegNet and SqueezeBert are unrelated model code paths; shared decorator migration alone is not enough for duplication." }, { "left": "pull_request:44010", "right": "pull_request:44129", "accept": false, - "reason": "SqueezeBert and SpeechT5 modify different model internals; they share a migration theme but not the same underlying code-path or mergeable fix." + "reason": "SqueezeBert and SpeechT5 address different model internals and output semantics; not the same concrete fix." } ] }, @@ -19204,12 +19201,12 @@ "pull_request:44002|pull_request:46106", "pull_request:43998|pull_request:46106" ], - "prepared_review_unit_hash": "e54f3abde29f60ddf451c83bea5769ec72ad1b11bf48d82ada681238371e448e", + "prepared_review_unit_hash": "907983b1b73ad8ec5a92e896544ed32c60d398c871404c7f03b71207c9ac5307", "component_budget": { "node_count": 14, "item_count": 14, "soft_pair_count": 45, - "serialized_chars": 43786, + "serialized_chars": 43785, "estimated_input_tokens": 10947, "estimated_eval_tokens": 22150 }, @@ -19217,7 +19214,7 @@ "node_count": 14, "item_count": 14, "soft_pair_count": 45, - "serialized_chars": 43786, + "serialized_chars": 43785, "estimated_input_tokens": 10947, "estimated_eval_tokens": 22150 }, @@ -19226,282 +19223,282 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are not duplicate PRs; they are separate output-tracing refactors applied to different model code paths, plus one unrelated docs change. The cluster is best treated as a thematic group, not a merge target set.", - "confidence": 0.81, + "summary": "These are mostly separate model-specific output-tracing refactors (plus one docs toctree tweak), not duplicates of the same underlying bug/change.", + "confidence": 0.98, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #44003 is the most representative core refactor: it spans two related model files and introduces the central output-capturing plumbing (`can_return_tuple`/`capture_outputs`).", + "canonical_pr_reason": null, "best_issue_reason": null, - "best_pr_reason": "PR #44003 is the strongest single representative of the cluster because it makes the broadest model-side tracing change and is closest to the shared refactor pattern seen across the group.", + "best_pr_reason": "PR 44003 is the broadest representative of the pattern, touching both `mamba` and `falcon_mamba`; however, it is still not a duplicate target for the other PRs.", "soft_edge_verdicts": [ { "left": "pull_request:43998", "right": "pull_request:44002", "accept": false, - "reason": "Different model families and different forward paths; same refactor theme only." + "reason": "Different model families/files; same refactor pattern, not the same concrete change." }, { "left": "pull_request:43998", "right": "pull_request:44000", "accept": false, - "reason": "Both are output-tracing refactors, but they affect unrelated models and code paths." + "reason": "Different model families/files; same refactor pattern, not the same concrete change." }, { "left": "pull_request:43998", "right": "pull_request:44001", "accept": false, - "reason": "Separate model implementation, not the same underlying bug or change." + "reason": "Different model families/files; same refactor pattern, not the same concrete change." }, { "left": "pull_request:43998", "right": "pull_request:44003", "accept": false, - "reason": "Different architectures and files; shared tracing pattern is too broad to merge as duplicates." + "reason": "Timm backbone vs Mamba/Falcon-Mamba code paths; only the tracing theme is shared." }, { "left": "pull_request:43998", "right": "pull_request:44029", "accept": false, - "reason": "Unrelated model code paths despite similar output-capturing refactor style." + "reason": "Unrelated architectures (`timm_backbone` vs `rwkv`); not the same bug or change." }, { "left": "pull_request:43999", "right": "pull_request:44001", "accept": false, - "reason": "MobileNetV1 and UnivNet are unrelated; this is only a similar refactor motif." + "reason": "MobileNetV1 vs UnivNet are unrelated model paths; similar refactor style only." }, { "left": "pull_request:43999", "right": "pull_request:44002", "accept": false, - "reason": "Different model subsystems and no shared concrete bug." + "reason": "MobileNetV1 vs UperNet are different code paths; not a duplicate change." }, { "left": "pull_request:43999", "right": "pull_request:44025", "accept": false, - "reason": "Different models; same broad tracing cleanup is not enough for duplication." + "reason": "MobileNetV1 vs Depth Anything; same tracing refactor theme, but different models." }, { "left": "pull_request:43999", "right": "pull_request:44026", "accept": false, - "reason": "Different encoder-decoder vs vision model code paths." + "reason": "MobileNetV1 vs VisionEncoderDecoder are unrelated implementations." }, { "left": "pull_request:43999", "right": "pull_request:44027", "accept": false, - "reason": "Unrelated model implementations; only the refactor pattern matches." + "reason": "MobileNetV1 vs SpeechEncoderDecoder are different model wrappers; not the same fix." }, { "left": "pull_request:44000", "right": "pull_request:44001", "accept": false, - "reason": "Different models and output types; not the same concrete change." + "reason": "Vision-Text dual encoder vs UnivNet are unrelated code paths." }, { "left": "pull_request:44000", "right": "pull_request:44002", "accept": false, - "reason": "Separate model families, no shared code path to merge." + "reason": "Vision-Text dual encoder vs UperNet are different model families." }, { "left": "pull_request:44000", "right": "pull_request:44003", "accept": false, - "reason": "Same general tracing theme, but different model architectures and files." + "reason": "Vision-Text dual encoder vs Mamba/Falcon-Mamba are unrelated internals." }, { "left": "pull_request:44000", "right": "pull_request:44025", "accept": false, - "reason": "Different models; similar refactor does not indicate duplication." + "reason": "Vision-Text dual encoder vs Depth Anything are separate model implementations." }, { "left": "pull_request:44001", "right": "pull_request:44002", "accept": false, - "reason": "Different model implementations with no single shared bug/change." + "reason": "UnivNet vs UperNet are different code paths; not a duplicate." }, { "left": "pull_request:44001", "right": "pull_request:44003", "accept": false, - "reason": "Unrelated model families; not mergeable as one PR." + "reason": "UnivNet vs Mamba/Falcon-Mamba are unrelated model internals." }, { "left": "pull_request:44001", "right": "pull_request:44025", "accept": false, - "reason": "Separate models and code paths, despite similar output-tracing work." + "reason": "UnivNet vs Depth Anything share only the tracing-refactor theme." }, { "left": "pull_request:44002", "right": "pull_request:44029", "accept": false, - "reason": "Different architectures; this is a broad subsystem similarity only." + "reason": "UperNet vs RWKV are unrelated model families." }, { "left": "pull_request:44002", "right": "pull_request:44003", "accept": false, - "reason": "No shared concrete change beyond the refactor style." + "reason": "UperNet vs Mamba/Falcon-Mamba are different code paths." }, { "left": "pull_request:44003", "right": "pull_request:44028", "accept": false, - "reason": "Mamba/FalconMamba vs SuperPoint are unrelated model paths." + "reason": "Mamba/Falcon-Mamba vs SuperPoint are unrelated model implementations." }, { "left": "pull_request:44003", "right": "pull_request:44004", "accept": false, - "reason": "Both touch output handling, but in unrelated models and call paths." + "reason": "Mamba/Falcon-Mamba refactor vs CodeGen refactor; same theme, different fix." }, { "left": "pull_request:43998", "right": "pull_request:44004", "accept": false, - "reason": "Different model code paths; same high-level refactor theme only." + "reason": "Timm backbone vs CodeGen are unrelated architectures." }, { "left": "pull_request:44001", "right": "pull_request:44004", "accept": false, - "reason": "Different models and signatures; not the same underlying fix." + "reason": "UnivNet vs CodeGen are different model paths." }, { "left": "pull_request:44002", "right": "pull_request:44004", "accept": false, - "reason": "Unrelated implementations despite similar output tracing changes." + "reason": "UperNet vs CodeGen are unrelated code paths." }, { "left": "pull_request:44000", "right": "pull_request:44004", "accept": false, - "reason": "Similar refactor pattern, but different model code and behavior." + "reason": "Vision-Text dual encoder vs CodeGen are different implementations." }, { "left": "pull_request:44025", "right": "pull_request:44026", "accept": false, - "reason": "Different model wrappers/code paths; not the same bug or change." + "reason": "Depth Anything vs VisionEncoderDecoder are separate model families." }, { "left": "pull_request:44025", "right": "pull_request:44027", "accept": false, - "reason": "Separate encoder-decoder models; shared tracing cleanup is too broad." + "reason": "Depth Anything vs SpeechEncoderDecoder are unrelated code paths." }, { "left": "pull_request:44025", "right": "pull_request:44028", "accept": false, - "reason": "Different model families and forward implementations." + "reason": "Depth Anything vs SuperPoint are unrelated model implementations." }, { "left": "pull_request:44025", "right": "pull_request:44029", "accept": false, - "reason": "Different model architectures; same refactor pattern is insufficient." + "reason": "Depth Anything vs RWKV are different architectures." }, { "left": "pull_request:44025", "right": "pull_request:44030", "accept": false, - "reason": "Unrelated models; not a single concrete issue." + "reason": "Depth Anything vs DPR are unrelated model paths." }, { "left": "pull_request:44026", - "right": "pull_request:44029", + "right": "pull_request:44027", "accept": false, - "reason": "Different model families and code paths." + "reason": "VisionEncoderDecoder vs SpeechEncoderDecoder are distinct wrappers; not the same change." }, { "left": "pull_request:44026", - "right": "pull_request:44027", + "right": "pull_request:44028", "accept": false, - "reason": "Vision encoder-decoder and speech encoder-decoder are distinct implementations." + "reason": "VisionEncoderDecoder vs SuperPoint are unrelated code paths." }, { "left": "pull_request:44026", - "right": "pull_request:44028", + "right": "pull_request:44029", "accept": false, - "reason": "Different model code paths; only the tracing refactor style overlaps." + "reason": "VisionEncoderDecoder vs RWKV are different model families." }, { "left": "pull_request:44026", "right": "pull_request:44030", "accept": false, - "reason": "Different models and outputs, no shared concrete bug." + "reason": "VisionEncoderDecoder vs DPR are unrelated implementations." }, { "left": "pull_request:44027", "right": "pull_request:44028", "accept": false, - "reason": "Separate model families; not mergeable as one PR." + "reason": "SpeechEncoderDecoder vs SuperPoint are different model paths." }, { "left": "pull_request:44027", "right": "pull_request:44029", "accept": false, - "reason": "Different architectures and forward paths." + "reason": "SpeechEncoderDecoder vs RWKV are unrelated architectures." }, { "left": "pull_request:44027", "right": "pull_request:44030", "accept": false, - "reason": "Unrelated model implementations despite similar cleanup." + "reason": "SpeechEncoderDecoder vs DPR are different code paths." }, { "left": "pull_request:44028", "right": "pull_request:44029", "accept": false, - "reason": "Different model families; same subsystem theme is too broad." + "reason": "SuperPoint vs RWKV are unrelated model implementations." }, { "left": "pull_request:44028", "right": "pull_request:44030", "accept": false, - "reason": "Different code paths and model types." + "reason": "SuperPoint vs DPR are different code paths." }, { "left": "pull_request:44029", "right": "pull_request:44030", "accept": false, - "reason": "RWKV and DPR are unrelated; no shared concrete fix." + "reason": "RWKV vs DPR are unrelated model families." }, { "left": "pull_request:44027", "right": "pull_request:46106", "accept": false, - "reason": "Docs toctree change is unrelated to the model tracing refactors." + "reason": "Model code refactor vs docs-only toctree edit; unrelated changes." }, { "left": "pull_request:44026", "right": "pull_request:46106", "accept": false, - "reason": "Documentation-only PR; not the same underlying change." + "reason": "Model code refactor vs docs-only toctree edit; unrelated changes." }, { "left": "pull_request:44001", "right": "pull_request:46106", "accept": false, - "reason": "Docs navigation change is unrelated to model code refactors." + "reason": "Model code refactor vs docs-only toctree edit; unrelated changes." }, { "left": "pull_request:44002", "right": "pull_request:46106", "accept": false, - "reason": "Unrelated documentation fix, not a duplicate of any model PR." + "reason": "Model code refactor vs docs-only toctree edit; unrelated changes." }, { "left": "pull_request:43998", "right": "pull_request:46106", "accept": false, - "reason": "Documentation toctree edit is not the same bug/change as model output tracing." + "reason": "Model code refactor vs docs-only toctree edit; unrelated changes." } ] }, @@ -19534,7 +19531,7 @@ "pull_request:44662|pull_request:45075", "pull_request:45075|pull_request:45401" ], - "prepared_review_unit_hash": "3fa1da49f4e766d83950d713db45ab8a390339e5b21807dbca9a419e3a90b3e0", + "prepared_review_unit_hash": "5abf70594dffdac0306fb70326d5db90eca828859196b9e7e2076574713ca9b8", "component_budget": { "node_count": 3, "item_count": 3, @@ -19556,30 +19553,30 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are three separate model-addition PRs for different models (PenguinVL, DeepSeek-OCR-2, Voxtral TTS). They share only Transformers boilerplate and docs/auto-registry changes, not the same underlying change.", - "confidence": 0.96, + "summary": "These are three separate model-introduction PRs (PenguinVL, DeepSeek-OCR-2, Voxtral TTS). They share generic HF wiring files, but they target different models, different code paths, and different upstream tasks, so they are not duplicates.", + "confidence": 0.98, "canonical_issue_reason": null, - "canonical_pr_reason": "If a single representative is needed, PR 45075 is the most mature/complete implementation in the set (conversion script, broader integration, highest review activity), but it is not a duplicate of the others.", + "canonical_pr_reason": null, "best_issue_reason": null, - "best_pr_reason": "PR 45075 is the strongest standalone representative because it has the most development/review activity and a fuller model integration path, but it is still a distinct model addition rather than the same fix as the others.", + "best_pr_reason": "PR 45075 is the strongest standalone representative: it is a full model addition with conversion code, tests, and broad integration, while the others are either a draft or an unrelated model implementation.", "soft_edge_verdicts": [ { "left": "pull_request:44662", "right": "pull_request:45401", "accept": false, - "reason": "Different models and different code paths; shared registry/docs boilerplate is not enough to call them duplicates." + "reason": "Different model families and different implementations (PenguinVL vs Voxtral TTS). The overlap is only shared registration/docs plumbing, not the same bug or change." }, { "left": "pull_request:44662", "right": "pull_request:45075", "accept": false, - "reason": "Separate model implementations with different architectures/assets; only generic Transformers scaffolding overlaps." + "reason": "Unrelated model additions for different architectures and tasks. Shared auto-mapping and docs files are expected for any new model, not evidence of duplication." }, { "left": "pull_request:45075", "right": "pull_request:45401", "accept": false, - "reason": "These add support for different models and are not plausibly mergeable into one PR fixing the same concrete bug or feature." + "reason": "Different concrete changes: DeepSeek-OCR-2 vision/OCR model vs Voxtral TTS speech model. They touch similar integration files but fix/add entirely separate code paths." } ] }, @@ -19609,12 +19606,12 @@ "soft_pairs": [ "pull_request:44979|pull_request:45363" ], - "prepared_review_unit_hash": "9931a2b89cc87b4d3978b14c61e38f97b8b5d7de109d3ea0b1d3d6fca38acf31", + "prepared_review_unit_hash": "0128bde623a28b227d67692204923896334f6472aa9796418a3bb986c043b3fd", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4407, + "serialized_chars": 4408, "estimated_input_tokens": 1102, "estimated_eval_tokens": 2460 }, @@ -19622,7 +19619,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4407, + "serialized_chars": 4408, "estimated_input_tokens": 1102, "estimated_eval_tokens": 2460 }, @@ -19631,18 +19628,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are related around module fusion, but they are not the same change: one adds a generic Module Fusion API, while the other implements a specific n-to-1 kernel-fusion path via KernelConfig and integration hooks.", - "confidence": 0.9, + "summary": "These are related kernel-fusion feature PRs, but they implement different layers of the stack: one introduces a generic module fusion API, while the other adds KernelConfig-driven n-to-1 fusion integration and an example. They are not duplicates.", + "confidence": 0.91, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #45363 is the more concrete and integrated feature implementation: it wires kernel fusion through KernelConfig and touching runtime integration points, rather than just adding a standalone abstraction layer.", + "canonical_pr_reason": "PR #44979 is the more foundational and self-contained feature: it adds a new `Module Fusion API` with its own implementation and tests, making it the best representative of the underlying cluster theme.", "best_issue_reason": null, - "best_pr_reason": "PR #45363 is the strongest representative of the cluster because it targets one specific fusion mechanism end-to-end and is closer to a shippable change.", + "best_pr_reason": "PR #45363 is the broader end-user-facing integration of kernel fusion via `KernelConfig`, touching runtime plumbing and an example, so it is the strongest single PR for the practical feature direction.", "soft_edge_verdicts": [ { "left": "pull_request:44979", "right": "pull_request:45363", "accept": false, - "reason": "They share the broad theme of fusion, but #44979 adds a generic module-fusion API while #45363 implements a specific kernel-based n-to-1 fusion path; they do not look like the same underlying bug or mergeable change." + "reason": "Related feature area, but not the same concrete change. #44979 adds a generic module-fusion API, while #45363 adds KernelConfig-based n-to-1 fusion integration and example usage; they would not plausibly merge as one duplicate PR." } ] }, @@ -19672,7 +19669,7 @@ "soft_pairs": [ "pull_request:45086|pull_request:45317" ], - "prepared_review_unit_hash": "1d8ffe5179239848731e67f7c92c76c8701bddf9e9a0b4d769fa899634b583ab", + "prepared_review_unit_hash": "2d800f4c74ef4dd36824e0f767f7478ee7697429eb17359f73984b8c3d0aa3ec", "component_budget": { "node_count": 2, "item_count": 2, @@ -19694,18 +19691,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both PRs fix the same AttributeError in `_patch_mistral_regex` by switching from `backend_tokenizer.pre_tokenizer` to `tokenizer.pre_tokenizer`; 45317 is the more complete version because it includes the regression test.", - "confidence": 0.97, + "summary": "Both PRs address the same Mistral regex patching AttributeError in the same function and file, with 45317 appearing to be a follow-up/expanded version of 45086 that adds tests.", + "confidence": 0.98, "canonical_issue_reason": null, - "canonical_pr_reason": "pull_request:45317 is the better canonical PR because it implements the same fix as 45086 and adds a test covering the regression.", + "canonical_pr_reason": "PR 45317 is the better canonical PR because it covers the same fix as 45086 but is more complete, including test changes and more discussion activity.", "best_issue_reason": null, - "best_pr_reason": "pull_request:45317 is the strongest overall PR candidate: same concrete bug fix plus test coverage, making it the most complete change.", + "best_pr_reason": "45317 is the strongest representative of the cluster: same concrete bug fix, same code path, and it adds validation via tests, making it the best stand-alone PR for this issue.", "soft_edge_verdicts": [ { "left": "pull_request:45086", "right": "pull_request:45317", "accept": true, - "reason": "Same underlying AttributeError fix in the same code path; 45317 is essentially 45086 plus a regression test." + "reason": "Same underlying AttributeError in _patch_mistral_regex, same file and issue target, and the second PR is essentially the same fix plus tests rather than a different change." } ] }, @@ -19735,12 +19732,12 @@ "soft_pairs": [ "pull_request:45115|pull_request:45569" ], - "prepared_review_unit_hash": "ddd7e412bef6c27b5213b18a69c99a830cce30662eedcdf1d0acdcd614e2dc41", + "prepared_review_unit_hash": "8fb0f926f81cb1f4135ca2b2bb29ce30f99835e864588285ea85bac1deb19ad0", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 6066, + "serialized_chars": 6065, "estimated_input_tokens": 1517, "estimated_eval_tokens": 3290 }, @@ -19748,7 +19745,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 6066, + "serialized_chars": 6065, "estimated_input_tokens": 1517, "estimated_eval_tokens": 3290 }, @@ -19757,18 +19754,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Two NemotronH-related PRs share some model files, but they target different changes: one refactors NemotronH to inherit GraniteMoeHybrid/conversion mapping, while the other splits NemotronH into dense and sparse model families with docs/tests updates. They are related, not duplicates.", - "confidence": 0.87, + "summary": "The two PRs are related to NemotronH, but they are not the same change: #45115 is a conversion/refactor toward GraniteMoeHybrid-compatible inheritance, while #45569 is a broader model split introducing NemotronH dense/sparse variants plus docs and auto-mapping updates. They should not be deduplicated.", + "confidence": 0.9, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 45569 is the more representative and broader NemotronH change set: it introduces the dense/sparse model split, updates public docs, auto mappings, and tests, so it better captures the main architectural direction of the cluster.", + "canonical_pr_reason": "No clear canonical duplicate PR: #45115 and #45569 address different code paths and scopes within NemotronH rather than the same concrete fix.", "best_issue_reason": null, - "best_pr_reason": "PR 45569 is the better single PR to track because it covers the end-user-facing API/documentation and the concrete model-family restructuring, whereas PR 45115 is a narrower refactor of conversion/mapping details.", + "best_pr_reason": "#45569 is the better representative PR for the cluster because it is broader and more complete, covering the dedicated NemotronH dense/sparse model split and associated integration changes.", "soft_edge_verdicts": [ { "left": "pull_request:45115", "right": "pull_request:45569", "accept": false, - "reason": "They overlap on NemotronH internals, but the concrete change sets differ: 45115 is a refactor around GraniteMoeHybrid-compatible inheritance and weight conversion, while 45569 creates distinct dense/sparse implementations plus docs/tests. They are not the same underlying change and would not plausibly be merged as one duplicate PR." + "reason": "Both touch NemotronH internals, but #45115 is a targeted conversion-mapping/inheritance refactor while #45569 restructures the model into dense/sparse variants with docs and auto-class updates. Different underlying changes, not mergeable as one duplicate PR." } ] }, @@ -19798,7 +19795,7 @@ "soft_pairs": [ "pull_request:45332|pull_request:45333" ], - "prepared_review_unit_hash": "2af8257da5116d522d71ee7459df9bba97ac9e2ea2c7dfdac55f480e4361ede8", + "prepared_review_unit_hash": "6289851ccc737cd6b8fc57e885c7770ad7fa9ef1213d05014240fde1315a0ece", "component_budget": { "node_count": 2, "item_count": 2, @@ -19820,18 +19817,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are two related but distinct feature PRs in the same heterogeneity effort: one adds per-layer config support, the other adds heterogeneous model/runtime support. They share files and abstractions, but they do not look like the same concrete change and should remain separate.", - "confidence": 0.78, + "summary": "The two PRs are related but not duplicates: one adds heterogeneous configuration plumbing, while the other builds broader heterogeneous model/runtime support on top of it. They touch overlapping files because the model-support PR depends on the config support, but they are distinct changes and not the same fix/feature.", + "confidence": 0.91, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #45332 is the broader umbrella change: it adds heterogeneous model support and touches runtime/cache/masking/modeling code in addition to the shared heterogeneity config pieces. It is the better representative of the overall feature cluster.", + "canonical_pr_reason": "PR #45332 is the better canonical representative because it is the broader end-to-end feature PR: it includes the runtime/model changes, cache/masking updates, and tests, while #45333 is narrower foundational config plumbing.", "best_issue_reason": null, - "best_pr_reason": "PR #45332 is the best single representative because it covers the larger end-to-end code-path change, while #45333 is narrower and mostly establishes the per-layer configuration foundation.", + "best_pr_reason": "PR #45332 is the strongest cluster anchor since it covers the most user-visible functionality and the wider code path impacted by heterogeneous support.", "soft_edge_verdicts": [ { "left": "pull_request:45332", "right": "pull_request:45333", "accept": false, - "reason": "Related feature work, but not the same underlying bug/change. #45333 introduces heterogeneous config plumbing; #45332 builds on that with model/cache/masking behavior changes. They could be merged independently and are not strict duplicates." + "reason": "Shared heterogeneity/config infrastructure, but different scope: #45333 adds per-layer config support, while #45332 adds heterogeneous modeling/runtime behavior. They would not plausibly be merged as one duplicate PR." } ] }, @@ -19861,7 +19858,7 @@ "soft_pairs": [ "pull_request:45415|pull_request:45425" ], - "prepared_review_unit_hash": "3d14b7f6dd99313aadafd57a4e42558739ad4d305c66c50cdd7a704166066d4d", + "prepared_review_unit_hash": "788b84e897a4369a96032119ae1332cdc2b3cecf44cb44f923d2c76879435e1f", "component_budget": { "node_count": 2, "item_count": 2, @@ -19883,18 +19880,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are both typing-related PRs in transformers, but they address different scopes: one is a broad type-checking sweep across many files, while the other is a focused typing-compatibility update around modeling_utils and related protocols. They are not duplicates.", - "confidence": 0.91, + "summary": "These are two separate typing-focused PRs. PR 45415 is a broad mechanical type-checking pass across many transformer modules, while PR 45425 adds specific typing protocols/helpers and related adjustments. They overlap in a few files but not in the same concrete change or bug fix.", + "confidence": 0.92, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 45425 is the more focused, self-contained typing change and better represents the cluster\u2019s narrow technical theme; PR 45415 is a much broader repository-wide type-checking sweep.", + "canonical_pr_reason": "PR 45425 is the better canonical representative because it targets a specific typing gap with concrete helper protocols and narrow follow-up changes, whereas PR 45415 is a large broad-sweep type-checking pass.", "best_issue_reason": null, - "best_pr_reason": "PR 45425 is the best representative because it targets a specific typing abstraction problem and is plausibly mergeable on its own, unlike the broad lint/type-check enablement in PR 45415.", + "best_pr_reason": "PR 45425 is the stronger standalone PR: it is more focused, more cohesive, and easier to merge as a single change set than the wide-ranging type-annotation sweep in PR 45415.", "soft_edge_verdicts": [ { "left": "pull_request:45415", "right": "pull_request:45425", "accept": false, - "reason": "Do not merge: 45415 is a broad type-checking pass over many unrelated files plus pyproject lint settings, while 45425 adds specific typing protocols/aliases for modeling_utils, peft, and flash-attention utilities. They do not fix the same concrete code-path problem and would not plausibly be one PR." + "reason": "Both are about typing, but they do not fix the same underlying problem. PR 45415 is a broad annotation/type-ignore cleanup across many files; PR 45425 adds specific typing protocols and utility fixes in a different code path. Shared files alone are not enough to treat them as duplicates." } ] }, @@ -19931,12 +19928,12 @@ "pull_request:46109|pull_request:46151", "pull_request:46135|pull_request:46151" ], - "prepared_review_unit_hash": "30d1ba34e06d24400be643a86d16d7e22339432acd8f5eada37110c6859d109a", + "prepared_review_unit_hash": "be8cbaedf4f262355bfc732659cb71f8c3abe5037793bf873db5f1e1a2fb230d", "component_budget": { "node_count": 5, "item_count": 5, "soft_pair_count": 5, - "serialized_chars": 12529, + "serialized_chars": 12530, "estimated_input_tokens": 3133, "estimated_eval_tokens": 6522 }, @@ -19944,7 +19941,7 @@ "node_count": 5, "item_count": 5, "soft_pair_count": 5, - "serialized_chars": 12529, + "serialized_chars": 12530, "estimated_input_tokens": 3133, "estimated_eval_tokens": 6522 }, @@ -19953,42 +19950,42 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These PRs are mostly separate device-compatibility or test-expectation changes across unrelated areas. They share a broad theme of XPU/model-parallel support, but the concrete code paths and affected tests/models differ enough that none should be merged as duplicates.", - "confidence": 0.79, + "summary": "These PRs are not duplicates of one another overall: they split into three different themes\u2014model device-placement fixes, distributed-test wording updates, and test expectation updates for unrelated suites. The only close resemblance is between the model-parallel PRs, but they still target different model code paths and should not be merged as one change.", + "confidence": 0.91, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 45487 is the strongest representative of the cluster: it fixes an actual runtime model-parallel bug in model code, spans two affected model implementations, and includes targeted tests. It is more concrete and substantive than the test-only expectation updates.", + "canonical_pr_reason": "PR 45487 is the broadest and most substantive fix: it addresses an actual model-parallel device placement bug across several model implementations, not just a test adjustment.", "best_issue_reason": null, - "best_pr_reason": "PR 45487 is the best merge candidate because it addresses a specific device-placement failure in production code rather than just adjusting test expectations, and the fix is narrowly scoped to the affected model-parallel paths.", + "best_pr_reason": "PR 45487 is the best representative of the cluster because it fixes a real runtime bug in core model code and has the clearest production impact.", "soft_edge_verdicts": [ { "left": "pull_request:45487", "right": "pull_request:46101", "accept": false, - "reason": "Both are device-placement fixes in model code, but they target different models and different failure points (AltCLIP/ChineseCLIP buffered token types vs Qwen2-Audio audio mask creation). Same broad class, not the same underlying bug." + "reason": "Both are device-placement fixes for model parallelism, but they touch different concrete model code paths (AltCLIP/ChineseCLIP vs Qwen2-Audio) and do not look like the same underlying bug." }, { "left": "pull_request:45487", "right": "pull_request:46151", "accept": false, - "reason": "45487 fixes model-parallel runtime behavior; 46151 only adds XPU expectations to a pipeline test. Different artifact type and no shared concrete code path." + "reason": "45487 fixes model code; 46151 only updates test expectations for pipeline outputs. Same broad theme of test/runtime compatibility, but not the same bug or change." }, { "left": "pull_request:46101", "right": "pull_request:46151", "accept": false, - "reason": "Qwen2-Audio runtime device handling and token-classification test expectations are unrelated changes; they do not fix the same bug." + "reason": "Qwen2-Audio runtime masking logic and token-classification test expectations are unrelated changes, despite both mentioning accelerator compatibility." }, { "left": "pull_request:46109", "right": "pull_request:46151", "accept": false, - "reason": "Both touch distributed/XPU testing, but one updates trainer distributed device assertions while the other adds pipeline token-classification expectations. These are distinct test suites and distinct behaviors." + "reason": "These are both test-only changes, but they affect different distributed-trainer vs pipeline test suites and different issues; not one underlying change." }, { "left": "pull_request:46135", "right": "pull_request:46151", "accept": false, - "reason": "Both are test expectation changes, but they concern different models and scenarios (Kosmos2.5 integration vs token-classification pipeline). No shared underlying bug/change." + "reason": "Both are test expectation updates, but for different model/pipeline areas with no shared code path or same concrete bug." } ] }, @@ -20020,12 +20017,12 @@ "pull_request:45513|pull_request:46084", "pull_request:46084|pull_request:46098" ], - "prepared_review_unit_hash": "5f02483eef44a25747a49b54508098f501547b85c940a46beba45825f84dcb54", + "prepared_review_unit_hash": "e5890af0825d84e02e814b57a6f4e042b526fed41454c837e9ee58357ab32d86", "component_budget": { "node_count": 3, "item_count": 3, "soft_pair_count": 2, - "serialized_chars": 7427, + "serialized_chars": 7426, "estimated_input_tokens": 1857, "estimated_eval_tokens": 3970 }, @@ -20033,7 +20030,7 @@ "node_count": 3, "item_count": 3, "soft_pair_count": 2, - "serialized_chars": 7427, + "serialized_chars": 7426, "estimated_input_tokens": 1857, "estimated_eval_tokens": 3970 }, @@ -20042,30 +20039,30 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Two PRs are effectively the same Mamba2 cache-seq_len>1 bugfix and should cluster together; the Qwen3.5 multi-token cached-forward PR is a separate model-family change and should remain separate.", + "summary": "Two PRs are about the same Mamba2 cache-length bug, while the Qwen3.5 PR is a separate model-specific fix. The Mamba2 docs+code PR is the best representative of the shared bug fix.", "confidence": 0.91, "canonical_issue_reason": null, - "canonical_pr_reason": "PR #46098 is the clearest representative: it directly names the bug, explicitly targets issue #46032, and appears to contain the full source/test/doc fix rather than just the code patch.", + "canonical_pr_reason": "pull_request:46098 is the broadest, most complete Mamba2 fix: it targets the same issue, covers the same code path, and includes tests/docs in addition to the implementation change.", "best_issue_reason": null, - "best_pr_reason": "PR #46098 is the strongest standalone duplicate target because it is explicit about the failure mode, includes the shared issue link, and covers the broadest complete fix set for the same Mamba2 code path.", + "best_pr_reason": "pull_request:46098 best represents the cluster because it addresses the shared Mamba2 use_cache seq_len>1 bug with the most complete scope and supporting updates.", "soft_edge_verdicts": [ { "left": "pull_request:45513", "right": "pull_request:46084", "accept": false, - "reason": "Different underlying models and code paths: Qwen3.5/OLMo hybrid linear-attention cached forward vs Mamba2 cached SSM/conv handling. They are only similar at a high level, not the same change." + "reason": "Different model families and different code paths: Qwen3.5 GDN linear attention vs Mamba2 cache handling. Similar symptom, but not the same underlying change." }, { "left": "pull_request:46084", "right": "pull_request:46098", "accept": true, - "reason": "Both address the same Mamba2 bug: `use_cache` with `seq_len > 1` producing wrong results. They touch the same files, share the same issue target (#46032), and the second PR looks like a broader version of the same fix." + "reason": "Same underlying Mamba2 use_cache seq_len>1 bug in the same module and same issue target; 46098 appears to be the broader, more complete version of the same fix." } ] }, "evaluator_result": { "accept": true, - "feedback": "The summary is grounded in the packet: 45513 is a different model family/code path from the Mamba2 fixes, while 46084 and 46098 clearly address the same Mamba2 `use_cache` / `seq_len > 1` bug, share issue #46032, and appear mergeable as one bugfix PR. The soft-edge verdicts are appropriately conservative overall." + "feedback": "The clustering is grounded in the packet: 46084 and 46098 share the same Mamba2 `use_cache`/`seq_len > 1` issue target and code path, while 45513 is a different model-family/code-path fix. The mergeability judgment for 46084 + 46098 is conservative and reasonable, and the best-PR choice for 46098 is supported by its broader scope (docs + tests + implementation)." }, "evaluator_used": true, "retried": false, @@ -20092,12 +20089,12 @@ "soft_pairs": [ "pull_request:45548|pull_request:45662" ], - "prepared_review_unit_hash": "f8ee4dc83b9de712bb2b3918a5598dcbe7b5055d81cbbd4a0a0f4ab933b23f4d", + "prepared_review_unit_hash": "dee3acf86fc055c7478981185eb4a3b15fbdcba1b4ad5c608aeb5b4f28ff672b", "component_budget": { "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4506, + "serialized_chars": 4508, "estimated_input_tokens": 1127, "estimated_eval_tokens": 2510 }, @@ -20105,7 +20102,7 @@ "node_count": 2, "item_count": 2, "soft_pair_count": 1, - "serialized_chars": 4506, + "serialized_chars": 4508, "estimated_input_tokens": 1127, "estimated_eval_tokens": 2510 }, @@ -20114,18 +20111,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are related EP-loading fixes, but they address different failure modes: one is about EP + ZeRO-3 loading via accelerate launch, the other about experts being overwritten by rank-0 broadcast under FSDP2. They should not be merged as duplicates.", - "confidence": 0.96, + "summary": "These are related only at a high level (expert-parallel loading), but they fix different bugs in different code paths: one is config/ZeRO-3 launch handling, the other is shard-preservation during EP/FSDP2 parameter loading.", + "confidence": 0.91, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 45662 is the stronger representative of the EP-loading bug class here because it fixes a concrete overwrite path in model loading and includes broader code-path changes; PR 45548 is a narrower, distinct accelerate/ZeRO-3 serialization/loading fix.", + "canonical_pr_reason": "PR 45662 is the stronger representative because it fixes the actual parameter-assignment path and spans the EP/FSDP2 loading logic where the corruption occurs; PR 45548 is narrower and mostly about config serialization/feature detection for a different loading scenario.", "best_issue_reason": null, - "best_pr_reason": "PR 45662 is the better overall candidate to represent the cluster because it targets the more specific underlying runtime bug with explicit code-path changes in loading/sharding behavior, whereas PR 45548 is a separate, narrower loading compatibility fix.", + "best_pr_reason": "PR 45662 is the best overall fit for the cluster because it addresses a concrete runtime failure mode in loading/sharding behavior, with broader and more central code-path changes than 45548.", "soft_edge_verdicts": [ { "left": "pull_request:45548", "right": "pull_request:45662", "accept": false, - "reason": "Both concern expert parallelism, but the bugs differ: 45548 is about EP + DeepSpeed ZeRO-3 loading via accelerate launch, while 45662 fixes experts being silently overwritten by rank-0 broadcast under FSDP2. Shared subsystem alone is not enough to treat them as the same change." + "reason": "Both concern expert-parallel loading, but 45548 targets distributed_config serialization and an EP capability flag for DeepSpeed ZeRO-3/accelerate launch, while 45662 fixes parameter wrapping during model loading to stop expert weights being overwritten under FSDP2. Different underlying bugs and not plausibly one merged PR." } ] }, @@ -20155,7 +20152,7 @@ "soft_pairs": [ "pull_request:45639|pull_request:45645" ], - "prepared_review_unit_hash": "dfc7c620270fe27897db0b4230518e48549a9e08f889555d1b2c5eb118c71e4e", + "prepared_review_unit_hash": "1071af12868f40796d4fb774c7ec29d14db6adc3ed0714e94b323ad2b20b112b", "component_budget": { "node_count": 2, "item_count": 2, @@ -20177,18 +20174,18 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "Both PRs address the same xdist-related captured_info logging collision in testing utilities; PR 45645 is the broader, more complete fix and subsumes the core change from 45639.", - "confidence": 0.96, + "summary": "Both PRs target the same xdist-safe captured_info logging problem in testing_utils; 45645 is the broader, more complete fix and appears to subsume 45639.", + "confidence": 0.93, "canonical_issue_reason": null, - "canonical_pr_reason": "PR 45645 is the stronger canonical PR because it contains the shared testing_utils fix plus the CI/workflow adjustments needed to preserve and aggregate the new per-worker logs.", + "canonical_pr_reason": "PR 45645 is the stronger canonical choice because it contains the core testing_utils fix plus the CI/workflow handling needed to preserve and aggregate multiple captured_info files, making it the more complete implementation.", "best_issue_reason": null, - "best_pr_reason": "PR 45645 is the best representative of the cluster since it covers the same underlying bug and includes the additional plumbing to make the fix usable in CI.", + "best_pr_reason": "PR 45645 is the best overall PR because it addresses the same underlying bug with the xdist-safe output path change and also updates downstream consumers, so it is more likely to fully resolve the issue.", "soft_edge_verdicts": [ { "left": "pull_request:45639", "right": "pull_request:45645", "accept": true, - "reason": "Accept: both PRs fix the same concrete xdist collision in captured_info handling via the same testing_utils path; 45645 extends that same fix with workflow support rather than changing to a different bug or subsystem." + "reason": "Same underlying captured_info xdist collision fix. 45639 is a narrower subset of the broader 45645 change set, and the overlapping testing_utils/test adjustments make them plausibly mergeable into one PR." } ] }, @@ -20287,7 +20284,7 @@ "pull_request:46069|pull_request:46154", "pull_request:46073|pull_request:46154" ], - "prepared_review_unit_hash": "209a611aa9e9a42d81a8da3a8ef6da1cf6b7781ede3c1c1268b512c6cc0ab9c5", + "prepared_review_unit_hash": "a49e24056b0490895a8cdd3da402498368a40575ef0c42bbe31b2e91ce74d5d6", "component_budget": { "node_count": 3, "item_count": 3, @@ -20309,30 +20306,30 @@ "aggressively_trimmed": false, "split": false, "analyst_result": { - "summary": "These are three separate model-specific PRs adding attention dispatch support, not duplicates of the same underlying fix. They share a theme (`attn_implementation`/SDPA support) but target different architectures and code paths.", + "summary": "These PRs are all model-specific additions for `attn_implementation`/SDPA dispatch, but they target different architectures and code paths. They are related in theme, not duplicates.", "confidence": 0.95, "canonical_issue_reason": null, - "canonical_pr_reason": "No true canonical duplicate exists in this cluster. If one PR must be chosen as the representative, 46154 is the most complete model-specific dispatch change, with the broadest test coverage and the clearest SDPA implementation work.", + "canonical_pr_reason": "PR 46154 is the strongest representative: it has the clearest feature scope (`attn_implementation=sdpa` dispatch), touches model code plus multiple test files, and looks like the most complete implementation among the three.", "best_issue_reason": null, - "best_pr_reason": "46154 is the best representative PR for the cluster because it is the most substantive and self-contained of the three, but it should still not be treated as a duplicate of the others.", + "best_pr_reason": "PR 46154 is the best single PR to represent this cluster because it is the most fully fleshed-out SDPA dispatch support change and has broader validation coverage.", "soft_edge_verdicts": [ { "left": "pull_request:46069", "right": "pull_request:46073", "accept": false, - "reason": "Both add attention-dispatch support, but for different models (ShieldGemma2 vs CTRL) and different code paths. Same subsystem, not the same fix." + "reason": "Both add SDPA/attention dispatch support, but for different models (ShieldGemma2 vs CTRL) with different files and code paths, so they are not the same underlying change." }, { "left": "pull_request:46069", "right": "pull_request:46154", "accept": false, - "reason": "Model-specific PRs for different architectures (ShieldGemma2 vs T5). They are analogous changes, not one mergeable duplicate change." + "reason": "Same broad feature area, but these are separate model implementations (ShieldGemma2 vs T5) and not a single mergeable code-path fix." }, { "left": "pull_request:46073", "right": "pull_request:46154", "accept": false, - "reason": "Both touch SDPA dispatch, but each PR targets a different model implementation and test set. Not the same underlying bug or change." + "reason": "Both are model-specific SDPA dispatch PRs, but they target different architectures and cannot plausibly be merged into one duplicate fix." } ] }, diff --git a/analysis/current/manifest.json b/analysis/current/manifest.json index fbffe1d3ec575478117b35171d7d5261a5df8d6d..451e7c1817191c8a67052444138ef96fb5ffe525 100644 --- a/analysis/current/manifest.json +++ b/analysis/current/manifest.json @@ -1,8 +1,8 @@ { - "analysis_id": "hybrid-model-20260523t180021z", + "analysis_id": "hybrid-model-20260524t000134z", "archived_artifacts": { - "hybrid": "snapshots/20260523T180021Z/analysis-runs/hybrid-model-20260523t180021z/analysis-report-hybrid.json", - "hybrid_reviews": "snapshots/20260523T180021Z/analysis-runs/hybrid-model-20260523t180021z/analysis-report-hybrid.llm-reviews.json" + "hybrid": "snapshots/20260524T000134Z/analysis-runs/hybrid-model-20260524t000134z/analysis-report-hybrid.json", + "hybrid_reviews": "snapshots/20260524T000134Z/analysis-runs/hybrid-model-20260524t000134z/analysis-report-hybrid.llm-reviews.json" }, "artifacts": { "hybrid": "analysis/current/analysis-report-hybrid.json", @@ -10,9 +10,9 @@ }, "channel": "canonical", "model": null, - "published_at": "2026-05-23T18:07:28Z", + "published_at": "2026-05-24T00:08:56Z", "repo": "huggingface/transformers", "schema_version": 1, - "snapshot_id": "20260523T180021Z", + "snapshot_id": "20260524T000134Z", "variant": "hybrid" } diff --git a/snapshots/20260524T000134Z/analysis-runs/hybrid-model-20260524t000134z/analysis-report-hybrid.json b/snapshots/20260524T000134Z/analysis-runs/hybrid-model-20260524t000134z/analysis-report-hybrid.json new file mode 100644 index 0000000000000000000000000000000000000000..7b11d5e1bde566bd596460f8552e0f9b0b1a16e7 --- /dev/null +++ b/snapshots/20260524T000134Z/analysis-runs/hybrid-model-20260524t000134z/analysis-report-hybrid.json @@ -0,0 +1,2156 @@ +{ + "schema_version": "1.0", + "repo": "huggingface/transformers", + "snapshot_id": "20260524T000134Z", + "generated_at": "2026-05-24T00:08:54Z", + "evidence_quality": "full", + "llm_enrichment": true, + "meta_bugs": [ + { + "cluster_id": "cluster-43979-11", + "summary": "Cluster of 1 issues and 10 PRs centered on issue #43979.", + "status": "open", + "confidence": 0.8, + "canonical_issue_number": 43979, + "canonical_pr_number": 44007, + "issue_numbers": [ + 43979 + ], + "pr_numbers": [ + 43996, + 44007, + 44013, + 44044, + 44066, + 44072, + 44085, + 44129, + 44154, + 44722 + ], + "evidence_types": [ + "closing_reference", + "shared_issue_target", + "soft_similarity" + ], + "pr_comparisons": [ + { + "left_pr_number": 43996, + "right_pr_number": 44007, + "code_similarity": 0.179, + "size_similarity": 0.576, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.429, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 43996, + "right_pr_number": 44013, + "code_similarity": 0.122, + "size_similarity": 0.318, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.392, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 43996, + "right_pr_number": 44044, + "code_similarity": 0.245, + "size_similarity": 0.864, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.479, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 43996, + "right_pr_number": 44066, + "code_similarity": 0.225, + "size_similarity": 0.818, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.408, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 43996, + "right_pr_number": 44072, + "code_similarity": 0.14, + "size_similarity": 0.303, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.528, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 43996, + "right_pr_number": 44085, + "code_similarity": 0.216, + "size_similarity": 0.783, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.398, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 43996, + "right_pr_number": 44129, + "code_similarity": 0.163, + "size_similarity": 0.643, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.229, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 43996, + "right_pr_number": 44154, + "code_similarity": 0.153, + "size_similarity": 0.535, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.31, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 43996, + "right_pr_number": 44722, + "code_similarity": 0.225, + "size_similarity": 0.848, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.368, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44007, + "right_pr_number": 44013, + "code_similarity": 0.19, + "size_similarity": 0.553, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.531, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44007, + "right_pr_number": 44044, + "code_similarity": 0.186, + "size_similarity": 0.667, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.354, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44007, + "right_pr_number": 44066, + "code_similarity": 0.188, + "size_similarity": 0.704, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.315, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44007, + "right_pr_number": 44072, + "code_similarity": 0.212, + "size_similarity": 0.526, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.708, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44007, + "right_pr_number": 44085, + "code_similarity": 0.195, + "size_similarity": 0.735, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.318, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44007, + "right_pr_number": 44129, + "code_similarity": 0.103, + "size_similarity": 0.37, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.191, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44007, + "right_pr_number": 44154, + "code_similarity": 0.238, + "size_similarity": 0.93, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.344, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44007, + "right_pr_number": 44722, + "code_similarity": 0.178, + "size_similarity": 0.679, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.28, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44013, + "right_pr_number": 44044, + "code_similarity": 0.126, + "size_similarity": 0.368, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.351, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44013, + "right_pr_number": 44066, + "code_similarity": 0.127, + "size_similarity": 0.389, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.325, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44013, + "right_pr_number": 44072, + "code_similarity": 0.29, + "size_similarity": 0.952, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.667, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44013, + "right_pr_number": 44085, + "code_similarity": 0.131, + "size_similarity": 0.406, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.329, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44013, + "right_pr_number": 44129, + "code_similarity": 0.07, + "size_similarity": 0.205, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.192, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44013, + "right_pr_number": 44154, + "code_similarity": 0.177, + "size_similarity": 0.594, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.389, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44013, + "right_pr_number": 44722, + "code_similarity": 0.118, + "size_similarity": 0.375, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.287, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44044, + "right_pr_number": 44066, + "code_similarity": 0.25, + "size_similarity": 0.947, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.404, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44044, + "right_pr_number": 44072, + "code_similarity": 0.136, + "size_similarity": 0.351, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.442, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44044, + "right_pr_number": 44085, + "code_similarity": 0.24, + "size_similarity": 0.906, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.394, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44044, + "right_pr_number": 44129, + "code_similarity": 0.147, + "size_similarity": 0.555, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.243, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44044, + "right_pr_number": 44154, + "code_similarity": 0.17, + "size_similarity": 0.62, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.306, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44044, + "right_pr_number": 44722, + "code_similarity": 0.257, + "size_similarity": 0.982, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.402, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44066, + "right_pr_number": 44072, + "code_similarity": 0.133, + "size_similarity": 0.37, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.393, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44066, + "right_pr_number": 44085, + "code_similarity": 0.763, + "size_similarity": 0.957, + "file_overlap": 0.5, + "area_overlap": 0.825, + "patch_similarity": 0.887, + "shared_filenames": [ + "src/transformers/models/gptj/modeling_gptj.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/models/gptj/modeling_gptj.py", + "left_ranges": [ + [ + 33, + 41 + ], + [ + 174, + 181 + ], + [ + 250, + 257 + ], + [ + 398, + 419 + ], + [ + 425, + 434 + ], + [ + 459, + 466 + ], + [ + 471, + 488 + ], + [ + 518, + 540 + ], + [ + 554, + 560 + ], + [ + 566, + 575 + ], + [ + 580, + 598 + ], + [ + 601, + 606 + ], + [ + 634, + 640 + ], + [ + 646, + 653 + ], + [ + 658, + 674 + ], + [ + 717, + 722 + ], + [ + 738, + 744 + ], + [ + 749, + 772 + ], + [ + 790, + 795 + ] + ], + "right_ranges": [ + [ + 33, + 40 + ], + [ + 173, + 180 + ], + [ + 249, + 256 + ], + [ + 397, + 405 + ], + [ + 408, + 419 + ], + [ + 425, + 434 + ], + [ + 459, + 465 + ], + [ + 470, + 489 + ], + [ + 517, + 539 + ], + [ + 553, + 559 + ], + [ + 565, + 574 + ], + [ + 579, + 597 + ], + [ + 600, + 611 + ], + [ + 633, + 639 + ], + [ + 645, + 652 + ], + [ + 657, + 673 + ], + [ + 716, + 728 + ], + [ + 737, + 743 + ], + [ + 748, + 771 + ], + [ + 789, + 794 + ] + ] + } + ] + }, + { + "left_pr_number": 44066, + "right_pr_number": 44129, + "code_similarity": 0.145, + "size_similarity": 0.526, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.263, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44066, + "right_pr_number": 44154, + "code_similarity": 0.174, + "size_similarity": 0.654, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.286, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44066, + "right_pr_number": 44722, + "code_similarity": 0.907, + "size_similarity": 0.964, + "file_overlap": 1.0, + "area_overlap": 0.808, + "patch_similarity": 0.874, + "shared_filenames": [ + "src/transformers/models/codegen/modeling_codegen.py", + "src/transformers/models/gptj/modeling_gptj.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/models/codegen/modeling_codegen.py", + "left_ranges": [ + [ + 245, + 266 + ] + ], + "right_ranges": [ + [ + 228, + 234 + ] + ] + }, + { + "filename": "src/transformers/models/gptj/modeling_gptj.py", + "left_ranges": [ + [ + 33, + 41 + ], + [ + 174, + 181 + ], + [ + 250, + 257 + ], + [ + 398, + 419 + ], + [ + 425, + 434 + ], + [ + 459, + 466 + ], + [ + 471, + 488 + ], + [ + 518, + 540 + ], + [ + 554, + 560 + ], + [ + 566, + 575 + ], + [ + 580, + 598 + ], + [ + 601, + 606 + ], + [ + 634, + 640 + ], + [ + 646, + 653 + ], + [ + 658, + 674 + ], + [ + 717, + 722 + ], + [ + 738, + 744 + ], + [ + 749, + 772 + ], + [ + 790, + 795 + ] + ], + "right_ranges": [ + [ + 33, + 42 + ], + [ + 175, + 182 + ], + [ + 251, + 258 + ], + [ + 399, + 420 + ], + [ + 426, + 435 + ], + [ + 460, + 467 + ], + [ + 472, + 489 + ], + [ + 519, + 540 + ], + [ + 554, + 560 + ], + [ + 566, + 575 + ], + [ + 580, + 598 + ], + [ + 601, + 612 + ], + [ + 634, + 640 + ], + [ + 646, + 653 + ], + [ + 658, + 663 + ], + [ + 666, + 674 + ], + [ + 717, + 722 + ], + [ + 738, + 744 + ], + [ + 749, + 772 + ], + [ + 790, + 795 + ] + ] + } + ] + }, + { + "left_pr_number": 44072, + "right_pr_number": 44085, + "code_similarity": 0.137, + "size_similarity": 0.387, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.398, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44072, + "right_pr_number": 44129, + "code_similarity": 0.074, + "size_similarity": 0.195, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.231, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44072, + "right_pr_number": 44154, + "code_similarity": 0.175, + "size_similarity": 0.566, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.414, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44072, + "right_pr_number": 44722, + "code_similarity": 0.124, + "size_similarity": 0.357, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.347, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44085, + "right_pr_number": 44129, + "code_similarity": 0.141, + "size_similarity": 0.503, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.272, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44085, + "right_pr_number": 44154, + "code_similarity": 0.18, + "size_similarity": 0.684, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.289, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44085, + "right_pr_number": 44722, + "code_similarity": 0.728, + "size_similarity": 0.923, + "file_overlap": 0.5, + "area_overlap": 0.791, + "patch_similarity": 0.78, + "shared_filenames": [ + "src/transformers/models/gptj/modeling_gptj.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/models/gptj/modeling_gptj.py", + "left_ranges": [ + [ + 33, + 40 + ], + [ + 173, + 180 + ], + [ + 249, + 256 + ], + [ + 397, + 405 + ], + [ + 408, + 419 + ], + [ + 425, + 434 + ], + [ + 459, + 465 + ], + [ + 470, + 489 + ], + [ + 517, + 539 + ], + [ + 553, + 559 + ], + [ + 565, + 574 + ], + [ + 579, + 597 + ], + [ + 600, + 611 + ], + [ + 633, + 639 + ], + [ + 645, + 652 + ], + [ + 657, + 673 + ], + [ + 716, + 728 + ], + [ + 737, + 743 + ], + [ + 748, + 771 + ], + [ + 789, + 794 + ] + ], + "right_ranges": [ + [ + 33, + 42 + ], + [ + 175, + 182 + ], + [ + 251, + 258 + ], + [ + 399, + 420 + ], + [ + 426, + 435 + ], + [ + 460, + 467 + ], + [ + 472, + 489 + ], + [ + 519, + 540 + ], + [ + 554, + 560 + ], + [ + 566, + 575 + ], + [ + 580, + 598 + ], + [ + 601, + 612 + ], + [ + 634, + 640 + ], + [ + 646, + 653 + ], + [ + 658, + 663 + ], + [ + 666, + 674 + ], + [ + 717, + 722 + ], + [ + 738, + 744 + ], + [ + 749, + 772 + ], + [ + 790, + 795 + ] + ] + } + ] + }, + { + "left_pr_number": 44129, + "right_pr_number": 44154, + "code_similarity": 0.099, + "size_similarity": 0.344, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.199, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44129, + "right_pr_number": 44722, + "code_similarity": 0.146, + "size_similarity": 0.545, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.247, + "shared_filenames": [], + "shared_file_areas": [] + }, + { + "left_pr_number": 44154, + "right_pr_number": 44722, + "code_similarity": 0.164, + "size_similarity": 0.631, + "file_overlap": 0.0, + "area_overlap": 0.0, + "patch_similarity": 0.255, + "shared_filenames": [], + "shared_file_areas": [] + } + ] + }, + { + "cluster_id": "cluster-41211-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #41211.", + "status": "open", + "confidence": 0.55, + "canonical_issue_number": 41211, + "canonical_pr_number": 44339, + "issue_numbers": [ + 41211 + ], + "pr_numbers": [ + 41356, + 44339 + ], + "evidence_types": [ + "closing_reference" + ], + "pr_comparisons": [ + { + "left_pr_number": 41356, + "right_pr_number": 44339, + "code_similarity": 0.155, + "size_similarity": 0.094, + "file_overlap": 0.273, + "area_overlap": 0.078, + "patch_similarity": 0.183, + "shared_filenames": [ + "docs/source/en/model_doc/deimv2.md", + "src/transformers/models/deimv2/__init__.py", + "src/transformers/models/deimv2/configuration_deimv2.py", + "src/transformers/models/deimv2/modeling_deimv2.py", + "tests/models/deimv2/__init__.py", + "tests/models/deimv2/test_modeling_deimv2.py" + ], + "shared_file_areas": [ + { + "filename": "docs/source/en/model_doc/deimv2.md", + "left_ranges": [ + [ + 1, + 132 + ] + ], + "right_ranges": [ + [ + 1, + 65 + ] + ] + }, + { + "filename": "src/transformers/models/deimv2/__init__.py", + "left_ranges": [ + [ + 1, + 15 + ] + ], + "right_ranges": [ + [ + 1, + 29 + ] + ] + }, + { + "filename": "src/transformers/models/deimv2/configuration_deimv2.py", + "left_ranges": [ + [ + 1, + 74 + ] + ], + "right_ranges": [ + [ + 1, + 266 + ] + ] + }, + { + "filename": "tests/models/deimv2/test_modeling_deimv2.py", + "left_ranges": [ + [ + 1, + 15 + ] + ], + "right_ranges": [ + [ + 1, + 1734 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-43240-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #43240.", + "status": "open", + "confidence": 0.75, + "canonical_issue_number": 43240, + "canonical_pr_number": 43251, + "issue_numbers": [ + 43240 + ], + "pr_numbers": [ + 43251, + 43254 + ], + "evidence_types": [ + "closing_reference", + "shared_issue_target" + ], + "pr_comparisons": [ + { + "left_pr_number": 43251, + "right_pr_number": 43254, + "code_similarity": 0.794, + "size_similarity": 0.64, + "file_overlap": 1.0, + "area_overlap": 0.667, + "patch_similarity": 0.882, + "shared_filenames": [ + "src/transformers/loss/loss_utils.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/loss/loss_utils.py", + "left_ranges": [ + [ + 30, + 50 + ] + ], + "right_ranges": [ + [ + 30, + 43 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-46032-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #46032.", + "status": "open", + "confidence": 0.75, + "canonical_issue_number": 46032, + "canonical_pr_number": 46084, + "issue_numbers": [ + 46032 + ], + "pr_numbers": [ + 46084, + 46098 + ], + "evidence_types": [ + "closing_reference", + "shared_issue_target" + ], + "pr_comparisons": [ + { + "left_pr_number": 46084, + "right_pr_number": 46098, + "code_similarity": 0.43, + "size_similarity": 0.5, + "file_overlap": 0.667, + "area_overlap": 0.226, + "patch_similarity": 0.339, + "shared_filenames": [ + "src/transformers/models/mamba2/modeling_mamba2.py", + "tests/models/mamba2/test_modeling_mamba2.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/models/mamba2/modeling_mamba2.py", + "left_ranges": [ + [ + 256, + 262 + ], + [ + 303, + 353 + ], + [ + 461, + 467 + ], + [ + 471, + 491 + ], + [ + 496, + 502 + ], + [ + 599, + 609 + ] + ], + "right_ranges": [ + [ + 255, + 373 + ], + [ + 481, + 495 + ] + ] + }, + { + "filename": "tests/models/mamba2/test_modeling_mamba2.py", + "left_ranges": [ + [ + 206, + 236 + ], + [ + 285, + 322 + ] + ], + "right_ranges": [ + [ + 476, + 542 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-41115-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #41115.", + "status": "open", + "confidence": 0.55, + "canonical_issue_number": 41115, + "canonical_pr_number": 45613, + "issue_numbers": [ + 41115 + ], + "pr_numbers": [ + 41116, + 45613 + ], + "evidence_types": [ + "closing_reference" + ], + "pr_comparisons": [ + { + "left_pr_number": 41116, + "right_pr_number": 45613, + "code_similarity": 0.805, + "size_similarity": 0.885, + "file_overlap": 0.75, + "area_overlap": 0.854, + "patch_similarity": 0.693, + "shared_filenames": [ + "docs/source/en/_toctree.yml", + "docs/source/en/model_doc/minicpm3.md", + "src/transformers/models/auto/modeling_auto.py", + "src/transformers/models/minicpm3/__init__.py", + "src/transformers/models/minicpm3/configuration_minicpm3.py", + "src/transformers/models/minicpm3/modeling_minicpm3.py", + "src/transformers/models/minicpm3/modular_minicpm3.py", + "tests/models/minicpm3/__init__.py", + "tests/models/minicpm3/test_modeling_minicpm3.py" + ], + "shared_file_areas": [ + { + "filename": "docs/source/en/_toctree.yml", + "left_ranges": [ + [ + 713, + 720 + ] + ], + "right_ranges": [ + [ + 713, + 720 + ] + ] + }, + { + "filename": "docs/source/en/model_doc/minicpm3.md", + "left_ranges": [ + [ + 1, + 67 + ] + ], + "right_ranges": [ + [ + 1, + 45 + ] + ] + }, + { + "filename": "src/transformers/models/auto/modeling_auto.py", + "left_ranges": [ + [ + 285, + 291 + ], + [ + 700, + 706 + ], + [ + 1298, + 1304 + ] + ], + "right_ranges": [ + [ + 699, + 705 + ], + [ + 1301, + 1307 + ] + ] + }, + { + "filename": "src/transformers/models/minicpm3/__init__.py", + "left_ranges": [ + [ + 1, + 29 + ] + ], + "right_ranges": [ + [ + 1, + 29 + ] + ] + }, + { + "filename": "src/transformers/models/minicpm3/configuration_minicpm3.py", + "left_ranges": [ + [ + 1, + 141 + ] + ], + "right_ranges": [ + [ + 1, + 126 + ] + ] + }, + { + "filename": "src/transformers/models/minicpm3/modeling_minicpm3.py", + "left_ranges": [ + [ + 1, + 544 + ] + ], + "right_ranges": [ + [ + 1, + 522 + ] + ] + }, + { + "filename": "src/transformers/models/minicpm3/modular_minicpm3.py", + "left_ranges": [ + [ + 1, + 444 + ] + ], + "right_ranges": [ + [ + 1, + 342 + ] + ] + }, + { + "filename": "tests/models/minicpm3/test_modeling_minicpm3.py", + "left_ranges": [ + [ + 1, + 109 + ] + ], + "right_ranges": [ + [ + 1, + 136 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-45647-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #45647.", + "status": "open", + "confidence": 0.75, + "canonical_issue_number": 45647, + "canonical_pr_number": 45738, + "issue_numbers": [ + 45647 + ], + "pr_numbers": [ + 45737, + 45738 + ], + "evidence_types": [ + "closing_reference", + "shared_issue_target" + ], + "pr_comparisons": [ + { + "left_pr_number": 45737, + "right_pr_number": 45738, + "code_similarity": 0.526, + "size_similarity": 0.11, + "file_overlap": 0.5, + "area_overlap": 1.0, + "patch_similarity": 0.024, + "shared_filenames": [ + "src/transformers/models/musicgen_melody/modeling_musicgen_melody.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/models/musicgen_melody/modeling_musicgen_melody.py", + "left_ranges": [ + [ + 491, + 497 + ] + ], + "right_ranges": [ + [ + 491, + 497 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-45081-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #45081.", + "status": "open", + "confidence": 0.8, + "canonical_issue_number": 45081, + "canonical_pr_number": 45317, + "issue_numbers": [ + 45081 + ], + "pr_numbers": [ + 45086, + 45317 + ], + "evidence_types": [ + "closing_reference", + "shared_issue_target", + "soft_similarity" + ], + "pr_comparisons": [ + { + "left_pr_number": 45086, + "right_pr_number": 45317, + "code_similarity": 0.589, + "size_similarity": 0.34, + "file_overlap": 0.5, + "area_overlap": 1.0, + "patch_similarity": 0.136, + "shared_filenames": [ + "src/transformers/tokenization_utils_tokenizers.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/tokenization_utils_tokenizers.py", + "left_ranges": [ + [ + 1360, + 1370 + ], + [ + 1374, + 1380 + ] + ], + "right_ranges": [ + [ + 1360, + 1370 + ], + [ + 1374, + 1380 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-45561-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #45561.", + "status": "open", + "confidence": 0.8, + "canonical_issue_number": 45561, + "canonical_pr_number": 45645, + "issue_numbers": [ + 45561 + ], + "pr_numbers": [ + 45639, + 45645 + ], + "evidence_types": [ + "closing_reference", + "shared_issue_target", + "soft_similarity" + ], + "pr_comparisons": [ + { + "left_pr_number": 45639, + "right_pr_number": 45645, + "code_similarity": 0.532, + "size_similarity": 0.619, + "file_overlap": 0.5, + "area_overlap": 0.543, + "patch_similarity": 0.451, + "shared_filenames": [ + "src/transformers/testing_utils.py", + "tests/utils/test_testing_utils.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/testing_utils.py", + "left_ranges": [ + [ + 3529, + 3536 + ], + [ + 3753, + 3787 + ] + ], + "right_ranges": [ + [ + 3525, + 3558 + ], + [ + 3782, + 3788 + ] + ] + }, + { + "filename": "tests/utils/test_testing_utils.py", + "left_ranges": [ + [ + 1, + 86 + ] + ], + "right_ranges": [ + [ + 1, + 114 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-43698-3", + "summary": "Cluster of 1 issues and 2 PRs centered on issue #43698.", + "status": "open", + "confidence": 0.75, + "canonical_issue_number": 43698, + "canonical_pr_number": 43779, + "issue_numbers": [ + 43698 + ], + "pr_numbers": [ + 43779, + 43816 + ], + "evidence_types": [ + "closing_reference", + "shared_issue_target" + ], + "pr_comparisons": [ + { + "left_pr_number": 43779, + "right_pr_number": 43816, + "code_similarity": 0.418, + "size_similarity": 0.538, + "file_overlap": 1.0, + "area_overlap": 0.02, + "patch_similarity": 0.02, + "shared_filenames": [ + "src/transformers/integrations/integration_utils.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/integrations/integration_utils.py", + "left_ranges": [ + [ + 2227, + 2241 + ], + [ + 2303, + 2309 + ] + ], + "right_ranges": [ + [ + 2278, + 2291 + ], + [ + 2309, + 2322 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-46100-2", + "summary": "Cluster of 2 related pull requests linked by soft_similarity.", + "status": "open", + "confidence": 0.5, + "canonical_issue_number": null, + "canonical_pr_number": 46100, + "issue_numbers": [], + "pr_numbers": [ + 46100, + 46114 + ], + "evidence_types": [ + "soft_similarity" + ], + "pr_comparisons": [ + { + "left_pr_number": 46100, + "right_pr_number": 46114, + "code_similarity": 1.0, + "size_similarity": 1.0, + "file_overlap": 1.0, + "area_overlap": 1.0, + "patch_similarity": 1.0, + "shared_filenames": [ + "tests/models/vit/test_modeling_vit.py" + ], + "shared_file_areas": [ + { + "filename": "tests/models/vit/test_modeling_vit.py", + "left_ranges": [ + [ + 216, + 226 + ] + ], + "right_ranges": [ + [ + 216, + 226 + ] + ] + } + ] + } + ] + }, + { + "cluster_id": "cluster-44018-2", + "summary": "Cluster of 2 related pull requests linked by soft_similarity.", + "status": "open", + "confidence": 0.5, + "canonical_issue_number": null, + "canonical_pr_number": 44068, + "issue_numbers": [], + "pr_numbers": [ + 44018, + 44068 + ], + "evidence_types": [ + "soft_similarity" + ], + "pr_comparisons": [ + { + "left_pr_number": 44018, + "right_pr_number": 44068, + "code_similarity": 0.766, + "size_similarity": 0.939, + "file_overlap": 1.0, + "area_overlap": 0.425, + "patch_similarity": 0.866, + "shared_filenames": [ + "src/transformers/models/gpt_neo/modeling_gpt_neo.py", + "tests/models/gpt_neo/test_modeling_gpt_neo.py" + ], + "shared_file_areas": [ + { + "filename": "src/transformers/models/gpt_neo/modeling_gpt_neo.py", + "left_ranges": [ + [ + 26, + 46 + ], + [ + 138, + 143 + ], + [ + 182, + 187 + ], + [ + 283, + 295 + ], + [ + 327, + 341 + ], + [ + 348, + 354 + ], + [ + 360, + 369 + ], + [ + 399, + 406 + ], + [ + 411, + 419 + ], + [ + 428, + 433 + ], + [ + 472, + 492 + ], + [ + 507, + 513 + ], + [ + 519, + 528 + ], + [ + 541, + 559 + ], + [ + 562, + 567 + ], + [ + 595, + 601 + ], + [ + 607, + 614 + ], + [ + 627, + 643 + ], + [ + 685, + 690 + ], + [ + 708, + 714 + ], + [ + 720, + 727 + ], + [ + 740, + 757 + ], + [ + 761, + 766 + ], + [ + 780, + 786 + ], + [ + 791, + 798 + ], + [ + 807, + 822 + ], + [ + 840, + 845 + ] + ], + "right_ranges": [ + [ + 26, + 31 + ], + [ + 34, + 44 + ], + [ + 136, + 141 + ], + [ + 180, + 185 + ], + [ + 281, + 293 + ], + [ + 325, + 339 + ], + [ + 346, + 352 + ], + [ + 358, + 367 + ], + [ + 397, + 404 + ], + [ + 409, + 417 + ], + [ + 426, + 434 + ], + [ + 463, + 483 + ], + [ + 498, + 504 + ], + [ + 510, + 519 + ], + [ + 532, + 550 + ], + [ + 553, + 564 + ], + [ + 586, + 592 + ], + [ + 598, + 605 + ], + [ + 618, + 634 + ], + [ + 676, + 688 + ], + [ + 699, + 705 + ], + [ + 711, + 718 + ], + [ + 731, + 748 + ], + [ + 752, + 762 + ], + [ + 771, + 777 + ], + [ + 782, + 789 + ], + [ + 798, + 813 + ], + [ + 831, + 836 + ] + ] + }, + { + "filename": "tests/models/gpt_neo/test_modeling_gpt_neo.py", + "left_ranges": [ + [ + 458, + 464 + ] + ], + "right_ranges": [ + [ + 458, + 464 + ] + ] + } + ] + } + ] + } + ], + "duplicate_issues": [], + "duplicate_prs": [ + { + "cluster_id": "cluster-41115-3", + "canonical_pr_number": 45613, + "duplicate_pr_numbers": [ + 41116 + ], + "target_issue_number": 41115, + "reason": "PRs in cluster-41115-3 are treated as duplicates because they converge on issue #41115 with closing_reference evidence." + }, + { + "cluster_id": "cluster-41211-3", + "canonical_pr_number": 44339, + "duplicate_pr_numbers": [ + 41356 + ], + "target_issue_number": 41211, + "reason": "PRs in cluster-41211-3 are treated as duplicates because they converge on issue #41211 with closing_reference evidence." + }, + { + "cluster_id": "cluster-43240-3", + "canonical_pr_number": 43251, + "duplicate_pr_numbers": [ + 43254 + ], + "target_issue_number": 43240, + "reason": "PRs in cluster-43240-3 are treated as duplicates because they converge on issue #43240 with closing_reference, shared_issue_target evidence." + }, + { + "cluster_id": "cluster-43698-3", + "canonical_pr_number": 43779, + "duplicate_pr_numbers": [ + 43816 + ], + "target_issue_number": 43698, + "reason": "PRs in cluster-43698-3 are treated as duplicates because they converge on issue #43698 with closing_reference, shared_issue_target evidence." + }, + { + "cluster_id": "cluster-43979-11", + "canonical_pr_number": 44007, + "duplicate_pr_numbers": [ + 43996, + 44013, + 44044, + 44066, + 44072, + 44085, + 44129, + 44154, + 44722 + ], + "target_issue_number": 43979, + "reason": "PRs in cluster-43979-11 are treated as duplicates because they converge on issue #43979 with closing_reference, shared_issue_target, soft_similarity evidence." + }, + { + "cluster_id": "cluster-44018-2", + "canonical_pr_number": 44068, + "duplicate_pr_numbers": [ + 44018 + ], + "target_issue_number": null, + "reason": "PRs in cluster-44018-2 are treated as duplicates because they share soft_similarity evidence." + }, + { + "cluster_id": "cluster-45081-3", + "canonical_pr_number": 45317, + "duplicate_pr_numbers": [ + 45086 + ], + "target_issue_number": 45081, + "reason": "PRs in cluster-45081-3 are treated as duplicates because they converge on issue #45081 with closing_reference, shared_issue_target, soft_similarity evidence." + }, + { + "cluster_id": "cluster-45561-3", + "canonical_pr_number": 45645, + "duplicate_pr_numbers": [ + 45639 + ], + "target_issue_number": 45561, + "reason": "PRs in cluster-45561-3 are treated as duplicates because they converge on issue #45561 with closing_reference, shared_issue_target, soft_similarity evidence." + }, + { + "cluster_id": "cluster-45647-3", + "canonical_pr_number": 45738, + "duplicate_pr_numbers": [ + 45737 + ], + "target_issue_number": 45647, + "reason": "PRs in cluster-45647-3 are treated as duplicates because they converge on issue #45647 with closing_reference, shared_issue_target evidence." + }, + { + "cluster_id": "cluster-46032-3", + "canonical_pr_number": 46084, + "duplicate_pr_numbers": [ + 46098 + ], + "target_issue_number": 46032, + "reason": "PRs in cluster-46032-3 are treated as duplicates because they converge on issue #46032 with closing_reference, shared_issue_target evidence." + }, + { + "cluster_id": "cluster-46100-2", + "canonical_pr_number": 46100, + "duplicate_pr_numbers": [ + 46114 + ], + "target_issue_number": null, + "reason": "PRs in cluster-46100-2 are treated as duplicates because they share soft_similarity evidence." + } + ], + "best_issue": { + "cluster_id": "cluster-45647-3", + "issue_number": 45647, + "reason": "Issue #45647 is the strongest global issue candidate because it is open, belongs to a cluster with 3 artifacts, and carries 12 discussion comments plus 8 inbound references.", + "score": 162.0 + }, + "best_pr": { + "cluster_id": "cluster-43979-11", + "pr_number": 44007, + "reason": "PR #44007 is the strongest global PR candidate because it is open, belongs to a cluster with 11 artifacts, links to 1 issues, and carries 2 review/discussion events.", + "score": 238.886 + } +} diff --git a/snapshots/20260524T000134Z/analysis-runs/hybrid-model-20260524t000134z/analysis-report-hybrid.llm-reviews.json b/snapshots/20260524T000134Z/analysis-runs/hybrid-model-20260524t000134z/analysis-report-hybrid.llm-reviews.json new file mode 100644 index 0000000000000000000000000000000000000000..c3c683f067f876f556e4de43aa4e093c27a022e5 --- /dev/null +++ b/snapshots/20260524T000134Z/analysis-runs/hybrid-model-20260524t000134z/analysis-report-hybrid.llm-reviews.json @@ -0,0 +1,20344 @@ +{ + "schema_version": "1.0", + "repo": "huggingface/transformers", + "snapshot_id": "20260524T000134Z", + "generated_at": "2026-05-24T00:08:54Z", + "model": "gpt-5.4-mini?service_tier=flex", + "reviews": [ + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 1, + "review_unit_count": 49, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-426", + "nodes": [ + "issue:24643", + "issue:30277", + "issue:34634", + "issue:34928", + "issue:35545", + "issue:36010", + "issue:36331", + "issue:39290", + "issue:41720", + "issue:41762", + "issue:42915", + "issue:43039", + "issue:43425", + "issue:43541", + "issue:43572", + "issue:43596", + "issue:43638", + "issue:43716", + "issue:43827", + "issue:43828", + "issue:43866", + "issue:43901", + "issue:43975", + "issue:44292", + "issue:44322", + "issue:44457", + "issue:44509", + "issue:44512", + "issue:44560", + "issue:44805", + "issue:44841", + "issue:44863", + "issue:44918", + "issue:45081", + "issue:45161", + "issue:45237", + "issue:45464", + "issue:45507", + "issue:45685", + "issue:45812", + "issue:45823", + "issue:45859", + "issue:45874", + "issue:45901", + "issue:45902", + "issue:45920", + "issue:46015", + "issue:46165" + ], + "soft_pairs": [ + "issue:44322|issue:45464", + "issue:45237|issue:45685", + "issue:43716|issue:45237", + "issue:43828|issue:45237", + "issue:45237|issue:45507", + "issue:44292|issue:45237", + "issue:30277|issue:35545", + "issue:35545|issue:42915", + "issue:35545|issue:41720", + "issue:45901|issue:46015", + "issue:44509|issue:44512", + "issue:35545|issue:36010", + "issue:44805|issue:45823", + "issue:44918|issue:46165", + "issue:43638|issue:44805", + "issue:43638|issue:46165", + "issue:43596|issue:44805", + "issue:44805|issue:45161", + "issue:44560|issue:46165", + "issue:43596|issue:46165", + "issue:44805|issue:44918", + "issue:43572|issue:46165", + "issue:34634|issue:35545", + "issue:30277|issue:43638", + "issue:43596|issue:43638", + "issue:43039|issue:45161", + "issue:43975|issue:44457", + "issue:44292|issue:44457", + "issue:34928|issue:44805", + "issue:24643|issue:30277", + "issue:45859|issue:45902", + "issue:43638|issue:45161", + "issue:34928|issue:36331", + "issue:43596|issue:45161", + "issue:43716|issue:45507", + "issue:39290|issue:45081", + "issue:45823|issue:45874", + "issue:43425|issue:44292", + "issue:34928|issue:41720", + "issue:43828|issue:45161", + "issue:43716|issue:45685", + "issue:44918|issue:45161", + "issue:43828|issue:44292", + "issue:43541|issue:43716", + "issue:44292|issue:44918", + "issue:43716|issue:44292", + "issue:45507|issue:45685", + "issue:43039|issue:44918", + "issue:41762|issue:43638", + "issue:43827|issue:44512", + "issue:43716|issue:44560", + "issue:43828|issue:45685", + "issue:44918|issue:45685", + "issue:43039|issue:43828", + "issue:30277|issue:36331", + "issue:45237|issue:46015", + "issue:34928|issue:36010", + "issue:43039|issue:44805", + "issue:41762|issue:43596", + "issue:39290|issue:41720", + "issue:43901|issue:44512", + "issue:45237|issue:45901", + "issue:44841|issue:45823", + "issue:43039|issue:44560", + "issue:43975|issue:46015", + "issue:45812|issue:45920", + "issue:44863|issue:46015", + "issue:43866|issue:44863", + "issue:44292|issue:46015" + ], + "prepared_review_unit_hash": "ff874a85474cdc886c0f715b028ee4b8a25dcae056743f81e04870d26bc5b7d1", + "component_budget": { + "node_count": 426, + "item_count": 426, + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 69, + "serialized_chars": 48406, + "estimated_input_tokens": 12102, + "estimated_eval_tokens": 24460 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is mostly a mix of unrelated issue reports, with only a couple of true duplicate-style overlaps. The clearest duplicates are the DeepSpeed ZeRO-3 BERT index error pair and the AutoTokenizer wrong-token-ID regression pair.", + "confidence": 0.86, + "canonical_issue_reason": "issue:45920 is the broadest and most representative report of the AutoTokenizer v5 regression, covering multiple affected model families rather than a single model family.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45920 best captures the underlying tokenizer regression and is open, broader, and more suitable as the main tracking issue.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44322", + "right": "issue:45464", + "accept": false, + "reason": "Different failures: missing config attribute vs streaming chat/completions API behavior." + }, + { + "left": "issue:45237", + "right": "issue:45685", + "accept": false, + "reason": "Different models and backends; one is AMD GPU support for GPT-OSS, the other is an MPS histogram kernel issue in MoE code." + }, + { + "left": "issue:43716", + "right": "issue:45237", + "accept": false, + "reason": "Different code paths: image preprocessor dtype mismatch vs GPT-OSS running on AMD GPUs." + }, + { + "left": "issue:43828", + "right": "issue:45237", + "accept": false, + "reason": "Both are model/backend errors, but the concrete bugs are different dtype/autocast handling vs AMD GPU execution." + }, + { + "left": "issue:45237", + "right": "issue:45507", + "accept": false, + "reason": "Different underlying bug: AMD GPU runtime failure vs invalid method call in GraniteMoEHybrid." + }, + { + "left": "issue:44292", + "right": "issue:45237", + "accept": false, + "reason": "Different model/runtime failures; NVFP4 execution error is not the same as AMD GPU incompatibility." + }, + { + "left": "issue:30277", + "right": "issue:35545", + "accept": false, + "reason": "DeepSpeed collective mismatch and ONNX export failure are unrelated bugs." + }, + { + "left": "issue:35545", + "right": "issue:42915", + "accept": false, + "reason": "Different model/export path and error type; ONNX export is unrelated to FineGrainedFP8 config failure." + }, + { + "left": "issue:35545", + "right": "issue:41720", + "accept": false, + "reason": "Different subsystems and symptoms: export error vs device-mapping cuda assert." + }, + { + "left": "issue:45901", + "right": "issue:46015", + "accept": false, + "reason": "One is a task crash, the other is stale docs; same broad area but not the same bug." + }, + { + "left": "issue:44509", + "right": "issue:44512", + "accept": false, + "reason": "Both are v5 docs cleanups, but they concern different removed doc references and should stay separate." + }, + { + "left": "issue:35545", + "right": "issue:36010", + "accept": false, + "reason": "Different problems: ONNX export versus missing GenerationMixin import." + }, + { + "left": "issue:44805", + "right": "issue:45823", + "accept": false, + "reason": "Different failures: mask/tensor shape mismatch vs Gemma4 device placement mismatch." + }, + { + "left": "issue:44918", + "right": "issue:46165", + "accept": false, + "reason": "Different training failures with different triggers; not the same code-path bug." + }, + { + "left": "issue:43638", + "right": "issue:44805", + "accept": false, + "reason": "Both are IndexError reports, but one is DeepSpeed ZeRO-3 BERT init and the other is a mask shape mismatch." + }, + { + "left": "issue:43596", + "right": "issue:44805", + "accept": false, + "reason": "Same superficial error class, but different underlying failure modes and code paths." + }, + { + "left": "issue:44805", + "right": "issue:45161", + "accept": false, + "reason": "Mask-shape mismatch is unrelated to GPT-OSS tensor-parallel not working." + }, + { + "left": "issue:44560", + "right": "issue:46165", + "accept": false, + "reason": "Different model/runtime issue from a training resume/trust-remote-code problem." + }, + { + "left": "issue:43596", + "right": "issue:46165", + "accept": false, + "reason": "DeepSpeed ZeRO-3 BERT index error is unrelated to resume-training with remote code." + }, + { + "left": "issue:44805", + "right": "issue:44918", + "accept": false, + "reason": "Different symptoms and paths: masking error vs unpacking Qwen3.5 embeddings in TRL SFT." + }, + { + "left": "issue:43572", + "right": "issue:46165", + "accept": false, + "reason": "Config missing pad_token_idx is unrelated to training resume behavior." + }, + { + "left": "issue:34634", + "right": "issue:35545", + "accept": false, + "reason": "Processor voice preset bug is unrelated to ModernBERT ONNX export." + }, + { + "left": "issue:30277", + "right": "issue:43638", + "accept": false, + "reason": "Both involve DeepSpeed, but one is collective mismatch and the other is BERT zero-sized index on load." + }, + { + "left": "issue:43596", + "right": "issue:43638", + "accept": true, + "reason": "Same DeepSpeed ZeRO-3 BERT index-out-of-bounds failure; the titles describe the same underlying loading/training bug." + }, + { + "left": "issue:43039", + "right": "issue:45161", + "accept": false, + "reason": "Different issues in different parts of the stack; Liger cross-entropy call is not the same as GPT-OSS TP failure." + }, + { + "left": "issue:43975", + "right": "issue:44457", + "accept": false, + "reason": "Different bugs: detokenization regression vs LoRA merge/save/reload mismatch." + }, + { + "left": "issue:44292", + "right": "issue:44457", + "accept": false, + "reason": "Different failures and models; NVFP4 runtime error is not a LoRA merge regression." + }, + { + "left": "issue:34928", + "right": "issue:44805", + "accept": false, + "reason": "Activation checkpointing/FSDP recompute mismatch is unrelated to the mask-shape IndexError." + }, + { + "left": "issue:24643", + "right": "issue:30277", + "accept": false, + "reason": "Both mention DeepSpeed, but the concrete failures are different runtime errors." + }, + { + "left": "issue:45859", + "right": "issue:45902", + "accept": false, + "reason": "Same Qwen3.5 MoE embedding area, but the triggers differ (CPU offload vs meta-to-empty materialization), so this is not clearly the same bug." + }, + { + "left": "issue:43638", + "right": "issue:45161", + "accept": false, + "reason": "BERT ZeRO-3 loading error and GPT-OSS tensor-parallel failure are unrelated." + }, + { + "left": "issue:34928", + "right": "issue:36331", + "accept": false, + "reason": "Activation checkpointing/FSDP mismatch is unrelated to CustomTrainer compute_loss signature breakage." + }, + { + "left": "issue:43039", + "right": "issue:43828", + "accept": false, + "reason": "Different model/runtime bugs: Liger Kernel cross-entropy invocation vs autocast dtype mismatch." + }, + { + "left": "issue:30277", + "right": "issue:36331", + "accept": false, + "reason": "DeepSpeed collective mismatch and trainer compute_loss signature mismatch are unrelated." + }, + { + "left": "issue:45237", + "right": "issue:46015", + "accept": false, + "reason": "Model runtime failure vs docs cleanup; not the same bug or change." + }, + { + "left": "issue:34928", + "right": "issue:36010", + "accept": false, + "reason": "FSDP activation-checkpointing error is unrelated to GenerationMixin import failure." + }, + { + "left": "issue:43039", + "right": "issue:44805", + "accept": false, + "reason": "Different underlying failures: Liger kernel path vs tensor mask shape mismatch." + }, + { + "left": "issue:41762", + "right": "issue:43596", + "accept": false, + "reason": "Both are DeepSpeed-related, but Gemma3 ZeRO-3 loading and BERT ZeRO-3 index errors are different bugs." + }, + { + "left": "issue:39290", + "right": "issue:41720", + "accept": false, + "reason": "Different model/backend issues: Gemma3/vLLM attribute error vs Qwen3 auto device map cuda assert." + }, + { + "left": "issue:43901", + "right": "issue:44512", + "accept": false, + "reason": "Different docs issues referencing different removed API surface." + }, + { + "left": "issue:45237", + "right": "issue:45901", + "accept": false, + "reason": "GPT-OSS AMD runtime failure is unrelated to table-question-answering crashes." + }, + { + "left": "issue:44841", + "right": "issue:45823", + "accept": false, + "reason": "Different processor/model failures with no shared concrete bug." + }, + { + "left": "issue:43039", + "right": "issue:44560", + "accept": false, + "reason": "Liger Kernel cross-entropy and Qwen3-vl video stop-iteration are unrelated." + }, + { + "left": "issue:43975", + "right": "issue:46015", + "accept": false, + "reason": "Tokenizer detokenization regression is unrelated to outdated RAG examples." + }, + { + "left": "issue:45812", + "right": "issue:45920", + "accept": true, + "reason": "Same AutoTokenizer wrong-token-ID regression in v5; 45920 simply broadens the affected model list beyond Granite." + }, + { + "left": "issue:44863", + "right": "issue:46015", + "accept": false, + "reason": "NemotronH checkpoint loading failure is unrelated to docs/example updates." + }, + { + "left": "issue:43866", + "right": "issue:44863", + "accept": false, + "reason": "Ovis2 checkpoint corruption and NemotronH checkpoint loading are different problems." + }, + { + "left": "issue:44292", + "right": "issue:46015", + "accept": false, + "reason": "NVFP4 runtime error is unrelated to outdated RAG docs." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet and stays conservative. The two accepted overlaps are well-supported by the titles and descriptions: the DeepSpeed ZeRO-3 BERT index errors (43596/43638) and the AutoTokenizer wrong-token-ID regression (45812/45920). The rest of the pairwise rejections are cautious and consistent with the packet." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 2, + "review_unit_count": 49, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-426", + "nodes": [ + "issue:36010", + "issue:39186", + "issue:39290", + "issue:40990", + "issue:41093", + "issue:41762", + "issue:43054", + "issue:43296", + "issue:43366", + "issue:43404", + "issue:43425", + "issue:43531", + "issue:43541", + "issue:43572", + "issue:43596", + "issue:43638", + "issue:43645", + "issue:43716", + "issue:43792", + "issue:43828", + "issue:43854", + "issue:43866", + "issue:43975", + "issue:43994", + "issue:44291", + "issue:44292", + "issue:44387", + "issue:44589", + "issue:44841", + "issue:44918", + "issue:44960", + "issue:45005", + "issue:45020", + "issue:45070", + "issue:45084", + "issue:45161", + "issue:45362", + "issue:45399", + "issue:45464", + "issue:45542", + "issue:45632", + "issue:45685", + "issue:45698", + "issue:45759", + "issue:45800", + "issue:45823", + "issue:45901", + "issue:45902" + ], + "soft_pairs": [ + "issue:39290|issue:43531", + "issue:43854|issue:43866", + "issue:43792|issue:45901", + "issue:41093|issue:43638", + "issue:43425|issue:45070", + "issue:43596|issue:45800", + "issue:43975|issue:45901", + "issue:43792|issue:45070", + "issue:45362|issue:45464", + "issue:44292|issue:45901", + "issue:39290|issue:45823", + "issue:45464|issue:45685", + "issue:43404|issue:43866", + "issue:43541|issue:45685", + "issue:43541|issue:43828", + "issue:44841|issue:45084", + "issue:36010|issue:39290", + "issue:45800|issue:45823", + "issue:41093|issue:41762", + "issue:43572|issue:45070", + "issue:43296|issue:43366", + "issue:43541|issue:45161", + "issue:44387|issue:45005", + "issue:43828|issue:45464", + "issue:44291|issue:45070", + "issue:43366|issue:45464", + "issue:45020|issue:45800", + "issue:43531|issue:45070", + "issue:43541|issue:44589", + "issue:41093|issue:43596", + "issue:43054|issue:43994", + "issue:45542|issue:45800", + "issue:45759|issue:45902", + "issue:39186|issue:44918", + "issue:43716|issue:44918", + "issue:40990|issue:44960", + "issue:45632|issue:45698", + "issue:43645|issue:45399", + "issue:43404|issue:45005" + ], + "prepared_review_unit_hash": "0c9974d3f32ac16fb6d7cd08ef94fa643f3dc5220aded1163b773100684cc3ae", + "component_budget": { + "node_count": 426, + "item_count": 426, + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 39, + "serialized_chars": 39128, + "estimated_input_tokens": 9782, + "estimated_eval_tokens": 19820 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a grab bag of distinct Transformers regressions and model-specific failures rather than true duplicates. The best umbrella issue is the broad remote_code regression report, but I did not find any soft pair that looks like the same concrete bug.", + "confidence": 0.63, + "canonical_issue_reason": "Issue 45020 is the broadest and most central report here: it describes recent-version breakage for models using `remote_code`, which is the closest umbrella for several compatibility-style regressions in the set.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 45020 is the best global representative because it captures a class of recent Transformers compatibility breakages affecting custom/model-loading paths, not just one model instance.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:39290", + "right": "issue:43531", + "accept": false, + "reason": "Both mention `sliding_window`, but they are different models with different failure modes; not the same concrete bug." + }, + { + "left": "issue:43854", + "right": "issue:43866", + "accept": false, + "reason": "Unrelated model-loading problem vs checkpoint corruption; no shared bug path." + }, + { + "left": "issue:43792", + "right": "issue:45901", + "accept": false, + "reason": "Different tasks and model families; both are crashes, but not the same underlying issue." + }, + { + "left": "issue:41093", + "right": "issue:43638", + "accept": false, + "reason": "Both are index-related errors under deepspeed, but one is a mask shape mismatch and the other is a zero3/Bert init failure." + }, + { + "left": "issue:43425", + "right": "issue:45070", + "accept": false, + "reason": "Generic Torch version incompatibility is not the same as the specific pydantic `PretrainedConfig` regression." + }, + { + "left": "issue:43596", + "right": "issue:45800", + "accept": false, + "reason": "Different root causes: zero3/Bert loading index error vs a torch/transformers version mismatch." + }, + { + "left": "issue:43975", + "right": "issue:45901", + "accept": false, + "reason": "Detokenization bug and table QA crash are unrelated." + }, + { + "left": "issue:43792", + "right": "issue:45070", + "accept": false, + "reason": "Completely different issue classes: Whisper runtime failure vs config/model-schema regression." + }, + { + "left": "issue:45362", + "right": "issue:45464", + "accept": false, + "reason": "Same model family, but one is a chat crash and the other is a streaming chat/completions failure; too little evidence of one code-path bug." + }, + { + "left": "issue:44292", + "right": "issue:45901", + "accept": false, + "reason": "`init_empty_weights` argument handling is unrelated to table-question-answering crashes." + }, + { + "left": "issue:39290", + "right": "issue:45823", + "accept": false, + "reason": "Different models and different runtime errors; no shared bug." + }, + { + "left": "issue:45464", + "right": "issue:45685", + "accept": false, + "reason": "Streaming chat API failure is unrelated to an MPS histogram-on-Int error." + }, + { + "left": "issue:43404", + "right": "issue:43866", + "accept": false, + "reason": "Untied weights in Mistral3 is unrelated to a corrupted Ovis2 checkpoint." + }, + { + "left": "issue:43541", + "right": "issue:45685", + "accept": false, + "reason": "Torch dynamo grouped_mm failure and MPS histogram failure are unrelated." + }, + { + "left": "issue:43541", + "right": "issue:43828", + "accept": false, + "reason": "Both are dtype-related in MoE contexts, but the model paths and failure mechanisms differ." + }, + { + "left": "issue:44841", + "right": "issue:45084", + "accept": false, + "reason": "Processor load failure and template-node compilation failure are unrelated." + }, + { + "left": "issue:36010", + "right": "issue:39290", + "accept": false, + "reason": "ImportError for `GenerationMixin` and a Gemma3 sliding-window attribute error are unrelated." + }, + { + "left": "issue:45800", + "right": "issue:45823", + "accept": false, + "reason": "Torch/Transformers version incompatibility is not the same bug as Gemma4 device-map mismatch." + }, + { + "left": "issue:41093", + "right": "issue:41762", + "accept": false, + "reason": "Both involve zero3 index errors, but the specific shapes and loading paths differ enough to treat them as separate bugs." + }, + { + "left": "issue:43572", + "right": "issue:45070", + "accept": false, + "reason": "Both are config-related v5 regressions, but they concern different fields and different breakage modes." + }, + { + "left": "issue:43296", + "right": "issue:43366", + "accept": false, + "reason": "vLLM loading failure and GGUF architecture support are different problems." + }, + { + "left": "issue:43541", + "right": "issue:45161", + "accept": false, + "reason": "Both are MoE-related, but one is grouped_mm tracing and the other is tensor-parallel-only GPT-OSS behavior." + }, + { + "left": "issue:44387", + "right": "issue:45005", + "accept": false, + "reason": "Quantization OOM and tied-weights regressions are different issues." + }, + { + "left": "issue:43828", + "right": "issue:45464", + "accept": false, + "reason": "Autocast dtype mismatch is unrelated to a streaming chat API failure." + }, + { + "left": "issue:44291", + "right": "issue:45070", + "accept": false, + "reason": "Unexpected `_is_hf_initialized` argument during init is a different path than the pydantic config-field regression." + }, + { + "left": "issue:43366", + "right": "issue:45464", + "accept": false, + "reason": "GGUF gpt-oss support and Qwen streaming inference are unrelated." + }, + { + "left": "issue:45020", + "right": "issue:45800", + "accept": false, + "reason": "`remote_code` breakage is broader, but this pair does not point to the same specific bug as the torch/Transformers version incompatibility report." + }, + { + "left": "issue:43531", + "right": "issue:45070", + "accept": false, + "reason": "Sliding-window handling and pydantic config regression are distinct." + }, + { + "left": "issue:43541", + "right": "issue:44589", + "accept": false, + "reason": "Grouped_mm tracing failure and missing Float8 storage object are unrelated." + }, + { + "left": "issue:41093", + "right": "issue:43596", + "accept": false, + "reason": "Both are index errors with zero3, but they occur in different contexts and likely different code paths." + }, + { + "left": "issue:43054", + "right": "issue:43994", + "accept": false, + "reason": "Same model family, but one is about poorer embeddings and the other about nonsensical pipeline output; not enough to merge as one bug." + }, + { + "left": "issue:45542", + "right": "issue:45800", + "accept": false, + "reason": "TensorBoard/TensorFlow backend setup issue is unrelated to torch version incompatibility." + }, + { + "left": "issue:45759", + "right": "issue:45902", + "accept": false, + "reason": "Config unwrapping for composite Qwen models and uninitialized rotary-embedding memory are different bugs." + }, + { + "left": "issue:39186", + "right": "issue:44918", + "accept": false, + "reason": "FSDP 2-D weight error and Qwen embedding unpacking in TRL are unrelated." + }, + { + "left": "issue:43716", + "right": "issue:44918", + "accept": false, + "reason": "Mistral3 image-preprocessor dtype mismatch and Qwen3.5 embedding handling are different failures." + }, + { + "left": "issue:40990", + "right": "issue:44960", + "accept": false, + "reason": "Perplexity regression on GPT-OSS and GLM5 are unrelated." + }, + { + "left": "issue:45632", + "right": "issue:45698", + "accept": false, + "reason": "Both involve custom-module loading, but one is a cache-path collision and the other is loading the wrong module after save/load; not clearly the same bug." + }, + { + "left": "issue:43645", + "right": "issue:45399", + "accept": false, + "reason": "Custom models in notebooks and flash-attn fallback gating are unrelated." + }, + { + "left": "issue:43404", + "right": "issue:45005", + "accept": false, + "reason": "Both mention tied weights, but they target different model classes and APIs; not enough to treat as one concrete bug." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 3, + "review_unit_count": 49, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-426", + "nodes": [ + "issue:36331", + "issue:38175", + "issue:39692", + "issue:41553", + "issue:41628", + "issue:41762", + "issue:42915", + "issue:42947", + "issue:43054", + "issue:43493", + "issue:43643", + "issue:43644", + "issue:43782", + "issue:43828", + "issue:43856", + "issue:43866", + "issue:43872", + "issue:43873", + "issue:43881", + "issue:43883", + "issue:43950", + "issue:43975", + "issue:44292", + "issue:44360", + "issue:44387", + "issue:44403", + "issue:44451", + "issue:44488", + "issue:44492", + "issue:44512", + "issue:44530", + "issue:44534", + "issue:44589", + "issue:44863", + "issue:44910", + "issue:44918", + "issue:44928", + "issue:44960", + "issue:45005", + "issue:45020", + "issue:45237", + "issue:45399", + "issue:45507", + "issue:45663", + "issue:45701", + "issue:45800", + "issue:45859", + "issue:45902" + ], + "soft_pairs": [ + "issue:42915|issue:45399", + "issue:44530|issue:45005", + "issue:43873|issue:45005", + "issue:44863|issue:45005", + "issue:44360|issue:44512", + "issue:44918|issue:45507", + "issue:44492|issue:44512", + "issue:43975|issue:44292", + "issue:44292|issue:45507", + "issue:43950|issue:44534", + "issue:43644|issue:43950", + "issue:39692|issue:43054", + "issue:43828|issue:44928", + "issue:41628|issue:45800", + "issue:36331|issue:41762", + "issue:43872|issue:44589", + "issue:43950|issue:45902", + "issue:43866|issue:44403", + "issue:43054|issue:45701", + "issue:41553|issue:43883", + "issue:38175|issue:43054", + "issue:44910|issue:45859", + "issue:45663|issue:45902", + "issue:41553|issue:43881", + "issue:42947|issue:44387", + "issue:43054|issue:43493", + "issue:44403|issue:44451", + "issue:43975|issue:45237", + "issue:43643|issue:45020", + "issue:44403|issue:44488", + "issue:42947|issue:43856", + "issue:41762|issue:42915", + "issue:43881|issue:43883", + "issue:43782|issue:44960" + ], + "prepared_review_unit_hash": "ad5fe355f962ae403aa08d6e61eb3f300e8d32782e5671b79981a29f12c07c01", + "component_budget": { + "node_count": 426, + "item_count": 426, + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 34, + "serialized_chars": 37510, + "estimated_input_tokens": 9378, + "estimated_eval_tokens": 19012 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly heterogeneous. The only clear duplicate subcluster is the Transformers v5 non-persistent-buffer corruption regression; the rest are separate loading/quantization/model-specific bugs.", + "confidence": 0.91, + "canonical_issue_reason": "Issue 43950 is the best representative of the one clear duplicate family here: it states the non-persistent-buffer corruption regression directly and broadly enough to cover the closely related reports.", + "canonical_pr_reason": null, + "best_issue_reason": "43950 is the strongest overall issue candidate because it is the clearest, most general description of the recurring buffer-corruption bug. That said, the wider cluster is not a single coherent duplicate set.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:42915", + "right": "issue:45399", + "accept": false, + "reason": "Different failure modes: FP8 config failure on Qwen3Moe vs flash-attn fallback gating when fa2 is absent." + }, + { + "left": "issue:44530", + "right": "issue:45005", + "accept": false, + "reason": "Both involve model/runtime behavior in v5, but one is a Qwen3.5 cache crash and the other is a translation-model tied-weights issue; not the same bug." + }, + { + "left": "issue:43873", + "right": "issue:45005", + "accept": false, + "reason": "Quantization/offloading behavior is too broad; these report different concrete problems and code paths." + }, + { + "left": "issue:44863", + "right": "issue:45005", + "accept": false, + "reason": "NemotronH checkpoint loading and tied-weights handling are unrelated failures." + }, + { + "left": "issue:44360", + "right": "issue:44512", + "accept": false, + "reason": "A DSA indexer ReLU discussion is unrelated to a documentation typo about the removed run command." + }, + { + "left": "issue:44918", + "right": "issue:45507", + "accept": false, + "reason": "Different models and subsystems: Qwen3.5 SFT embedding unpacking vs GraniteMoEHybrid invalid method calls." + }, + { + "left": "issue:44492", + "right": "issue:44512", + "accept": false, + "reason": "One is a cache-strategy typo, the other is stale docs about a removed command; not the same underlying issue." + }, + { + "left": "issue:43975", + "right": "issue:44292", + "accept": false, + "reason": "Different model-specific inference/tokenization failures with no shared code-path evident." + }, + { + "left": "issue:44292", + "right": "issue:45507", + "accept": false, + "reason": "NVFP4 runtime error and GraniteMoEHybrid invalid method are unrelated." + }, + { + "left": "issue:43950", + "right": "issue:44534", + "accept": true, + "reason": "Same regression: non-persistent buffers getting corrupted/junk-filled during loading in Transformers v5." + }, + { + "left": "issue:43644", + "right": "issue:43950", + "accept": true, + "reason": "These describe the same non-persistent-buffer junk/corruption regression in Transformers 5.x." + }, + { + "left": "issue:39692", + "right": "issue:43054", + "accept": false, + "reason": "Both concern SigLIP2, but one is a broken docs example and the other is degraded text embeddings; not the same bug." + }, + { + "left": "issue:43828", + "right": "issue:44928", + "accept": false, + "reason": "Different problems: autocast dtype mismatch in Phi-tiny-MoE vs Qwen3.5 embedding unpacking in TRL SFT." + }, + { + "left": "issue:41628", + "right": "issue:45800", + "accept": false, + "reason": "Import failure for AutoImageProcessor and torch/transformers version incompatibility are separate issues." + }, + { + "left": "issue:36331", + "right": "issue:41762", + "accept": false, + "reason": "Training custom loss API mismatch is unrelated to Gemma3 loading under ZeRO-3." + }, + { + "left": "issue:43872", + "right": "issue:44589", + "accept": false, + "reason": "bitsandbytes constructor incompatibility and missing Float8 storage type are different failures." + }, + { + "left": "issue:43950", + "right": "issue:45902", + "accept": false, + "reason": "The latter is about Qwen3_5Moe rotary embedding uninitialized memory after meta->to_empty, not the same as generic non-persistent-buffer corruption." + }, + { + "left": "issue:43866", + "right": "issue:44403", + "accept": false, + "reason": "Checkpoint corruption for Ovis2 and generic loader noise are not the same underlying bug." + }, + { + "left": "issue:43054", + "right": "issue:45701", + "accept": false, + "reason": "SigLIP2 embedding quality vs tokenization changes are not the same defect." + }, + { + "left": "issue:41553", + "right": "issue:43883", + "accept": false, + "reason": "Bad AutoTokenizer error message for Voxtral and missing all_tied_weights_keys for Molmo are unrelated." + }, + { + "left": "issue:38175", + "right": "issue:43054", + "accept": false, + "reason": "Unexpected zero probabilities and degraded SigLIP2 text embeddings are different model behaviors." + }, + { + "left": "issue:44910", + "right": "issue:45859", + "accept": false, + "reason": "Flash-Attention illegal access from 3D position_ids is unrelated to CPU offload incompatibility in Qwen3_5Moe rotary embeddings." + }, + { + "left": "issue:45663", + "right": "issue:45902", + "accept": false, + "reason": "Gemma4 FSDP2 shared_kv_states rebuilding and Qwen3_5Moe rotary embedding memory initialization are unrelated." + }, + { + "left": "issue:41553", + "right": "issue:43881", + "accept": false, + "reason": "Different model-loading failures with different root causes." + }, + { + "left": "issue:42947", + "right": "issue:44387", + "accept": false, + "reason": "Gradient checkpointing ineffectiveness with LoRA and int4 reserved-memory/OOM regression are not the same bug." + }, + { + "left": "issue:43054", + "right": "issue:43493", + "accept": false, + "reason": "Different SigLIP2 concerns: embedding quality versus HF/JAX implementation discrepancy." + }, + { + "left": "issue:44403", + "right": "issue:44451", + "accept": false, + "reason": "Generic loader noise and a specific model load failure are not the same underlying issue." + }, + { + "left": "issue:43975", + "right": "issue:45237", + "accept": false, + "reason": "Incorrect detokenization in v5 and AMD GPU model failure are unrelated." + }, + { + "left": "issue:43643", + "right": "issue:45020", + "accept": false, + "reason": "Missing fields from AutoConfig with trust_remote_code is too specific to merge with the broader remote_code breakage report." + }, + { + "left": "issue:44403", + "right": "issue:44488", + "accept": false, + "reason": "Generic loading noise is not the same as a specific model load failure." + }, + { + "left": "issue:42947", + "right": "issue:43856", + "accept": false, + "reason": "LoRA gradient checkpointing and Qwen3 MoE memory usage are different training-memory issues." + }, + { + "left": "issue:41762", + "right": "issue:42915", + "accept": false, + "reason": "Gemma3 ZeRO-3 load failure and Qwen3Moe FP8 failure are unrelated." + }, + { + "left": "issue:43881", + "right": "issue:43883", + "accept": false, + "reason": "Two separate loading errors with different symptoms and likely different code paths." + }, + { + "left": "issue:43782", + "right": "issue:44960", + "accept": false, + "reason": "Qwen3VL weight_only loading error and GLM5 issue are unrelated." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet and stays conservative: it correctly treats the cluster as mostly heterogeneous while identifying the non-persistent-buffer corruption issues as the only clear duplicate family. The accepted soft-edge pairs are well supported by the issue titles and wording, and the rejected pairs avoid over-merging distinct bugs." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 4, + "review_unit_count": 49, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-426", + "nodes": [ + "issue:24643", + "issue:30064", + "issue:36331", + "issue:41093", + "issue:41553", + "issue:42491", + "issue:42915", + "issue:42947", + "issue:43257", + "issue:43366", + "issue:43404", + "issue:43638", + "issue:43742", + "issue:43792", + "issue:43824", + "issue:43827", + "issue:43828", + "issue:43854", + "issue:43866", + "issue:43872", + "issue:43940", + "issue:43950", + "issue:44220", + "issue:44291", + "issue:44368", + "issue:44479", + "issue:44488", + "issue:44509", + "issue:44530", + "issue:44560", + "issue:44661", + "issue:44805", + "issue:44863", + "issue:44928", + "issue:44991", + "issue:45020", + "issue:45081", + "issue:45305", + "issue:45313", + "issue:45325", + "issue:45356", + "issue:45399", + "issue:45538", + "issue:45588", + "issue:45685", + "issue:45823", + "issue:45902", + "issue:45923" + ], + "soft_pairs": [ + "issue:45685|issue:45823", + "issue:43824|issue:43872", + "issue:44991|issue:45399", + "issue:45020|issue:45823", + "issue:42491|issue:44368", + "issue:44661|issue:45020", + "issue:43828|issue:45823", + "issue:45081|issue:45399", + "issue:45020|issue:45356", + "issue:43742|issue:45399", + "issue:43827|issue:44509", + "issue:43404|issue:43950", + "issue:43638|issue:45823", + "issue:42915|issue:45020", + "issue:43792|issue:44220", + "issue:43257|issue:43866", + "issue:43872|issue:44291", + "issue:42915|issue:43872", + "issue:41553|issue:44991", + "issue:43940|issue:45313", + "issue:36331|issue:41093", + "issue:43854|issue:44863", + "issue:43257|issue:45305", + "issue:42947|issue:45923", + "issue:44928|issue:45588", + "issue:43872|issue:45538", + "issue:41093|issue:44805", + "issue:41553|issue:44488", + "issue:44479|issue:44560", + "issue:44530|issue:44863", + "issue:24643|issue:30064", + "issue:43366|issue:43828", + "issue:45325|issue:45902" + ], + "prepared_review_unit_hash": "58b4d1f2ad7e2e4078b8d393052617350d4a72674f8154e6933c5ff9e3bb0d1e", + "component_budget": { + "node_count": 426, + "item_count": 426, + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 33, + "serialized_chars": 38073, + "estimated_input_tokens": 9519, + "estimated_eval_tokens": 19294 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Mostly a mixed bag of unrelated runtime regressions; only a few soft pairs look like true duplicates, concentrated around shared error paths in docs, audio feature extraction, kwarg propagation, mask-shape mismatches, and Qwen video regressions.", + "confidence": 0.73, + "canonical_issue_reason": "Issue 44291 is the clearest, most root-cause-focused report of the `_is_hf_initialized` regression and subsumes the closely related bitsandbytes/init-empty-weights failure mode.", + "canonical_pr_reason": null, + "best_issue_reason": "Best representative of the cluster: it names a concrete, reproducible framework regression with a specific cause and version boundary, making it the most actionable canonical issue.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45685", + "right": "issue:45823", + "accept": false, + "reason": "Different failure modes: MPS int-histogram limitation vs a PLE device-mismatch forward bug." + }, + { + "left": "issue:43824", + "right": "issue:43872", + "accept": false, + "reason": "Both are load/runtime errors, but one is a Qwen2.5-VL import problem and the other is a bitsandbytes kwarg incompatibility." + }, + { + "left": "issue:44991", + "right": "issue:45399", + "accept": false, + "reason": "Tokenizer load failure for a specific model versus flash-attn fallback gating; only broad version-regression theme overlaps." + }, + { + "left": "issue:45020", + "right": "issue:45823", + "accept": false, + "reason": "Remote-code compatibility regression is too broad to match a specific device-map forward mismatch." + }, + { + "left": "issue:42491", + "right": "issue:44368", + "accept": false, + "reason": "Both mention Qwen/LoRA warnings, but one is a version-compatibility failure and the other is a config warning about tied embeddings." + }, + { + "left": "issue:44661", + "right": "issue:45020", + "accept": false, + "reason": "Generic model-addition/tokenizer-mapping bug is not the same concrete problem as remote_code breakage." + }, + { + "left": "issue:43828", + "right": "issue:45823", + "accept": false, + "reason": "Dtype mismatch under autocast is unrelated to the Gemma4 PLE device-placement bug." + }, + { + "left": "issue:45081", + "right": "issue:45399", + "accept": false, + "reason": "Mistral tokenizer regex crash and flash-attn fallback gating are different code paths." + }, + { + "left": "issue:45020", + "right": "issue:45356", + "accept": false, + "reason": "Remote-code regressions are too broad; Kimi-K2.5 tokenizer codec handling is a separate tokenizer-specific bug." + }, + { + "left": "issue:43742", + "right": "issue:45399", + "accept": false, + "reason": "MobileLLM load key error is unrelated to flash-attn fallback checks." + }, + { + "left": "issue:43827", + "right": "issue:44509", + "accept": true, + "reason": "Same docs regression: both report stale pipeline-task references removed in v5." + }, + { + "left": "issue:43404", + "right": "issue:43950", + "accept": false, + "reason": "Tied lm_head warning and non-persistent buffer corruption are different model-state issues." + }, + { + "left": "issue:43638", + "right": "issue:45823", + "accept": false, + "reason": "ZeRO-3 zero-size index error is unrelated to Gemma4 PLE device mismatch." + }, + { + "left": "issue:42915", + "right": "issue:45020", + "accept": false, + "reason": "Specific Qwen3-MoE hf4.x/hf5.x incompatibility is not the same as generic remote_code breakage." + }, + { + "left": "issue:43792", + "right": "issue:44220", + "accept": true, + "reason": "Both point to the same audio feature-extraction path causing Whisper to fail." + }, + { + "left": "issue:43257", + "right": "issue:43866", + "accept": false, + "reason": "Qwen3 MoE weight conversion under accelerate+deepspeed does not match a corrupted Ovis2 checkpoint report." + }, + { + "left": "issue:43872", + "right": "issue:44291", + "accept": true, + "reason": "Same underlying `_is_hf_initialized` argument regression, just surfaced through different load/init paths." + }, + { + "left": "issue:42915", + "right": "issue:43872", + "accept": false, + "reason": "Different bugs: Qwen3-MoE cross-version compatibility versus the `_is_hf_initialized` kwarg issue." + }, + { + "left": "issue:41553", + "right": "issue:44991", + "accept": false, + "reason": "Bad Voxtral tokenizer error message is not the same as a tokenizer load failure for EMBEDDIA/est-roberta." + }, + { + "left": "issue:43940", + "right": "issue:45313", + "accept": false, + "reason": "Both are ZeRO-3 weight-loading failures, but the model variants and missing-parameter patterns are different enough to avoid merging." + }, + { + "left": "issue:36331", + "right": "issue:41093", + "accept": false, + "reason": "Unexpected compute_loss kwarg and mask-shape IndexError are unrelated training/runtime failures." + }, + { + "left": "issue:43854", + "right": "issue:44863", + "accept": false, + "reason": "GLM-4.7 test-loading failure and NemotronH checkpoint-loading failure are separate model-specific bugs." + }, + { + "left": "issue:43257", + "right": "issue:45305", + "accept": false, + "reason": "DeepSpeed is shared, but weight-conversion failure and gradient-averaging bug are different code paths." + }, + { + "left": "issue:42947", + "right": "issue:45923", + "accept": false, + "reason": "Gradient checkpointing/LoRA behavior is not the same bug as a missing supports_gradient_checkpointing flag on a remote-code model." + }, + { + "left": "issue:44928", + "right": "issue:45588", + "accept": false, + "reason": "NaN gradient explosion in RLHF is unrelated to flash-attention crash on sink-less models." + }, + { + "left": "issue:43872", + "right": "issue:45538", + "accept": false, + "reason": "The `model_max_length` CLIPTokenizer issue is unrelated to the `_is_hf_initialized` regression." + }, + { + "left": "issue:41093", + "right": "issue:44805", + "accept": true, + "reason": "These are the same mask/tensor shape mismatch error, with only the report details differing." + }, + { + "left": "issue:41553", + "right": "issue:44488", + "accept": false, + "reason": "Voxtral tokenizer error-message quality and sleng-bert loading failure are different tokenizer issues." + }, + { + "left": "issue:44479", + "right": "issue:44560", + "accept": true, + "reason": "Both describe the same 5.3.0 video-input regression across Qwen VL-family models." + }, + { + "left": "issue:44530", + "right": "issue:44863", + "accept": false, + "reason": "PagedAttentionCache linear_attention crash and NemotronH checkpoint loading are unrelated." + }, + { + "left": "issue:45325", + "right": "issue:45902", + "accept": false, + "reason": "Still-image rope scaling bug and uninitialized `inv_freq` after `to_empty(cuda)` are separate rotary-embedding regressions." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet and stays conservative about the cluster being mostly unrelated runtime regressions. The accepted soft pairs are also reasonably supported by the titles/body metadata, especially the docs, mask-shape, _is_hf_initialized, and Qwen video-regression matches." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 5, + "review_unit_count": 49, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-426", + "nodes": [ + "issue:36010", + "issue:38175", + "issue:39692", + "issue:41720", + "issue:42491", + "issue:42617", + "issue:42915", + "issue:43317", + "issue:43366", + "issue:43531", + "issue:43550", + "issue:43596", + "issue:43632", + "issue:43638", + "issue:43644", + "issue:43645", + "issue:43646", + "issue:43708", + "issue:43720", + "issue:43824", + "issue:43950", + "issue:43957", + "issue:43994", + "issue:44075", + "issue:44155", + "issue:44451", + "issue:44488", + "issue:44534", + "issue:44560", + "issue:44589", + "issue:44661", + "issue:44756", + "issue:44805", + "issue:44877", + "issue:44910", + "issue:44918", + "issue:44960", + "issue:44977", + "issue:45030", + "issue:45276", + "issue:45313", + "issue:45357", + "issue:45584", + "issue:45663", + "issue:45685", + "issue:45823", + "issue:45834" + ], + "soft_pairs": [ + "issue:45663|issue:45823", + "issue:43366|issue:43531", + "issue:44155|issue:45584", + "issue:43950|issue:45834", + "issue:44560|issue:45685", + "issue:43824|issue:43957", + "issue:43550|issue:44910", + "issue:42915|issue:43957", + "issue:43366|issue:44589", + "issue:43645|issue:43646", + "issue:44560|issue:44918", + "issue:42617|issue:43366", + "issue:36010|issue:42915", + "issue:44661|issue:45823", + "issue:44560|issue:44805", + "issue:43632|issue:45834", + "issue:38175|issue:43994", + "issue:43638|issue:44661", + "issue:43596|issue:44661", + "issue:44661|issue:44805", + "issue:44877|issue:45030", + "issue:43317|issue:44756", + "issue:43644|issue:44534", + "issue:38175|issue:39692", + "issue:42491|issue:43720", + "issue:41720|issue:42915", + "issue:44960|issue:44977", + "issue:45313|issue:45357", + "issue:44451|issue:44488", + "issue:42915|issue:44756", + "issue:42491|issue:45276", + "issue:43708|issue:44075" + ], + "prepared_review_unit_hash": "c50670ff3f2b4fa7f3aeb05f25bcbd1f57a7215ba757cf783875489d9e8724ca", + "component_budget": { + "node_count": 426, + "item_count": 426, + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 32, + "serialized_chars": 36625, + "estimated_input_tokens": 9157, + "estimated_eval_tokens": 18570 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This set is mostly heterogeneous, but a few true duplicate pockets stand out. The strongest canonical thread is the Transformers v5 non-persistent-buffer regression, with a near-verbatim follow-up and a couple of adjacent custom-init reports; a few other pairs are only loosely similar and should stay separate.", + "confidence": 0.76, + "canonical_issue_reason": "Issue 43644 is the cleanest representative of the main duplicate pocket: it states a specific Transformers 5.0.0 regression about non-persistent buffers, and 44534 is a near-verbatim later report of the same bug.", + "canonical_pr_reason": null, + "best_issue_reason": "43644 is the best standalone issue to keep as the canonical bug report because it is concrete, narrowly scoped, and has the clearest duplicate signal among the candidates.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45663", + "right": "issue:45823", + "accept": false, + "reason": "Different Gemma4 failures: FSDP2 shared_kv_states KeyError vs device_map=\"auto\" PLE device mismatch." + }, + { + "left": "issue:43366", + "right": "issue:43531", + "accept": false, + "reason": "Unrelated problems: gpt-oss GGUF support vs Qwen3-MoE sliding_window behavior." + }, + { + "left": "issue:44155", + "right": "issue:45584", + "accept": false, + "reason": "Different model paths and symptoms: AudioFlamingo3 batched leakage vs Whisper empty-transcription generation." + }, + { + "left": "issue:43950", + "right": "issue:45834", + "accept": false, + "reason": "Non-persistent buffer corruption and Kosmos2.5 long-OCR index errors are different bugs." + }, + { + "left": "issue:44560", + "right": "issue:45685", + "accept": false, + "reason": "Qwen3-VL video StopIteration and an MPS histogram_int limitation are unrelated." + }, + { + "left": "issue:43824", + "right": "issue:43957", + "accept": false, + "reason": "ImportError for a missing class and meta-device loading regressions are distinct code paths." + }, + { + "left": "issue:43550", + "right": "issue:44910", + "accept": false, + "reason": "Different attention/compile failures on different models; not the same concrete bug." + }, + { + "left": "issue:42915", + "right": "issue:43957", + "accept": false, + "reason": "Qwen3Moe FineGrainedFP8Config failure is separate from generic torch.device(\"meta\") loading issues." + }, + { + "left": "issue:43366", + "right": "issue:44589", + "accept": false, + "reason": "GGUF gpt-oss support and missing Float8 storage are unrelated loading problems." + }, + { + "left": "issue:43645", + "right": "issue:43646", + "accept": true, + "reason": "Both describe the same Transformers v5 custom-model initialization regression; the notebook detail in 43645 is incidental." + }, + { + "left": "issue:44560", + "right": "issue:44918", + "accept": false, + "reason": "Different Qwen3.5/Qwen3-VL failures with different triggers and stacks." + }, + { + "left": "issue:42617", + "right": "issue:43366", + "accept": false, + "reason": "3d_parallel.py execution failure is not the same bug as gpt-oss GGUF support." + }, + { + "left": "issue:36010", + "right": "issue:42915", + "accept": false, + "reason": "GenerationMixin import failure and Qwen3MoE FP8 issues are unrelated." + }, + { + "left": "issue:44661", + "right": "issue:45823", + "accept": false, + "reason": "Tokenizer-mapping registration bug and Gemma4 device mismatch are different problems." + }, + { + "left": "issue:44560", + "right": "issue:44805", + "accept": false, + "reason": "Qwen3-VL video StopIteration is not the same as a generic mask shape mismatch." + }, + { + "left": "issue:43632", + "right": "issue:45834", + "accept": false, + "reason": "_is_hf_initialized regression does not match the Kosmos2.5 long-input index error." + }, + { + "left": "issue:38175", + "right": "issue:43994", + "accept": true, + "reason": "Both concern broken SigLIP2-base-patch16-224 inference producing nonsensical/zero outputs through the same model/processor path." + }, + { + "left": "issue:43638", + "right": "issue:44661", + "accept": false, + "reason": "Zero3 Bert initialization and tokenizer-mapping registration are not the same underlying issue." + }, + { + "left": "issue:43596", + "right": "issue:44661", + "accept": false, + "reason": "Bert Zero3 index error is a different failure mode from the tokenizer mapping bug." + }, + { + "left": "issue:44661", + "right": "issue:44805", + "accept": false, + "reason": "Tokenizer mapping validation and mask/tensor shape mismatch are unrelated." + }, + { + "left": "issue:44877", + "right": "issue:45030", + "accept": false, + "reason": "Strict config blocking granite_speech and glm4v config validation are separate issues." + }, + { + "left": "issue:43317", + "right": "issue:44756", + "accept": false, + "reason": "dequantized model offload failure and mmap/OOM avoidance on Strix Halo are not the same bug." + }, + { + "left": "issue:43644", + "right": "issue:44534", + "accept": true, + "reason": "These are near-verbatim reports of the same Transformers v5 non-persistent-buffer junk regression." + }, + { + "left": "issue:38175", + "right": "issue:39692", + "accept": false, + "reason": "Both mention SigLIP2, but 39692 is a docs/example problem with additional quantization issues, not clearly the same concrete bug as 38175." + }, + { + "left": "issue:42491", + "right": "issue:43720", + "accept": false, + "reason": "LoRA cross-version compatibility for qwen3_moe and BitNet packed-weight loading are different code paths." + }, + { + "left": "issue:41720", + "right": "issue:42915", + "accept": false, + "reason": "cudaErrorAssert on A800 with auto device mapping is unrelated to FineGrainedFP8Config failure." + }, + { + "left": "issue:44960", + "right": "issue:44977", + "accept": false, + "reason": "GLM5 and Qwen3.5 flash-attention generation failures are different model-specific bugs." + }, + { + "left": "issue:45313", + "right": "issue:45357", + "accept": false, + "reason": "ZeRO-3 language_model weight loading and save_pretrained visual-key corruption are separate issues." + }, + { + "left": "issue:44451", + "right": "issue:44488", + "accept": true, + "reason": "Both report the same current-version regression loading specific BERT checkpoints, likely the same model-loading breakage." + }, + { + "left": "issue:42491", + "right": "issue:45276", + "accept": false, + "reason": "qwen3_moe LoRA incompatibility and Gemma4 resize_token_embeddings behavior are unrelated." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The canonical 43644/44534 duplicate pocket is well grounded, and the other accepted pairs are phrased cautiously enough to stay within the packet evidence. The rejected pairs also look appropriately conservative. No obvious overstatement stands out." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 6, + "review_unit_count": 49, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-426", + "nodes": [ + "issue:36010", + "issue:39404", + "issue:39692", + "issue:41720", + "issue:42491", + "issue:42947", + "issue:43257", + "issue:43335", + "issue:43381", + "issue:43404", + "issue:43425", + "issue:43454", + "issue:43644", + "issue:43645", + "issue:43854", + "issue:43856", + "issue:43873", + "issue:43931", + "issue:43957", + "issue:43994", + "issue:44155", + "issue:44164", + "issue:44291", + "issue:44292", + "issue:44387", + "issue:44402", + "issue:44488", + "issue:44756", + "issue:44910", + "issue:44912", + "issue:44928", + "issue:44960", + "issue:44991", + "issue:45216", + "issue:45276", + "issue:45292", + "issue:45335", + "issue:45362", + "issue:45406", + "issue:45488", + "issue:45588", + "issue:45663", + "issue:45704", + "issue:45759", + "issue:45812", + "issue:45823", + "issue:45859", + "issue:46050" + ], + "soft_pairs": [ + "issue:43854|issue:43931", + "issue:43257|issue:43854", + "issue:45488|issue:45812", + "issue:44488|issue:44991", + "issue:43644|issue:43645", + "issue:39692|issue:43994", + "issue:43856|issue:43873", + "issue:42491|issue:43931", + "issue:45362|issue:45406", + "issue:43856|issue:44387", + "issue:44960|issue:46050", + "issue:42491|issue:45292", + "issue:44910|issue:45588", + "issue:44164|issue:44291", + "issue:36010|issue:39404", + "issue:45276|issue:45335", + "issue:43404|issue:43454", + "issue:43425|issue:43957", + "issue:41720|issue:44155", + "issue:39692|issue:44960", + "issue:44402|issue:44488", + "issue:42947|issue:45704", + "issue:43873|issue:44387", + "issue:44291|issue:44756", + "issue:45663|issue:45759", + "issue:45663|issue:45859", + "issue:44155|issue:45823", + "issue:43335|issue:45663", + "issue:44292|issue:44912", + "issue:44387|issue:44756", + "issue:43957|issue:44756", + "issue:43854|issue:45216", + "issue:45276|issue:45292", + "issue:42947|issue:43381", + "issue:43381|issue:44928", + "issue:44164|issue:44756" + ], + "prepared_review_unit_hash": "84bf4590b2367a88d6d0a96b6b051b222e0300a2072204da53864b5ce195b28a", + "component_budget": { + "node_count": 426, + "item_count": 426, + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 36, + "serialized_chars": 38639, + "estimated_input_tokens": 9660, + "estimated_eval_tokens": 19576 + }, + "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 bugs. The only credible duplicate subcluster is around `resize_token_embeddings` not updating model-specific output embeddings, where one issue is the broader generic report and the other is a Gemma4-specific manifestation.", + "confidence": 0.64, + "canonical_issue_reason": "Issue 45292 is the best anchor because it is the broad, model-agnostic version of the only apparent duplicate family (`resize_token_embeddings` not affecting `output_embeddings`).", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 45292 is the strongest representative overall: it is the most general and concise description of the shared `resize_token_embeddings`/`output_embeddings` regression, while the nearby issues are narrower model-specific variants.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43854", + "right": "issue:43931", + "accept": false, + "reason": "Different bugs: GLM-4.7 unit-test load failure vs Qwen3-VL weight-shape mismatch." + }, + { + "left": "issue:43257", + "right": "issue:43854", + "accept": false, + "reason": "Different model-loading failures with unrelated root causes (Qwen3 MoE conversion vs GLM-4.7 test load)." + }, + { + "left": "issue:45488", + "right": "issue:45812", + "accept": false, + "reason": "Both are tokenizer regressions, but they target different model families and different failure modes (`LlamaTokenizer` pre-tokenizer override vs Granite token IDs)." + }, + { + "left": "issue:44488", + "right": "issue:44991", + "accept": false, + "reason": "Different tokenizer-load problems for different models; no clear shared code path." + }, + { + "left": "issue:43644", + "right": "issue:43645", + "accept": false, + "reason": "Both are Transformers 5.0 regressions, but one is junk non-persistent buffers and the other is custom-model initialization in Jupyter; unrelated." + }, + { + "left": "issue:39692", + "right": "issue:43994", + "accept": false, + "reason": "Same model family (SigLIP2) but different reported issues: docs/example errors and quantization failure vs nonsensical inference results." + }, + { + "left": "issue:43856", + "right": "issue:43873", + "accept": false, + "reason": "Different problems: Qwen3 MoE training memory efficiency vs quantization/offloading behavior." + }, + { + "left": "issue:42491", + "right": "issue:43931", + "accept": false, + "reason": "Unrelated issues: LoRA compatibility across HF versions vs Qwen3-VL load-time shape mismatch." + }, + { + "left": "issue:45362", + "right": "issue:45406", + "accept": false, + "reason": "Different model/runtime crashes: Qwen3.5 chat failure vs Gemma4Processor missing `_tokenizer` in serve." + }, + { + "left": "issue:43856", + "right": "issue:44387", + "accept": false, + "reason": "Both mention memory, but one is Qwen3 MoE training efficiency and the other is int4 quantization reserved-memory/OOM; not the same bug." + }, + { + "left": "issue:44960", + "right": "issue:46050", + "accept": false, + "reason": "Different model families and failure modes; GLM5 vs sinq save/load corruption." + }, + { + "left": "issue:42491", + "right": "issue:45292", + "accept": false, + "reason": "Different root causes; Qwen3 MoE HF4.x/HF5.x incompatibility is not the same as a generic `resize_token_embeddings` regression." + }, + { + "left": "issue:44910", + "right": "issue:45588", + "accept": false, + "reason": "Both involve attention code, but one is Qwen3.5 flash-attn illegal memory access from 3D `position_ids`, the other is `s_aux=None` handling in sink-less models." + }, + { + "left": "issue:44164", + "right": "issue:44291", + "accept": false, + "reason": "Both are loading/saving internals, but one is `extra_state` serialization and the other is `init_empty_weights` argument compatibility." + }, + { + "left": "issue:36010", + "right": "issue:39404", + "accept": false, + "reason": "Completely different issues: import path breakage vs Whisper pipeline `return_language` regression." + }, + { + "left": "issue:45276", + "right": "issue:45335", + "accept": false, + "reason": "Similar API symptom, but different model families and target embeddings (`Gemma4` output/embed tokens vs `t5gemma` decoder embed tokens); not clearly the same bug." + }, + { + "left": "issue:43404", + "right": "issue:43454", + "accept": false, + "reason": "Both are weight-tying issues, but they affect different model classes and are not clearly the same concrete code-path problem." + }, + { + "left": "issue:43425", + "right": "issue:43957", + "accept": false, + "reason": "Unrelated compatibility/loading bugs: Torch 2.10 vs meta-device loading regressions." + }, + { + "left": "issue:41720", + "right": "issue:44155", + "accept": false, + "reason": "Different subsystems and failures: Qwen3 auto device mapping CUDA assert vs AudioFlamingo3 batched inference leakage." + }, + { + "left": "issue:39692", + "right": "issue:44960", + "accept": false, + "reason": "Different model families and different problems; SigLIP2 docs/example issues are not the same as GLM5 model loading." + }, + { + "left": "issue:44402", + "right": "issue:44488", + "accept": false, + "reason": "Both are token/vocab-related model-loading issues, but they concern different models and different underlying tokenizer problems." + }, + { + "left": "issue:42947", + "right": "issue:45704", + "accept": false, + "reason": "Different bugs: PEFT LoRA gradient checkpointing vs T5 accidentally using apex RMSNorm." + }, + { + "left": "issue:43873", + "right": "issue:44387", + "accept": false, + "reason": "Both involve quantization/memory, but one is offloading not working and the other is increased CUDA reserved memory under int4; separate issues." + }, + { + "left": "issue:44291", + "right": "issue:44756", + "accept": false, + "reason": "Different concerns: `init_empty_weights` type error vs Strix Halo mmap/OOM." + }, + { + "left": "issue:45663", + "right": "issue:45759", + "accept": false, + "reason": "Different model architecture bugs: Gemma4 FSDP2 `shared_kv_states` vs Qwen 3.5/3.6 `text_config` unwrap behavior." + }, + { + "left": "issue:45663", + "right": "issue:45859", + "accept": false, + "reason": "Different code paths: Gemma4 FSDP2 attention state vs Qwen3_5 MoE rotary embedding CPU offload." + }, + { + "left": "issue:44155", + "right": "issue:45823", + "accept": false, + "reason": "No shared underlying bug; AudioFlamingo3 track leakage is unrelated to Gemma4 PLE device mismatch." + }, + { + "left": "issue:43335", + "right": "issue:45663", + "accept": false, + "reason": "Different model families and failures; SwitchTransformers sparse-layer creation is unrelated to Gemma4 FSDP2 key errors." + }, + { + "left": "issue:44292", + "right": "issue:44912", + "accept": false, + "reason": "Both involve quantization, but one is a Qwen-3 NVFP4 runtime error and the other is MXFP4 load fallback; not the same issue." + }, + { + "left": "issue:44387", + "right": "issue:44756", + "accept": false, + "reason": "Both mention memory, but one is int4 reserved-memory/OOM and the other is mmap on Strix Halo; unrelated." + }, + { + "left": "issue:43957", + "right": "issue:44756", + "accept": false, + "reason": "Different bugs: meta-device loading regressions vs Strix Halo mmap/OOM." + }, + { + "left": "issue:43854", + "right": "issue:45216", + "accept": false, + "reason": "Different load/save bugs on different models: GLM-4.7 unit-test loading vs Qwen3.5 checkpoint save regression." + }, + { + "left": "issue:45276", + "right": "issue:45292", + "accept": true, + "reason": "Same underlying `resize_token_embeddings` regression: both report output embeddings not being updated, with 45276 adding a Gemma4-specific manifestation." + }, + { + "left": "issue:42947", + "right": "issue:43381", + "accept": false, + "reason": "Both involve gradient checkpointing language, but one is ineffective LoRA checkpointing and the other is eval-mode incompatibility." + }, + { + "left": "issue:43381", + "right": "issue:44928", + "accept": false, + "reason": "Different checkpointing/SDPA problems; eval-mode checkpointing is unrelated to Qwen3.5 3D `position_ids` NaN behavior." + }, + { + "left": "issue:44164", + "right": "issue:44756", + "accept": false, + "reason": "Serialization `extra_state` handling is unrelated to mmap/OOM on Strix Halo." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is well grounded in the packet: most issues are clearly unrelated by model/failure mode, and the one accepted duplicate pair (45276/45292) is a conservative, defensible match. The soft-edge rejections are generally cautious rather than overstated. One minor omission is that 45335 also looks adjacent to the same resize_token_embeddings family, but that does not materially undermine the overall assessment." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 7, + "review_unit_count": 49, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-426", + "nodes": [ + "issue:32090", + "issue:36032", + "issue:39404", + "issue:43257", + "issue:43299", + "issue:43425", + "issue:43452", + "issue:43596", + "issue:43611", + "issue:43645", + "issue:43646", + "issue:43708", + "issue:43716", + "issue:43828", + "issue:43901", + "issue:43906", + "issue:43931", + "issue:43957", + "issue:43994", + "issue:44079", + "issue:44164", + "issue:44265", + "issue:44402", + "issue:44451", + "issue:44479", + "issue:44509", + "issue:44530", + "issue:44589", + "issue:44671", + "issue:44683", + "issue:44756", + "issue:44805", + "issue:44863", + "issue:44877", + "issue:44898", + "issue:44912", + "issue:44918", + "issue:44991", + "issue:45084", + "issue:45292", + "issue:45335", + "issue:45588", + "issue:45663", + "issue:45684", + "issue:45698", + "issue:45701", + "issue:45759", + "issue:45823" + ], + "soft_pairs": [ + "issue:45684|issue:45698", + "issue:44479|issue:44877", + "issue:45588|issue:45663", + "issue:43646|issue:43708", + "issue:43645|issue:44164", + "issue:43828|issue:43957", + "issue:45292|issue:45335", + "issue:44402|issue:44451", + "issue:43425|issue:43828", + "issue:43257|issue:44863", + "issue:44683|issue:45823", + "issue:44589|issue:44756", + "issue:39404|issue:45701", + "issue:43425|issue:45823", + "issue:43425|issue:43716", + "issue:43611|issue:43645", + "issue:44530|issue:45084", + "issue:43906|issue:44530", + "issue:39404|issue:44912", + "issue:44265|issue:44898", + "issue:44479|issue:44805", + "issue:43901|issue:44509", + "issue:43994|issue:44079", + "issue:43425|issue:44918", + "issue:44589|issue:45084", + "issue:43299|issue:43931", + "issue:43596|issue:44683", + "issue:39404|issue:44671", + "issue:44164|issue:45759", + "issue:36032|issue:43452", + "issue:44671|issue:44991", + "issue:32090|issue:39404" + ], + "prepared_review_unit_hash": "6c8dabfe29e2b70254f899c5f1167e8a9f026d7bdf91ee577ce66076b3ad40d0", + "component_budget": { + "node_count": 426, + "item_count": 426, + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 32, + "serialized_chars": 37406, + "estimated_input_tokens": 9352, + "estimated_eval_tokens": 18960 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is heterogeneous and mostly contains distinct Transformers regressions across model loading, tokenization, attention backends, and training/runtime edge cases. One clear duplicate-like pair appears around `resize_token_embeddings` not propagating to output embeddings.", + "confidence": 0.62, + "canonical_issue_reason": "`issue:43611` is the best broad representative: it describes a general Transformers 5.0 model-loading regression, which matches the dominant theme of the cluster better than the narrower model-specific reports.", + "canonical_pr_reason": null, + "best_issue_reason": "`issue:43611` is the most suitable anchor issue because it is broad, clearly regression-oriented, and centrally aligned with the repeated v5 loading breakages in this set.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45684", + "right": "issue:45698", + "accept": false, + "reason": "Both involve custom-model save/load, but one is about read-only permissions during `save_pretrained` and the other is about `from_pretrained` selecting the wrong custom module. Different bugs." + }, + { + "left": "issue:44479", + "right": "issue:44877", + "accept": false, + "reason": "Different models and failures: Qwen VL video-input regression vs strict config blocking `granite_speech` loading." + }, + { + "left": "issue:45588", + "right": "issue:45663", + "accept": false, + "reason": "`s_aux=None` crash in flash attention for sink-less models is unrelated to Gemma4 FSDP2 `shared_kv_states` rebuilding." + }, + { + "left": "issue:43646", + "right": "issue:43708", + "accept": false, + "reason": "Custom model initialization breakage and `resume_from_checkpoint` max-step miscalculation are separate code paths." + }, + { + "left": "issue:43645", + "right": "issue:44164", + "accept": false, + "reason": "Notebook custom model initialization and `extra_state` save/load handling are different issues." + }, + { + "left": "issue:43828", + "right": "issue:43957", + "accept": false, + "reason": "Phi-tiny-MoE autocast dtype mismatch and meta-device loading breakage are different runtime problems." + }, + { + "left": "issue:45292", + "right": "issue:45335", + "accept": true, + "reason": "Same underlying bug: `resize_token_embeddings` fails to update output embeddings; `t5gemma` is just a specific manifestation affecting `decoder.embed_tokens`." + }, + { + "left": "issue:44402", + "right": "issue:44451", + "accept": false, + "reason": "The titles both concern tokenizer/model loading, but they do not point to the same concrete failure or shared code path with enough evidence." + }, + { + "left": "issue:43425", + "right": "issue:43828", + "accept": false, + "reason": "Torch 2.10 compatibility is a different problem from the Phi-tiny-MoE autocast dtype mismatch." + }, + { + "left": "issue:43257", + "right": "issue:44863", + "accept": false, + "reason": "Qwen3 MOE weight conversion with accelerate+deepspeed and NemotronH checkpoint loading are different model-specific loading bugs." + }, + { + "left": "issue:44683", + "right": "issue:45823", + "accept": false, + "reason": "Compiled flex_attention on torch >=2.9 is unrelated to Gemma4 PLE device mismatch during forward." + }, + { + "left": "issue:44589", + "right": "issue:44756", + "accept": false, + "reason": "Missing Float8 storage support and disabling mmap to avoid OOM are unrelated failure modes." + }, + { + "left": "issue:39404", + "right": "issue:45701", + "accept": false, + "reason": "Whisper pipeline `return_language` regression and tokenization changes are not the same bug." + }, + { + "left": "issue:43425", + "right": "issue:45823", + "accept": false, + "reason": "General torch version incompatibility is not the same concrete issue as Gemma4 PLE device mismatch." + }, + { + "left": "issue:43425", + "right": "issue:43716", + "accept": false, + "reason": "Torch 2.10 incompatibility and Mistral-3 image preprocessor dtype mismatch are different problems." + }, + { + "left": "issue:43611", + "right": "issue:43645", + "accept": false, + "reason": "Both are v5 regressions, but one is generic model loading with `base_model_prefix` while the other is custom model initialization in notebooks." + }, + { + "left": "issue:44530", + "right": "issue:45084", + "accept": false, + "reason": "PagedAttentionCache `linear_attention` crash and non-template-node compilation failure are different attention/compiler issues." + }, + { + "left": "issue:43906", + "right": "issue:44530", + "accept": false, + "reason": "An isolated reproduction of another issue is not the same as the Qwen3.5 PagedAttentionCache crash." + }, + { + "left": "issue:39404", + "right": "issue:44912", + "accept": false, + "reason": "Whisper pipeline language regression and MXFP4 quantization fallback are unrelated." + }, + { + "left": "issue:44265", + "right": "issue:44898", + "accept": false, + "reason": "`torch.export.export` failing with `torch_compilable_check` is a different path from Perceiver non-default-resolution handling." + }, + { + "left": "issue:44479", + "right": "issue:44805", + "accept": false, + "reason": "A Qwen VL video regression is not the same as the mask-shape indexing error." + }, + { + "left": "issue:43901", + "right": "issue:44509", + "accept": false, + "reason": "Both are docs-related, but they concern different pipeline/task deprecations and should not be merged as one bug." + }, + { + "left": "issue:43994", + "right": "issue:44079", + "accept": false, + "reason": "Siglip2 producing bad outputs and `ModelOutput` key assignment when a previous key is `None` are different issues." + }, + { + "left": "issue:43425", + "right": "issue:44918", + "accept": false, + "reason": "Torch compatibility and TRL SFT embedding unpacking are unrelated." + }, + { + "left": "issue:44589", + "right": "issue:45084", + "accept": false, + "reason": "Float8 storage lookup failure and compilation of non-template nodes are separate failures." + }, + { + "left": "issue:43299", + "right": "issue:43931", + "accept": false, + "reason": "Both concern Qwen3-VL loading, but one is a dev0 regression and the other is a checkpoint shape mismatch; not enough evidence of the same underlying bug." + }, + { + "left": "issue:43596", + "right": "issue:44683", + "accept": false, + "reason": "Deepspeed ZeRO-3/BertModel indexing and compiled flex_attention on torch >=2.9 are unrelated." + }, + { + "left": "issue:39404", + "right": "issue:44671", + "accept": false, + "reason": "Whisper pipeline behavior and CamemBERT masked-LM predictions are different model-specific regressions." + }, + { + "left": "issue:44164", + "right": "issue:45759", + "accept": false, + "reason": "`extra_state` save/load handling and `from_config` not unwrapping `text_config` are different model-loading behaviors." + }, + { + "left": "issue:36032", + "right": "issue:43452", + "accept": false, + "reason": "Tokenizer method-name conflict and `gguf_file` handling for tokenizer/model loading are different bugs." + }, + { + "left": "issue:44671", + "right": "issue:44991", + "accept": false, + "reason": "CamemBERT prediction regression and tokenizer loading failure for EMBEDDIA/est-roberta are not the same issue." + }, + { + "left": "issue:32090", + "right": "issue:39404", + "accept": false, + "reason": "Trainer GPU broadcast `NoneType` error and Whisper `return_language` pipeline regression are unrelated." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The selection and summary are grounded in the packet: issue:43611 is a defensible broad representative for the repeated v5 model-loading regressions, and the lone accepted soft pair (issue:45292 / issue:45335) looks like the same resize_token_embeddings bug at a model-specific level. The rejected pairs are stated conservatively, with no clear overclaiming or unsupported mergeability." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 8, + "review_unit_count": 49, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-426", + "nodes": [ + "issue:34567", + "issue:36032", + "issue:40444", + "issue:40990", + "issue:42915", + "issue:43089", + "issue:43262", + "issue:43317", + "issue:43377", + "issue:43421", + "issue:43452", + "issue:43482", + "issue:43541", + "issue:43582", + "issue:43611", + "issue:43632", + "issue:43643", + "issue:43644", + "issue:43645", + "issue:43646", + "issue:43716", + "issue:43856", + "issue:43883", + "issue:44164", + "issue:44206", + "issue:44220", + "issue:44451", + "issue:44464", + "issue:44479", + "issue:44488", + "issue:44530", + "issue:44534", + "issue:44661", + "issue:44683", + "issue:44737", + "issue:44756", + "issue:44811", + "issue:44849", + "issue:44987", + "issue:44991", + "issue:45081", + "issue:45084", + "issue:45325", + "issue:45507", + "issue:45584", + "issue:45859", + "issue:45987" + ], + "soft_pairs": [ + "issue:44991|issue:45081", + "issue:43856|issue:44756", + "issue:44164|issue:44479", + "issue:44737|issue:45859", + "issue:44661|issue:44683", + "issue:43582|issue:43632", + "issue:44488|issue:45081", + "issue:44488|issue:44987", + "issue:43611|issue:43644", + "issue:40990|issue:43421", + "issue:43716|issue:45084", + "issue:44206|issue:44220", + "issue:36032|issue:42915", + "issue:43643|issue:44530", + "issue:43883|issue:44488", + "issue:34567|issue:40444", + "issue:44811|issue:45584", + "issue:43541|issue:45084", + "issue:44451|issue:44987", + "issue:42915|issue:44530", + "issue:44464|issue:45987", + "issue:43644|issue:43646", + "issue:43452|issue:43482", + "issue:44849|issue:45325", + "issue:36032|issue:43645", + "issue:43317|issue:43482", + "issue:44987|issue:44991", + "issue:45084|issue:45507", + "issue:43883|issue:44991", + "issue:43262|issue:43377", + "issue:43089|issue:45987", + "issue:43645|issue:44534" + ], + "prepared_review_unit_hash": "557ab533fdfcc7ae3e096ad04e5a873b35329a28fe70599a929f573847b31618", + "component_budget": { + "node_count": 426, + "item_count": 426, + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 32, + "serialized_chars": 36655, + "estimated_input_tokens": 9164, + "estimated_eval_tokens": 18584 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This set is mostly a grab bag of unrelated issue reports that share broad Transformers/version/model vocabulary but not the same underlying bug. The soft pairs are all false positives and should be kept separate.", + "confidence": 0.87, + "canonical_issue_reason": "No single canonical duplicate stands out: the items cover many distinct regressions across different models, loaders, and runtime paths.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 43611 is the broadest representative of the Transformers v5 regression family here, but it is still only one specific loading bug and should not absorb the others.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44991", + "right": "issue:45081", + "accept": false, + "reason": "Both involve tokenizer/model loading regressions, but one is a tokenizer failure for EMBEDDIA/est-roberta and the other is a Mistral regex patch crash; different root causes and code paths." + }, + { + "left": "issue:43856", + "right": "issue:44756", + "accept": false, + "reason": "One is MoE training memory inefficiency; the other is mmap-related OOM on Strix Halo. Same broad area, not the same bug." + }, + { + "left": "issue:44164", + "right": "issue:44479", + "accept": false, + "reason": "save/from_pretrained extra_state handling is unrelated to the Qwen video input regression." + }, + { + "left": "issue:44737", + "right": "issue:45859", + "accept": false, + "reason": "Both mention device placement, but XLNet CPU arange and Qwen3.5 MoE CPU-offload rotary embedding failures are different code paths." + }, + { + "left": "issue:44661", + "right": "issue:44683", + "accept": false, + "reason": "Tokenizer mapping/add-new-model-like logic is unrelated to compiled flex_attention failing on torch >= 2.9." + }, + { + "left": "issue:43582", + "right": "issue:43632", + "accept": false, + "reason": "Apple Silicon allocator warmup TypeError and the _is_hf_initialized flag regression are separate issues." + }, + { + "left": "issue:44488", + "right": "issue:45081", + "accept": false, + "reason": "Both are loading failures, but one is a model/tokenizer load issue for a specific repo and the other is a Mistral tokenizer regex-patching crash." + }, + { + "left": "issue:44488", + "right": "issue:44987", + "accept": false, + "reason": "These are both model-loading failures, but they affect different models and likely different breakpoints in the load pipeline." + }, + { + "left": "issue:43611", + "right": "issue:43644", + "accept": false, + "reason": "Both are Transformers v5 regressions, but one is base_model_prefix loading and the other is non-persistent buffer initialization; not the same bug." + }, + { + "left": "issue:40990", + "right": "issue:43421", + "accept": false, + "reason": "Perplexity on gpt-oss-20b is unrelated to runtime post-processor updates when special tokens change." + }, + { + "left": "issue:43716", + "right": "issue:45084", + "accept": false, + "reason": "Image preprocessor dtype mismatch and template compilation errors are different subsystems with different failure modes." + }, + { + "left": "issue:44206", + "right": "issue:44220", + "accept": false, + "reason": "Both are audio-related, but one is an unsupported center arg in a feature extractor and the other is a low-level fbank extraction issue." + }, + { + "left": "issue:36032", + "right": "issue:42915", + "accept": false, + "reason": "T5 tokenizer add_special_tokens conflict and Qwen3Moe FP8 config failure are unrelated." + }, + { + "left": "issue:43643", + "right": "issue:44530", + "accept": false, + "reason": "AutoConfig trust_remote_code field loss is not the same as PagedAttentionCache rejecting linear_attention groups." + }, + { + "left": "issue:43883", + "right": "issue:44488", + "accept": false, + "reason": "Missing all_tied_weights_keys in Molmo is unrelated to a model-loading failure for cjvt/sleng-bert." + }, + { + "left": "issue:34567", + "right": "issue:40444", + "accept": false, + "reason": "TrainerState token counting and Qwen2.5-VL iterable-dataset multi-image finetuning are distinct bugs." + }, + { + "left": "issue:44811", + "right": "issue:45584", + "accept": false, + "reason": "Both involve Whisper decoding, but one ignores skip_special_tokens and the other fails on empty transcription after align_special_tokens." + }, + { + "left": "issue:43541", + "right": "issue:45084", + "accept": false, + "reason": "Mixtral grouped_mm tracing errors and non-template-node compilation errors are not the same underlying issue." + }, + { + "left": "issue:44451", + "right": "issue:44987", + "accept": false, + "reason": "Different models and different loading failures; the shared symptom of 'cannot load' is too broad." + }, + { + "left": "issue:42915", + "right": "issue:44530", + "accept": false, + "reason": "Both are Qwen-family regressions, but FP8 config failure and invalid group type in PagedAttentionCache are separate code paths." + }, + { + "left": "issue:44464", + "right": "issue:45987", + "accept": false, + "reason": "Chunked generation inconsistency with compiled forward is unrelated to StaticCache.get_seq_length returning the wrong type." + }, + { + "left": "issue:43644", + "right": "issue:43646", + "accept": false, + "reason": "Both are v5 regressions around model initialization, but non-persistent buffer junk and custom-model init breakage are different concrete defects." + }, + { + "left": "issue:43452", + "right": "issue:43482", + "accept": false, + "reason": "GGUF-related loading failures, but one is a generic gguf_file/from_pretrained issue and the other is a specific Qwen2.5-GGUF v5 compatibility break." + }, + { + "left": "issue:44849", + "right": "issue:45325", + "accept": false, + "reason": "Qwen3.5 output_hidden_states and Qwen2.5-VL rope index scaling are unrelated model-specific bugs." + }, + { + "left": "issue:36032", + "right": "issue:43645", + "accept": false, + "reason": "Tokenizer special-token method conflicts and notebook custom-model initialization failures do not share the same underlying defect." + }, + { + "left": "issue:43317", + "right": "issue:43482", + "accept": false, + "reason": "gpu+cpu offload dequantized-model loading and Qwen2.5-GGUF loading are different failure modes." + }, + { + "left": "issue:44987", + "right": "issue:44991", + "accept": false, + "reason": "Both are Transformers>=5 loading regressions, but they affect different artifacts and appear to stem from different loader paths." + }, + { + "left": "issue:45084", + "right": "issue:45507", + "accept": false, + "reason": "Template compilation and GraniteMoEHybrid invalid-method calls are unrelated." + }, + { + "left": "issue:43883", + "right": "issue:44991", + "accept": false, + "reason": "Molmo weight-key attribute handling is unrelated to tokenizer loading for est-roberta." + }, + { + "left": "issue:43262", + "right": "issue:43377", + "accept": false, + "reason": "Audio chat-template sampling-rate defaults and MIMI batched-vs-single output differences are separate bugs." + }, + { + "left": "issue:43089", + "right": "issue:45987", + "accept": false, + "reason": "Generation performance overhead and StaticCache length type contract are not the same issue." + }, + { + "left": "issue:43645", + "right": "issue:44534", + "accept": false, + "reason": "Both are v5 regressions, but Jupyter custom-model initialization and non-persistent buffer junk are distinct bugs." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 9, + "review_unit_count": 49, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-426", + "nodes": [ + "issue:36032", + "issue:40444", + "issue:40990", + "issue:41553", + "issue:41720", + "issue:42175", + "issue:42915", + "issue:43296", + "issue:43322", + "issue:43377", + "issue:43404", + "issue:43425", + "issue:43531", + "issue:43540", + "issue:43582", + "issue:43611", + "issue:43632", + "issue:43646", + "issue:43742", + "issue:43792", + "issue:43819", + "issue:43994", + "issue:44062", + "issue:44079", + "issue:44220", + "issue:44265", + "issue:44291", + "issue:44403", + "issue:44451", + "issue:44466", + "issue:44479", + "issue:44534", + "issue:44589", + "issue:44610", + "issue:44683", + "issue:44756", + "issue:44811", + "issue:44877", + "issue:44987", + "issue:45020", + "issue:45081", + "issue:45084", + "issue:45127", + "issue:45216", + "issue:45325", + "issue:45357", + "issue:45542", + "issue:45663" + ], + "soft_pairs": [ + "issue:43404|issue:44466", + "issue:43646|issue:44534", + "issue:44403|issue:45081", + "issue:43322|issue:43540", + "issue:42915|issue:45084", + "issue:43994|issue:44610", + "issue:43377|issue:43819", + "issue:44220|issue:44683", + "issue:44265|issue:44610", + "issue:44451|issue:45081", + "issue:44987|issue:45081", + "issue:42915|issue:44291", + "issue:45216|issue:45357", + "issue:44062|issue:44987", + "issue:43632|issue:44756", + "issue:36032|issue:43742", + "issue:42915|issue:44589", + "issue:44479|issue:45325", + "issue:43404|issue:45127", + "issue:43296|issue:43531", + "issue:36032|issue:41720", + "issue:40444|issue:45325", + "issue:43531|issue:44877", + "issue:43611|issue:44534", + "issue:43425|issue:44403", + "issue:40990|issue:44811", + "issue:43792|issue:44589", + "issue:43632|issue:44534", + "issue:44079|issue:45663", + "issue:41553|issue:43582", + "issue:44683|issue:45020", + "issue:42175|issue:45542" + ], + "prepared_review_unit_hash": "cc463fac0119b0f097d1a1e2a574524cc09941fd3e5fdd49b2f8f667f970447d", + "component_budget": { + "node_count": 426, + "item_count": 426, + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 32, + "serialized_chars": 37206, + "estimated_input_tokens": 9302, + "estimated_eval_tokens": 18860 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a grab-bag of unrelated Transformers bug reports, many of them v5 regression/loading issues. Only the Qwen3.5 save_pretrained reports look like a likely duplicate pair; the rest are too different in model, failure mode, or code path to merge.", + "confidence": 0.68, + "canonical_issue_reason": "issue:43646 is the broadest representative of the v5 regression/theme here: a general loading/initialization breakage that fits the cluster better than the more model-specific reports.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43646 is the best single triage representative because it is broad, clear, and central to the repeated Transformers v5 breakage pattern in this set.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43404", + "right": "issue:44466", + "accept": false, + "reason": "Related tied-weights symptoms, but one is model init tying and the other is save_pretrained serialization; different code paths." + }, + { + "left": "issue:43646", + "right": "issue:44534", + "accept": false, + "reason": "Both are v5 regressions, but custom model initialization and non-persistent buffer junk are distinct bugs." + }, + { + "left": "issue:44403", + "right": "issue:45081", + "accept": false, + "reason": "Unrelated failures: noisy loading message vs Mistral tokenizer regex patch crash." + }, + { + "left": "issue:43322", + "right": "issue:43540", + "accept": false, + "reason": "Different multimodal failures on different models and inputs; not the same underlying bug." + }, + { + "left": "issue:42915", + "right": "issue:45084", + "accept": false, + "reason": "FP8 config failure and template compilation error are unrelated." + }, + { + "left": "issue:43994", + "right": "issue:44610", + "accept": false, + "reason": "Both vision-related, but different models and root causes." + }, + { + "left": "issue:43377", + "right": "issue:43819", + "accept": false, + "reason": "Both audio-related, but different models and distinct defects." + }, + { + "left": "issue:44220", + "right": "issue:44683", + "accept": false, + "reason": "Feature extraction bug vs torch-version flex_attention compile failure; unrelated." + }, + { + "left": "issue:44265", + "right": "issue:44610", + "accept": false, + "reason": "torch.export/check logic and processor image-size mismatch are different bugs." + }, + { + "left": "issue:44451", + "right": "issue:45081", + "accept": false, + "reason": "Different loading problems: ScandiBERT load failure vs Mistral regex crash." + }, + { + "left": "issue:44987", + "right": "issue:45081", + "accept": false, + "reason": "Remote-code loading regression and tokenizer regex crash do not share the same code path." + }, + { + "left": "issue:42915", + "right": "issue:44291", + "accept": false, + "reason": "Both are v5-related, but the failures are distinct: FP8 config vs unexpected _is_hf_initialized arg." + }, + { + "left": "issue:45216", + "right": "issue:45357", + "accept": true, + "reason": "Same Qwen3.5 save_pretrained regression; the later issue narrows the same underlying wrong-checkpoint/keys problem to visual encoder keys." + }, + { + "left": "issue:44062", + "right": "issue:44987", + "accept": false, + "reason": "AddedToken special-argument bug and remote_code loading failure are unrelated." + }, + { + "left": "issue:43632", + "right": "issue:44756", + "accept": false, + "reason": "Different problems: _is_hf_initialized flag handling vs mmap/OOM behavior." + }, + { + "left": "issue:36032", + "right": "issue:43742", + "accept": false, + "reason": "Tokenizer special-token conflict and MobileLLM key error are different bugs." + }, + { + "left": "issue:42915", + "right": "issue:44589", + "accept": false, + "reason": "Different FP8/storage issues with no shared failing path." + }, + { + "left": "issue:44479", + "right": "issue:45325", + "accept": false, + "reason": "Both touch Qwen vision models, but video-input regression and still-image rope indexing are different defects." + }, + { + "left": "issue:43404", + "right": "issue:45127", + "accept": false, + "reason": "Both involve tied weights, but one is missing tying in the model and the other is LoRA-merge collapse after vocab extension." + }, + { + "left": "issue:43296", + "right": "issue:43531", + "accept": false, + "reason": "Different model families and different failures." + }, + { + "left": "issue:36032", + "right": "issue:41720", + "accept": false, + "reason": "Tokenizer load conflict and CUDA assert from auto device mapping are unrelated." + }, + { + "left": "issue:40444", + "right": "issue:45325", + "accept": false, + "reason": "Multiple-image dataset failure and Qwen2.5-VL rope scaling bug are different issues." + }, + { + "left": "issue:43531", + "right": "issue:44877", + "accept": false, + "reason": "Sliding-window bug and granite_speech config loading are unrelated." + }, + { + "left": "issue:43611", + "right": "issue:44534", + "accept": false, + "reason": "Base-model-prefix loading regression and junk-filled non-persistent buffers are different v5 bugs." + }, + { + "left": "issue:43425", + "right": "issue:44403", + "accept": false, + "reason": "Torch compatibility issue and loading-noise issue are unrelated." + }, + { + "left": "issue:40990", + "right": "issue:44811", + "accept": false, + "reason": "Perplexity regression and Whisper batch_decode skip_special_tokens bug are unrelated." + }, + { + "left": "issue:43792", + "right": "issue:44589", + "accept": false, + "reason": "Whisper runtime failure and Float8 storage lookup error are different code paths." + }, + { + "left": "issue:43632", + "right": "issue:44534", + "accept": false, + "reason": "Distinct v5 regressions: _is_hf_initialized handling vs buffer serialization." + }, + { + "left": "issue:44079", + "right": "issue:45663", + "accept": false, + "reason": "ModelOutput key assignment and Gemma4 FSDP2 attention KeyError are unrelated." + }, + { + "left": "issue:41553", + "right": "issue:43582", + "accept": false, + "reason": "Bad AutoTokenizer error message and AppleSilicon warmup TypeError are unrelated." + }, + { + "left": "issue:44683", + "right": "issue:45020", + "accept": false, + "reason": "flex_attention compile failure and remote_code loading regression are unrelated." + }, + { + "left": "issue:42175", + "right": "issue:45542", + "accept": false, + "reason": "Both concern TensorFlow packaging/backend availability, but they describe different failure paths and are not clearly the same bug." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet: most items are clearly different bugs by model, failure mode, or code path, and the single accepted Qwen3.5 save_pretrained pair looks plausibly mergeable as one regression. The rejected soft-edge verdicts are generally conservative rather than overclaiming duplicates." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 10, + "review_unit_count": 49, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-426", + "nodes": [ + "issue:33666", + "issue:35707", + "issue:40444", + "issue:41720", + "issue:42175", + "issue:42915", + "issue:43296", + "issue:43317", + "issue:43377", + "issue:43381", + "issue:43421", + "issue:43454", + "issue:43482", + "issue:43525", + "issue:43531", + "issue:43611", + "issue:43632", + "issue:43644", + "issue:43646", + "issue:43749", + "issue:43792", + "issue:43818", + "issue:43856", + "issue:43957", + "issue:43994", + "issue:44079", + "issue:44265", + "issue:44291", + "issue:44403", + "issue:44462", + "issue:44466", + "issue:44534", + "issue:44568", + "issue:44589", + "issue:44829", + "issue:44898", + "issue:44991", + "issue:45072", + "issue:45081", + "issue:45325", + "issue:45357", + "issue:45381", + "issue:45479", + "issue:45491", + "issue:45538", + "issue:45812", + "issue:45834", + "issue:45923" + ], + "soft_pairs": [ + "issue:44403|issue:44991", + "issue:33666|issue:40444", + "issue:43421|issue:44568", + "issue:35707|issue:40444", + "issue:43818|issue:45357", + "issue:44462|issue:45812", + "issue:43381|issue:45923", + "issue:44291|issue:44589", + "issue:43482|issue:45538", + "issue:43317|issue:43856", + "issue:44829|issue:45479", + "issue:43482|issue:43525", + "issue:44079|issue:44534", + "issue:43531|issue:44291", + "issue:43377|issue:45491", + "issue:41720|issue:45081", + "issue:40444|issue:45381", + "issue:43749|issue:44079", + "issue:44265|issue:45072", + "issue:43644|issue:44079", + "issue:42175|issue:43296", + "issue:45325|issue:45381", + "issue:43632|issue:45072", + "issue:43792|issue:43957", + "issue:43611|issue:43646", + "issue:43994|issue:44898", + "issue:45325|issue:45834", + "issue:43454|issue:44466", + "issue:42915|issue:43482" + ], + "prepared_review_unit_hash": "1bbb2b861515981d502d518182f5c5f71915a50198214d3202a05ffaa12b5439", + "component_budget": { + "node_count": 426, + "item_count": 426, + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36625, + "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 set of unrelated issue reports. The only plausible duplicate-like pair is the Transformers v5 model-initialization/loading regression pair (43611/43646); the rest are distinct bugs across different models, features, and failure modes.", + "confidence": 0.68, + "canonical_issue_reason": "Issue 43646 is the broadest and most central of the near-duplicate loading regressions: it describes a Transformers v5 custom model initialization break, which likely subsumes the more specific base_model_prefix loading symptom in 43611.", + "canonical_pr_reason": null, + "best_issue_reason": "43646 is the best representative issue in this set because it captures the underlying v5 model initialization regression at a higher level than the narrower 43611 report.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44403", + "right": "issue:44991", + "accept": false, + "reason": "Both involve tokenizer/model loading, but one is generic loading noise and the other is a specific tokenizer failure for EMBEDDIA/est-roberta; different bugs." + }, + { + "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 IterableDataset/multi-image finetuning; not the same failure mode." + }, + { + "left": "issue:43421", + "right": "issue:44568", + "accept": false, + "reason": "Both touch special-token handling, but one is a runtime post-processor update issue and the other is add_special_tokens not inserting BOS/EOS; distinct code paths." + }, + { + "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:43818", + "right": "issue:45357", + "accept": false, + "reason": "Different models and different defects: missing temporal attention/weight sharing vs incorrect save_pretrained visual encoder keys." + }, + { + "left": "issue:44462", + "right": "issue:45812", + "accept": false, + "reason": "Both are tokenizer-related, but one ignores tokenizer.json while the other produces wrong IDs for Granite; not the same bug." + }, + { + "left": "issue:43381", + "right": "issue:45923", + "accept": false, + "reason": "One is a usage restriction for gradient checkpointing in eval mode; the other is a missing supports_gradient_checkpointing flag on a remote-code model." + }, + { + "left": "issue:44291", + "right": "issue:44589", + "accept": false, + "reason": "Both are loading regressions in v5, but one is an init_empty_weights argument mismatch and the other is a Float8 storage lookup error; different concrete failures." + }, + { + "left": "issue:43482", + "right": "issue:45538", + "accept": false, + "reason": "Completely different areas: Qwen2.5-GGUF loading under v5 vs CLIPTokenizer model_max_length behavior." + }, + { + "left": "issue:43317", + "right": "issue:43856", + "accept": false, + "reason": "Device-map/offload loading failure is unrelated to Qwen3 MoE memory usage during training." + }, + { + "left": "issue:44829", + "right": "issue:45479", + "accept": false, + "reason": "Both describe degenerate training behavior, but one is tied to flash_attention_3 and the other to problem_type=num_labels=1 classification loss handling." + }, + { + "left": "issue:43482", + "right": "issue:43525", + "accept": false, + "reason": "Different models and symptoms: GGUF loading failure vs missing pad_token_id on Llama4Config." + }, + { + "left": "issue:44079", + "right": "issue:44534", + "accept": false, + "reason": "ModelOutput key assignment and non-persistent buffer initialization are separate internal issues." + }, + { + "left": "issue:43531", + "right": "issue:44291", + "accept": false, + "reason": "Qwen3-MoE sliding_window behavior is unrelated to init_empty_weights/_is_hf_initialized constructor loading errors." + }, + { + "left": "issue:43377", + "right": "issue:45491", + "accept": false, + "reason": "Both involve batching/padding, but one is a MIMI encoder padding-mask bug and the other is a Gemma3 sliding-window/all-padding-window NaN issue." + }, + { + "left": "issue:41720", + "right": "issue:45081", + "accept": false, + "reason": "Qwen3 auto device-mapping cudaErrorAssert and Mistral tokenizer regex patch crashes are unrelated." + }, + { + "left": "issue:40444", + "right": "issue:45381", + "accept": false, + "reason": "Both are Qwen2.5-VL issues, but one is multi-image finetuning and the other is video vision_position_ids; different scenarios and likely different fixes." + }, + { + "left": "issue:43749", + "right": "issue:44079", + "accept": false, + "reason": "FSDP CPU RAM efficient loading and ModelOutput key assignment are unrelated." + }, + { + "left": "issue:44265", + "right": "issue:45072", + "accept": false, + "reason": "torch.export/export-time checks and bfloat16 dtype mismatches in inference are distinct problems." + }, + { + "left": "issue:43644", + "right": "issue:44079", + "accept": false, + "reason": "Non-persistent buffer junk fill and ModelOutput key assignment are separate regressions." + }, + { + "left": "issue:42175", + "right": "issue:43296", + "accept": false, + "reason": "Packaging/backends issue vs a specific PaddleOCR-VL model loading failure; no shared bug." + }, + { + "left": "issue:45325", + "right": "issue:45381", + "accept": false, + "reason": "Both are Qwen2.5-VL position-id issues, but one is still-image temporal scaling and the other is video vision_position_ids; too different to merge confidently." + }, + { + "left": "issue:43632", + "right": "issue:45072", + "accept": false, + "reason": "_is_hf_initialized breakage and bfloat16 dtype mismatch are unrelated Transformers v5 regressions." + }, + { + "left": "issue:43792", + "right": "issue:43957", + "accept": false, + "reason": "Whisper model runtime failure and torch.device('meta') loading regressions are not the same bug." + }, + { + "left": "issue:43611", + "right": "issue:43646", + "accept": true, + "reason": "These look like the same Transformers v5 model-loading/init regression: one reports breakage for models with base_model_prefix, the other for custom model initialization. A single fix to initialization/loading could plausibly address both." + }, + { + "left": "issue:43994", + "right": "issue:44898", + "accept": false, + "reason": "Different vision-model correctness issues: SigLIP2 nonsensical outputs vs Perceiver non-default resolution failure." + }, + { + "left": "issue:45325", + "right": "issue:45834", + "accept": false, + "reason": "Both are vision/sequence-length related, but one is a Qwen2.5-VL position-id scaling bug and the other is a Kosmos2.5 long-OCR index error." + }, + { + "left": "issue:43454", + "right": "issue:44466", + "accept": false, + "reason": "Both mention lm_head weight handling, but one is a tied-weights bug causing bad generation and the other is device-dependent serialization inconsistency." + }, + { + "left": "issue:42915", + "right": "issue:43482", + "accept": false, + "reason": "Qwen3Moe FineGrainedFP8Config failure and Qwen2.5-GGUF loading failure are unrelated." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The result is grounded in the packet and the soft-edge judgments are conservative overall. The single accepted pair, 43611/43646, is plausibly the same Transformers v5 loading/initialization regression, while the other pairs are rejected with appropriately narrow reasoning rather than overstating duplicate evidence." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 11, + "review_unit_count": 49, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-426", + "nodes": [ + "issue:34634", + "issue:35707", + "issue:39692", + "issue:40444", + "issue:40990", + "issue:41720", + "issue:42947", + "issue:43257", + "issue:43262", + "issue:43296", + "issue:43317", + "issue:43322", + "issue:43377", + "issue:43454", + "issue:43526", + "issue:43550", + "issue:43572", + "issue:43606", + "issue:43701", + "issue:43720", + "issue:43746", + "issue:43749", + "issue:43819", + "issue:43873", + "issue:43881", + "issue:43883", + "issue:44186", + "issue:44265", + "issue:44291", + "issue:44479", + "issue:44488", + "issue:44534", + "issue:44568", + "issue:44610", + "issue:44877", + "issue:44898", + "issue:44991", + "issue:45072", + "issue:45325", + "issue:45584", + "issue:45701", + "issue:45759", + "issue:45800", + "issue:45823", + "issue:45920", + "issue:45923", + "issue:46050" + ], + "soft_pairs": [ + "issue:43296|issue:45800", + "issue:44479|issue:45823", + "issue:43701|issue:45325", + "issue:44291|issue:44877", + "issue:43881|issue:44488", + "issue:43606|issue:44265", + "issue:43262|issue:43746", + "issue:43873|issue:45701", + "issue:40990|issue:45584", + "issue:34634|issue:35707", + "issue:44568|issue:45920", + "issue:35707|issue:41720", + "issue:44186|issue:44265", + "issue:43606|issue:43819", + "issue:41720|issue:45823", + "issue:43454|issue:43550", + "issue:39692|issue:40444", + "issue:43377|issue:44610", + "issue:43454|issue:45072", + "issue:43317|issue:43873", + "issue:43572|issue:45759", + "issue:43377|issue:44898", + "issue:43257|issue:45923", + "issue:40990|issue:42947", + "issue:43322|issue:46050", + "issue:43526|issue:43883", + "issue:43454|issue:43720", + "issue:41720|issue:43322", + "issue:43881|issue:44991", + "issue:43296|issue:44877", + "issue:43749|issue:44534" + ], + "prepared_review_unit_hash": "fc2a39c1128886c918f362a1d4971d109f3717d049f4a29ceaee7f1ea28e3ca5", + "component_budget": { + "node_count": 426, + "item_count": 426, + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 31, + "serialized_chars": 36541, + "estimated_input_tokens": 9136, + "estimated_eval_tokens": 18528 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Most items are distinct model-specific regressions, not duplicates. The only plausible duplicate family is offloading/quantization loading behavior, centered on 43873; the rest span unrelated tokenizer, multimodal, config, export, and compile bugs.", + "confidence": 0.81, + "canonical_issue_reason": "43873 is the broadest and most central issue in the only overlapping area here: offloading/quantization loading failures. It has the strongest discussion/reference signal and can anchor this cluster.", + "canonical_pr_reason": null, + "best_issue_reason": "43873 is the best representative issue because it is open, broad, and closest to the only near-duplicate pair in the set (43317).", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43296", + "right": "issue:45800", + "accept": false, + "reason": "Different failures: a model/processor loading problem in vLLM vs a general torch/transformers version incompatibility." + }, + { + "left": "issue:44479", + "right": "issue:45823", + "accept": false, + "reason": "Both involve multimodal models, but the concrete bugs differ: video-input regression vs device-mismatch during forward." + }, + { + "left": "issue:43701", + "right": "issue:45325", + "accept": false, + "reason": "Unrelated code paths: checkpoint key mismatch vs Qwen2.5-VL RoPE/position-id scaling." + }, + { + "left": "issue:44291", + "right": "issue:44877", + "accept": false, + "reason": "Different config/load errors: unexpected init arg vs strict config blocking granite_speech loading." + }, + { + "left": "issue:43881", + "right": "issue:44488", + "accept": false, + "reason": "Separate model-loading failures with no clear shared root cause." + }, + { + "left": "issue:43606", + "right": "issue:44265", + "accept": false, + "reason": "CPU-offload device mismatch and torch.export failure are unrelated bugs." + }, + { + "left": "issue:43262", + "right": "issue:43746", + "accept": false, + "reason": "Audio chat-template sampling-rate default and GraniteSpeech PEFT checkpoint loading are different problems." + }, + { + "left": "issue:43873", + "right": "issue:45701", + "accept": false, + "reason": "Offloading/quantization behavior is unrelated to a tokenizer-version regression." + }, + { + "left": "issue:40990", + "right": "issue:45584", + "accept": false, + "reason": "Different symptoms and subsystems: perplexity regression vs Whisper empty-transcription generation failure." + }, + { + "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:44568", + "right": "issue:45920", + "accept": false, + "reason": "Both are tokenizer regressions, but one is a single tokenizer special-token bug and the other is broader wrong-token-id behavior across multiple models." + }, + { + "left": "issue:35707", + "right": "issue:41720", + "accept": false, + "reason": "Progressive generation bug and Qwen3 device-map CUDA assert are different failures." + }, + { + "left": "issue:44186", + "right": "issue:44265", + "accept": false, + "reason": "Tokenizer padding/truncation crash and torch.export/torch_compilable_check are unrelated." + }, + { + "left": "issue:43606", + "right": "issue:43819", + "accept": false, + "reason": "Device-mismatch on offload vs DAC latent/STE mismatch are different code paths." + }, + { + "left": "issue:41720", + "right": "issue:45823", + "accept": false, + "reason": "Both mention device_map issues, but they affect different models and likely different underlying bugs." + }, + { + "left": "issue:43454", + "right": "issue:43550", + "accept": false, + "reason": "Weight-tying bug and torch.compile/SDPA failure are unrelated." + }, + { + "left": "issue:39692", + "right": "issue:40444", + "accept": false, + "reason": "Documentation/model-processor mismatch and IterableDataset multi-image fine-tuning failure are distinct." + }, + { + "left": "issue:43377", + "right": "issue:44610", + "accept": false, + "reason": "MIMI padding-mask inconsistency and OmDet-Turbo image-size mismatch are unrelated." + }, + { + "left": "issue:43454", + "right": "issue:45072", + "accept": false, + "reason": "Different model bugs: tied weights vs dtype mismatch in inference." + }, + { + "left": "issue:43317", + "right": "issue:43873", + "accept": true, + "reason": "Same offloading/quantization loading family: both describe device_map/offload failures when loading quantized or dequantized models, with overlapping accelerate loading behavior." + }, + { + "left": "issue:43572", + "right": "issue:45759", + "accept": false, + "reason": "A missing config field and an AutoModel text_config unwrapping bug are different API regressions." + }, + { + "left": "issue:43377", + "right": "issue:44898", + "accept": false, + "reason": "Padding-mask batch inconsistency and Perceiver resolution/interpolation failure are unrelated." + }, + { + "left": "issue:43257", + "right": "issue:45923", + "accept": false, + "reason": "Accelerate+deepspeed MOE conversion issue and missing gradient-checkpointing flag are separate problems." + }, + { + "left": "issue:40990", + "right": "issue:42947", + "accept": false, + "reason": "High perplexity and ineffective gradient checkpointing are different training/inference bugs." + }, + { + "left": "issue:43322", + "right": "issue:46050", + "accept": false, + "reason": "Llava loading segfault and sinq save/load quantization breakage are unrelated." + }, + { + "left": "issue:43526", + "right": "issue:43883", + "accept": false, + "reason": "BeitImageProcessorFast label reduction and Molmo tied-weights attribute error are unrelated." + }, + { + "left": "issue:43454", + "right": "issue:43720", + "accept": false, + "reason": "AyaVision weight tying and BitNet packed-weight unpacking are different loading/model-init bugs." + }, + { + "left": "issue:41720", + "right": "issue:43322", + "accept": false, + "reason": "Both are loading-time failures, but they affect different model families and likely different root causes." + }, + { + "left": "issue:43881", + "right": "issue:44991", + "accept": false, + "reason": "Model loading failure and tokenizer loading failure are not the same bug." + }, + { + "left": "issue:43296", + "right": "issue:44877", + "accept": false, + "reason": "vLLM model load failure and strict config rejection are different issues." + }, + { + "left": "issue:43749", + "right": "issue:44534", + "accept": false, + "reason": "FSDP CPU-efficient loading breakage and non-persistent buffer initialization junk are not the same underlying defect." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet and stays conservative overall. The single accepted pair (43317, 43873) is a plausible overlap in offloading/quantization loading behavior, and the rejected pairs are mostly distinguished at a high level by title/topic. No obvious overstatement stands out." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 12, + "review_unit_count": 49, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-426", + "nodes": [ + "issue:36010", + "issue:36331", + "issue:40990", + "issue:41720", + "issue:42617", + "issue:42915", + "issue:43295", + "issue:43316", + "issue:43317", + "issue:43404", + "issue:43450", + "issue:43526", + "issue:43531", + "issue:43550", + "issue:43572", + "issue:43577", + "issue:43606", + "issue:43708", + "issue:43720", + "issue:43873", + "issue:43881", + "issue:44008", + "issue:44186", + "issue:44387", + "issue:44423", + "issue:44466", + "issue:44479", + "issue:44610", + "issue:44671", + "issue:44743", + "issue:44756", + "issue:44877", + "issue:44898", + "issue:44913", + "issue:44918", + "issue:44977", + "issue:44991", + "issue:45072", + "issue:45081", + "issue:45161", + "issue:45357", + "issue:45406", + "issue:45632", + "issue:45684", + "issue:45701", + "issue:45704", + "issue:45812", + "issue:45823" + ], + "soft_pairs": [ + "issue:44466|issue:45357", + "issue:41720|issue:44977", + "issue:42617|issue:42915", + "issue:40990|issue:45701", + "issue:44008|issue:45823", + "issue:43404|issue:45357", + "issue:44387|issue:45704", + "issue:44186|issue:45072", + "issue:36331|issue:42617", + "issue:43577|issue:43708", + "issue:44877|issue:44991", + "issue:36331|issue:42915", + "issue:43873|issue:44756", + "issue:43450|issue:45701", + "issue:43550|issue:43720", + "issue:44423|issue:45406", + "issue:45632|issue:45684", + "issue:41720|issue:42617", + "issue:44479|issue:44918", + "issue:44610|issue:44898", + "issue:43550|issue:43606", + "issue:44186|issue:44610", + "issue:42915|issue:43317", + "issue:43295|issue:45701", + "issue:41720|issue:45161", + "issue:44671|issue:44743", + "issue:43526|issue:43881", + "issue:36010|issue:42617", + "issue:45081|issue:45812", + "issue:43531|issue:43572", + "issue:43316|issue:44913" + ], + "prepared_review_unit_hash": "291d53af0f91b8a200c5d1c4597cc1ad7d6524e66b7879a851c96e49923f7de9", + "component_budget": { + "node_count": 426, + "item_count": 426, + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 37223, + "estimated_input_tokens": 9306, + "estimated_eval_tokens": 18868 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These items form a loose grab-bag of Transformers bugs/regressions, but the soft pairs mostly point to different concrete code paths. I do not see any safe duplicate merges among the candidate pairs.", + "confidence": 0.64, + "canonical_issue_reason": "issue:43873 is the best umbrella representative: it is a concrete, open regression about offloading/quantization/device placement, has substantial discussion, and matches the strongest recurring theme across the set without being overly narrow.", + "canonical_pr_reason": null, + "best_issue_reason": "Among the issues, issue:43873 is the most suitable anchor for triage because it is specific enough to be actionable while still broad enough to represent the offloading/quantization/device-mapping subgroup in this cluster.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44466", + "right": "issue:45357", + "accept": false, + "reason": "Different bugs: tied lm_head serialization vs visual encoder key persistence in Qwen3.5." + }, + { + "left": "issue:41720", + "right": "issue:44977", + "accept": false, + "reason": "Same model family, but different failures and code paths: cuda assert on auto device map vs flash-attention generation regression." + }, + { + "left": "issue:42617", + "right": "issue:42915", + "accept": false, + "reason": "Unrelated failures: 3d_parallel.py execution vs Qwen3Moe FP8 config loading." + }, + { + "left": "issue:40990", + "right": "issue:45701", + "accept": false, + "reason": "Different scope: perplexity on one model/dataset vs broad tokenizer-version behavior change." + }, + { + "left": "issue:44008", + "right": "issue:45823", + "accept": false, + "reason": "Different Gemma generations and different symptoms: audio mask AttributeError vs PLE device mismatch." + }, + { + "left": "issue:43404", + "right": "issue:45357", + "accept": false, + "reason": "Different models and root causes: Mistral3 weight tying vs Qwen3.5 visual encoder save/load keys." + }, + { + "left": "issue:44387", + "right": "issue:45704", + "accept": false, + "reason": "Both mention memory, but one is int4 quantization OOM while the other is a T5/apex RMSNorm leak." + }, + { + "left": "issue:44186", + "right": "issue:45072", + "accept": false, + "reason": "Tokenizer crash vs dtype mismatch; different subsystems and failure modes." + }, + { + "left": "issue:36331", + "right": "issue:42617", + "accept": false, + "reason": "Custom trainer API breakage is unrelated to the 3d_parallel.py runtime failure." + }, + { + "left": "issue:43577", + "right": "issue:43708", + "accept": false, + "reason": "Different trainer/model issues: dtype remaining float32 vs checkpoint resume max_steps miscalculation." + }, + { + "left": "issue:44877", + "right": "issue:44991", + "accept": false, + "reason": "Different tokenizer/config problems: strict config rejection vs tokenizer loading failure." + }, + { + "left": "issue:36331", + "right": "issue:42915", + "accept": false, + "reason": "Trainer loss signature change is unrelated to Qwen3Moe FP8 config failure." + }, + { + "left": "issue:43873", + "right": "issue:44756", + "accept": false, + "reason": "Both involve memory/offload concerns, but one is quantization offloading behavior and the other is mmap/OOM on Strix Halo." + }, + { + "left": "issue:43450", + "right": "issue:45701", + "accept": false, + "reason": "Video batching shape bug is unrelated to a tokenization version regression." + }, + { + "left": "issue:43550", + "right": "issue:43720", + "accept": false, + "reason": "Different code paths: torch.compile+SDPA vs accelerate loading of packed BitNet weights." + }, + { + "left": "issue:44423", + "right": "issue:45406", + "accept": false, + "reason": "Serve crashes from different causes: multimodal tensor handling vs Gemma4Processor missing _tokenizer." + }, + { + "left": "issue:45632", + "right": "issue:45684", + "accept": false, + "reason": "Cache-path collision and file permission propagation are distinct bugs." + }, + { + "left": "issue:41720", + "right": "issue:42617", + "accept": false, + "reason": "Different failures: Qwen3 auto device mapping assert vs 3d_parallel.py runtime error." + }, + { + "left": "issue:44479", + "right": "issue:44918", + "accept": false, + "reason": "Both are Qwen-related, but one is a video-input regression and the other is an SFT embedding-unpack failure." + }, + { + "left": "issue:44610", + "right": "issue:44898", + "accept": false, + "reason": "Both concern image sizes, but the concrete model expectations and bug mechanisms differ." + }, + { + "left": "issue:43550", + "right": "issue:43606", + "accept": false, + "reason": "Torch.compile/SDPA issue is unrelated to bark CPU-offload device mismatch." + }, + { + "left": "issue:44186", + "right": "issue:44610", + "accept": false, + "reason": "Tokenizer crash and processor output-size mismatch are different bugs." + }, + { + "left": "issue:42915", + "right": "issue:43317", + "accept": false, + "reason": "FP8 config failure is distinct from dequantized model loading with gpu+cpu offload." + }, + { + "left": "issue:43295", + "right": "issue:45701", + "accept": false, + "reason": "Processor.tokenizer/image handling regression is not the same as general tokenization drift." + }, + { + "left": "issue:41720", + "right": "issue:45161", + "accept": false, + "reason": "Different model families and failures: Qwen3 auto device-map assert vs GPT-OSS tensor-parallel issue." + }, + { + "left": "issue:44671", + "right": "issue:44743", + "accept": false, + "reason": "CamemBERT MLM regression and Qwen3.5 recurrent-state reset are unrelated." + }, + { + "left": "issue:43526", + "right": "issue:43881", + "accept": false, + "reason": "ImageProcessor label reduction bug is unrelated to glm-4v loading failure." + }, + { + "left": "issue:36010", + "right": "issue:42617", + "accept": false, + "reason": "Importing GenerationMixin and running 3d_parallel.py are unrelated issues." + }, + { + "left": "issue:45081", + "right": "issue:45812", + "accept": false, + "reason": "Both are tokenizer-related, but one is a Mistral regex patch crash and the other is wrong Granite token IDs; different root causes." + }, + { + "left": "issue:43531", + "right": "issue:43572", + "accept": false, + "reason": "Qwen3-MoE sliding-window behavior and StableLmConfig pad_token_idx are different regressions." + }, + { + "left": "issue:43316", + "right": "issue:44913", + "accept": false, + "reason": "Gemma3TextConfig API mismatch and GPTNeoXConfig rotary_pct reload behavior are distinct config bugs." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 13, + "review_unit_count": 49, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-426", + "nodes": [ + "issue:29159", + "issue:30064", + "issue:33357", + "issue:34567", + "issue:34634", + "issue:35707", + "issue:36032", + "issue:41720", + "issue:42490", + "issue:42915", + "issue:43322", + "issue:43329", + "issue:43388", + "issue:43421", + "issue:43531", + "issue:43540", + "issue:43550", + "issue:43606", + "issue:43653", + "issue:43720", + "issue:43746", + "issue:43782", + "issue:43873", + "issue:43927", + "issue:44077", + "issue:44186", + "issue:44361", + "issue:44451", + "issue:44492", + "issue:44509", + "issue:44514", + "issue:44545", + "issue:44589", + "issue:44625", + "issue:44898", + "issue:44991", + "issue:45059", + "issue:45072", + "issue:45381", + "issue:45406", + "issue:45464", + "issue:45479", + "issue:45701", + "issue:45704", + "issue:45759", + "issue:45812", + "issue:45859" + ], + "soft_pairs": [ + "issue:43540|issue:43782", + "issue:43421|issue:45812", + "issue:42915|issue:43322", + "issue:44186|issue:44361", + "issue:29159|issue:36032", + "issue:43388|issue:45059", + "issue:45406|issue:45464", + "issue:33357|issue:35707", + "issue:30064|issue:34634", + "issue:41720|issue:43746", + "issue:43606|issue:43720", + "issue:43550|issue:43653", + "issue:43421|issue:45701", + "issue:43873|issue:45704", + "issue:44492|issue:44509", + "issue:43606|issue:43653", + "issue:34634|issue:41720", + "issue:42490|issue:43329", + "issue:43531|issue:45759", + "issue:34567|issue:35707", + "issue:44514|issue:44545", + "issue:43421|issue:43927", + "issue:44625|issue:45479", + "issue:44451|issue:44991", + "issue:44898|issue:45072", + "issue:43531|issue:44589", + "issue:43782|issue:45381", + "issue:44077|issue:44625", + "issue:45759|issue:45859" + ], + "prepared_review_unit_hash": "91fa3fe5332025ec4ec6b8b8aef858db88fdc8236599b96219e6a202adf6b460", + "component_budget": { + "node_count": 426, + "item_count": 426, + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 29, + "serialized_chars": 35968, + "estimated_input_tokens": 8992, + "estimated_eval_tokens": 18240 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a false-positive similarity set spanning unrelated models, processors, and failure modes. The only clear duplicate is the Qwen2_5_VLProcessor.apply_chat_template crash pair (44514/44545), which appears to be the same issue/same crash signature with the same padding=False condition.", + "confidence": 0.91, + "canonical_issue_reason": "Issue 44514 is the best canonical candidate because 44545 is the same Qwen2_5_VLProcessor.apply_chat_template crash report with the same padding=False condition; the rest of the cluster is heterogeneous.", + "canonical_pr_reason": "No pull requests are present in this cluster.", + "best_issue_reason": "44514 is the cleanest representative of the only true duplicate pair; it matches 44545 on the same underlying apply_chat_template crash signature.", + "best_pr_reason": "No pull requests are present in this cluster.", + "soft_edge_verdicts": [ + { + "left": "issue:43540", + "right": "issue:43782", + "accept": false, + "reason": "Different failure surfaces: Qwen3OmniMoe video processing ValueError vs Qwen3VL from_pretrained weight_only load error." + }, + { + "left": "issue:43421", + "right": "issue:45812", + "accept": false, + "reason": "Both concern tokenization, but one is runtime post-processor updates and the other is a Granite model token-ID regression." + }, + { + "left": "issue:42915", + "right": "issue:43322", + "accept": false, + "reason": "Unrelated models and bugs: FineGrainedFP8Config failure for Qwen3Moe vs Llava Next segmentation fault." + }, + { + "left": "issue:44186", + "right": "issue:44361", + "accept": false, + "reason": "Different tokenizers and symptoms: LayoutLMv2Tokenizer batching/NER crash vs MLukeTokenizer AttributeError." + }, + { + "left": "issue:29159", + "right": "issue:36032", + "accept": false, + "reason": "Tokenizer inconsistency report vs a T5Tokenizer name conflict on load; not the same bug." + }, + { + "left": "issue:43388", + "right": "issue:45059", + "accept": false, + "reason": "Metric gathering drops labels in the last batch, while SAM3 text/bbox behavior is a separate multimodal input issue." + }, + { + "left": "issue:45406", + "right": "issue:45464", + "accept": false, + "reason": "Gemma4Processor missing _tokenizer is unrelated to Qwen3.5-0.8B chat/completions streaming failures." + }, + { + "left": "issue:33357", + "right": "issue:35707", + "accept": false, + "reason": "MacOS CLIP bus error and progressive generation with inputs_embeds/past_key_values are different code paths." + }, + { + "left": "issue:30064", + "right": "issue:34634", + "accept": false, + "reason": "Void segmentation-map processing and BarkProcessor voice_preset are unrelated processor bugs." + }, + { + "left": "issue:41720", + "right": "issue:43746", + "accept": false, + "reason": "Qwen3 auto device mapping cuda assert is different from GraniteSpeech PEFT local-checkpoint loading." + }, + { + "left": "issue:43606", + "right": "issue:43720", + "accept": false, + "reason": "Both mention loading/offload, but CPU offload device mismatch and BitNet packed-weight unpacking are different concrete failures." + }, + { + "left": "issue:43550", + "right": "issue:43653", + "accept": false, + "reason": "torch.compile/SDPA bug is unrelated to BigBirdTokenizer special-token registration." + }, + { + "left": "issue:43421", + "right": "issue:45701", + "accept": false, + "reason": "Runtime special-token/post-processor syncing is not the same as a broad version-dependent tokenization change report." + }, + { + "left": "issue:43873", + "right": "issue:45704", + "accept": false, + "reason": "Quantization/offloading behavior is unrelated to the T5 apex FusedRMSNorm memory leak." + }, + { + "left": "issue:44492", + "right": "issue:44509", + "accept": false, + "reason": "A typo in cache strategies is unrelated to outdated docs about removed pipeline tasks." + }, + { + "left": "issue:43606", + "right": "issue:43653", + "accept": false, + "reason": "CPU offload device mismatch and tokenizer special-token registration are different bugs." + }, + { + "left": "issue:34634", + "right": "issue:41720", + "accept": false, + "reason": "BarkProcessor voice_preset and Qwen3 auto device mapping are unrelated." + }, + { + "left": "issue:42490", + "right": "issue:43329", + "accept": false, + "reason": "Local-path import warning is unrelated to the undefined video-branch helpers in multimodal token counting." + }, + { + "left": "issue:43531", + "right": "issue:45759", + "accept": false, + "reason": "Qwen3-MoE sliding_window behavior and AutoModelForCausalLM.from_config text_config unwrapping are different issues." + }, + { + "left": "issue:34567", + "right": "issue:35707", + "accept": false, + "reason": "TrainerState token counter not updating is unrelated to progressive generation with inputs_embeds/past_key_values." + }, + { + "left": "issue:44514", + "right": "issue:44545", + "accept": true, + "reason": "Same Qwen2_5_VLProcessor.apply_chat_template crash signature with the same padding=False condition; these are duplicate reports of the same bug." + }, + { + "left": "issue:43421", + "right": "issue:43927", + "accept": false, + "reason": "Runtime post-processor updates and DiaConfig save/load token-ID loss are different tokenization/config issues." + }, + { + "left": "issue:44625", + "right": "issue:45479", + "accept": false, + "reason": "num_labels propagation in Qwen3.5 is not the same as the single-label-classification zero-loss issue across models." + }, + { + "left": "issue:44451", + "right": "issue:44991", + "accept": false, + "reason": "Both are loading failures, but for different models and different root causes." + }, + { + "left": "issue:44898", + "right": "issue:45072", + "accept": false, + "reason": "Perceiver non-default resolution failure is unrelated to bfloat16 dtype mismatches in SwitchTransformers/TimmWrapperModel." + }, + { + "left": "issue:43531", + "right": "issue:44589", + "accept": false, + "reason": "Qwen3-MoE sliding_window bug is unrelated to the Float8 storage object lookup error." + }, + { + "left": "issue:43782", + "right": "issue:45381", + "accept": false, + "reason": "Qwen3VL weight_only load error and Qwen2.5-VL video position-id bug are different multimodal issues." + }, + { + "left": "issue:44077", + "right": "issue:44625", + "accept": false, + "reason": "patchtsmixer post_init validation and Qwen3.5 num_labels propagation do not share a code path." + }, + { + "left": "issue:45759", + "right": "issue:45859", + "accept": false, + "reason": "Composite config unwrapping and CPU-offload rotary-embedding incompatibility are unrelated." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "The overall clustering judgment is plausible, but the writeup is slightly overstated in a way that matters for precision: 44514 and 44545 do not have the exact same title text, only the same underlying bug wording. Please soften that claim to something like \"same issue/same crash signature\" rather than \"exact same title.\" The rest of the soft-edge rejections look conservative, and the no-PR note is grounded." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 14, + "review_unit_count": 49, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-426", + "nodes": [ + "issue:30064", + "issue:36010", + "issue:39692", + "issue:40444", + "issue:41720", + "issue:42886", + "issue:43381", + "issue:43421", + "issue:43531", + "issue:43540", + "issue:43596", + "issue:43618", + "issue:43645", + "issue:43653", + "issue:43697", + "issue:43720", + "issue:43746", + "issue:43819", + "issue:43844", + "issue:43873", + "issue:43994", + "issue:44008", + "issue:44062", + "issue:44186", + "issue:44220", + "issue:44368", + "issue:44451", + "issue:44610", + "issue:44661", + "issue:44671", + "issue:44843", + "issue:44918", + "issue:44936", + "issue:44987", + "issue:45042", + "issue:45081", + "issue:45183", + "issue:45356", + "issue:45357", + "issue:45479", + "issue:45588", + "issue:45663", + "issue:45704", + "issue:45759", + "issue:45859", + "issue:45907", + "issue:45923", + "issue:46050" + ], + "soft_pairs": [ + "issue:44987|issue:45042", + "issue:44368|issue:45357", + "issue:44062|issue:44451", + "issue:45588|issue:45759", + "issue:43746|issue:46050", + "issue:43421|issue:43653", + "issue:39692|issue:44610", + "issue:45704|issue:45923", + "issue:44918|issue:44936", + "issue:45663|issue:45923", + "issue:42886|issue:44843", + "issue:43381|issue:43844", + "issue:43421|issue:44186", + "issue:43531|issue:43645", + "issue:45704|issue:45859", + "issue:43994|issue:44671", + "issue:41720|issue:45704", + "issue:43596|issue:44220", + "issue:43746|issue:43873", + "issue:45081|issue:45356", + "issue:39692|issue:43873", + "issue:40444|issue:43746", + "issue:45356|issue:45907", + "issue:30064|issue:36010", + "issue:44220|issue:44661", + "issue:43720|issue:43819", + "issue:43653|issue:43720", + "issue:43540|issue:44008", + "issue:43618|issue:43697", + "issue:45183|issue:45479" + ], + "prepared_review_unit_hash": "183da652f83f324bb62c8ac9d4497d6e00f742faacdbb963fb9dda5320e6795d", + "component_budget": { + "node_count": 426, + "item_count": 426, + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 37220, + "estimated_input_tokens": 9305, + "estimated_eval_tokens": 18866 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These are mostly superficial text matches, not duplicate pairs. The issues span unrelated models, subsystems, and failure modes (tokenizers, quantization, offloading, vision processors, training, offline hub behavior), so I would not merge any soft pair as the same underlying bug/change.", + "confidence": 0.95, + "canonical_issue_reason": "issue:43873 is the strongest standalone anchor by activity and breadth, but the set is too heterogeneous to represent a true duplicate cluster.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43873 has the richest discussion/inbound activity and a concrete reproducible report, so it is the best representative issue here even though it is not a duplicate of the others.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44987", + "right": "issue:45042", + "accept": false, + "reason": "Different bugs: model loading regression vs image processor requiring torchvision." + }, + { + "left": "issue:44368", + "right": "issue:45357", + "accept": false, + "reason": "Unrelated Qwen3.5 warning about tie_word_embeddings vs wrong visual encoder keys on save." + }, + { + "left": "issue:44062", + "right": "issue:44451", + "accept": false, + "reason": "AddedToken keyword collision is a tokenizer API bug; ScandiBERT load failure is a separate model-loading issue." + }, + { + "left": "issue:45588", + "right": "issue:45759", + "accept": false, + "reason": "Flash-attention crash on s_aux=None is unrelated to multimodal config unwrapping for Qwen 3.5/3.6." + }, + { + "left": "issue:43746", + "right": "issue:46050", + "accept": false, + "reason": "PEFT adapter checkpoint loading and sinq quantization save/load are different code paths." + }, + { + "left": "issue:43421", + "right": "issue:43653", + "accept": false, + "reason": "Runtime post-processor syncing is not the same as BigBirdTokenizer special-token registration." + }, + { + "left": "issue:39692", + "right": "issue:44610", + "accept": false, + "reason": "Documentation/example errors for SigLIP2 are unrelated to OmDet-Turbo processor size mismatch." + }, + { + "left": "issue:45704", + "right": "issue:45923", + "accept": false, + "reason": "T5 apex RMSNorm leak is unrelated to a missing supports_gradient_checkpointing flag." + }, + { + "left": "issue:44918", + "right": "issue:44936", + "accept": false, + "reason": "TRL SFT unpacking failure is a different bug from trainer.evaluate() failing after train()." + }, + { + "left": "issue:45663", + "right": "issue:45923", + "accept": false, + "reason": "FSDP2 shared_kv_states rebuild bug is not the same as a missing gradient-checkpointing capability flag." + }, + { + "left": "issue:42886", + "right": "issue:44843", + "accept": false, + "reason": "Both involve offline hub behavior, but one is cache loading and the other is an unconditional model_info() call in a regex patch." + }, + { + "left": "issue:43381", + "right": "issue:43844", + "accept": false, + "reason": "Eval-mode gradient checkpointing restriction and ZeRO-3 gradient growth are distinct problems." + }, + { + "left": "issue:43421", + "right": "issue:44186", + "accept": false, + "reason": "Special-token post-processor syncing does not match LayoutLMv2 NER/batched padding crashes." + }, + { + "left": "issue:43531", + "right": "issue:43645", + "accept": false, + "reason": "Qwen3-MoE sliding-window behavior is unrelated to custom model initialization in notebooks." + }, + { + "left": "issue:45704", + "right": "issue:45859", + "accept": false, + "reason": "RMSNorm memory leak in T5 is unrelated to CPU offload compatibility in Qwen3.5 MoE rotary embedding." + }, + { + "left": "issue:43994", + "right": "issue:44671", + "accept": false, + "reason": "Different models and symptoms: SigLIP2 pipeline output issue vs CamemBERT MLM regression." + }, + { + "left": "issue:41720", + "right": "issue:45704", + "accept": false, + "reason": "CUDA assert with auto device mapping on Qwen3 is unrelated to T5 RMSNorm leak." + }, + { + "left": "issue:43596", + "right": "issue:44220", + "accept": false, + "reason": "Zero3/BertModel init failure and fbank feature extraction issue are unrelated." + }, + { + "left": "issue:43746", + "right": "issue:43873", + "accept": false, + "reason": "Local checkpoint loading with PEFT adapters is a different bug from quantization/offloading behavior." + }, + { + "left": "issue:45081", + "right": "issue:45356", + "accept": false, + "reason": "Shared tokenizer-regex terminology, but one is an AttributeError in _patch_mistral_regex and the other is a Kimi tokenizer codec regression." + }, + { + "left": "issue:39692", + "right": "issue:43873", + "accept": false, + "reason": "SigLIP2 doc/example problems are unrelated to offloading with quantization." + }, + { + "left": "issue:40444", + "right": "issue:43746", + "accept": false, + "reason": "Multi-image IterableDataset finetuning failure is unrelated to PEFT local checkpoint loading." + }, + { + "left": "issue:45356", + "right": "issue:45907", + "accept": false, + "reason": "Tokenizer codec/regression bug is unrelated to list_repo_files needing to forward revision." + }, + { + "left": "issue:30064", + "right": "issue:36010", + "accept": false, + "reason": "Void segmentation-map processing and GenerationMixin import failure are completely different issues." + }, + { + "left": "issue:44220", + "right": "issue:44661", + "accept": false, + "reason": "Audio fbank extraction and TOKENIZER_MAPPING_NAMES handling are unrelated." + }, + { + "left": "issue:43720", + "right": "issue:43819", + "accept": false, + "reason": "BitNet packed-weight loading and DAC.from_latents missing-STE behavior are different models and code paths." + }, + { + "left": "issue:43653", + "right": "issue:43720", + "accept": false, + "reason": "Special-token decode bug in BigBirdTokenizer is unrelated to BitNet accelerate-loading behavior." + }, + { + "left": "issue:43540", + "right": "issue:44008", + "accept": false, + "reason": "Video-processing ValueError in Qwen3OmniMoe is unrelated to Gemma 3n audio_mel_mask name collision." + }, + { + "left": "issue:43618", + "right": "issue:43697", + "accept": false, + "reason": "CLIP attentions assignment regression and RTDetrV2 output mismatch are separate model-specific bugs." + }, + { + "left": "issue:45183", + "right": "issue:45479", + "accept": false, + "reason": "Input validation error messaging is unrelated to zero-loss classification with num_labels=1." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 15, + "review_unit_count": 49, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-426", + "nodes": [ + "issue:30064", + "issue:36331", + "issue:39692", + "issue:41720", + "issue:43295", + "issue:43322", + "issue:43335", + "issue:43388", + "issue:43421", + "issue:43531", + "issue:43540", + "issue:43550", + "issue:43577", + "issue:43643", + "issue:43644", + "issue:43653", + "issue:43716", + "issue:43749", + "issue:43756", + "issue:43782", + "issue:43818", + "issue:43819", + "issue:43950", + "issue:44008", + "issue:44062", + "issue:44186", + "issue:44442", + "issue:44448", + "issue:44488", + "issue:44561", + "issue:44671", + "issue:44811", + "issue:44821", + "issue:44843", + "issue:44877", + "issue:44898", + "issue:44977", + "issue:44991", + "issue:45020", + "issue:45072", + "issue:45081", + "issue:45084", + "issue:45381", + "issue:45584", + "issue:45685", + "issue:45834", + "issue:45987", + "issue:46032" + ], + "soft_pairs": [ + "issue:43550|issue:45072", + "issue:43295|issue:45834", + "issue:43388|issue:46032", + "issue:43550|issue:43819", + "issue:43716|issue:44008", + "issue:44062|issue:44488", + "issue:43322|issue:45685", + "issue:43335|issue:43756", + "issue:44186|issue:44898", + "issue:43653|issue:43819", + "issue:43950|issue:45987", + "issue:43531|issue:43643", + "issue:44843|issue:45081", + "issue:43643|issue:44877", + "issue:43540|issue:45381", + "issue:30064|issue:36331", + "issue:44448|issue:44671", + "issue:43421|issue:44811", + "issue:43782|issue:44977", + "issue:41720|issue:44991", + "issue:43577|issue:43818", + "issue:44821|issue:45084", + "issue:39692|issue:45834", + "issue:45584|issue:45834", + "issue:44186|issue:44442", + "issue:44561|issue:45020", + "issue:43644|issue:43749" + ], + "prepared_review_unit_hash": "3bceb5a717713e7f592f2822df3a21298b7eb768bad6588e8f8ebd2e08cdd7b6", + "component_budget": { + "node_count": 426, + "item_count": 426, + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 27, + "serialized_chars": 36053, + "estimated_input_tokens": 9014, + "estimated_eval_tokens": 18284 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly heterogeneous: many pairs share broad Transformers/tokenizer/vision/remote_code wording, but they describe different failure modes, models, and code paths. I do not see any safe duplicate/merge candidates among the soft edges; the closest thematic grouping is around recent v5 / remote_code / tokenizer-loading regressions.", + "confidence": 0.84, + "canonical_issue_reason": "issue:45020 is the broadest umbrella among the set, covering recent Transformers regressions affecting remote_code-based models; it best matches the recurring theme behind several nearby tokenizer/config-loading reports.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45020 is the strongest representative issue for this cluster because it is general, clearly stated, and could subsume adjacent remote_code/tokenizer loading regressions better than the narrower bug reports.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43550", + "right": "issue:45072", + "accept": false, + "reason": "Different models and bugs: torch.compile/SDPA behavior for Bamba vs dtype mismatch in SwitchTransformers/TimmWrapperModel." + }, + { + "left": "issue:43295", + "right": "issue:45834", + "accept": false, + "reason": "Unrelated regressions: processor.tokenizer/images regression vs Kosmos2.5 long OCR index error." + }, + { + "left": "issue:43388", + "right": "issue:46032", + "accept": false, + "reason": "Different subsystems: metric label truncation in gather_for_metrics vs Mamba2 use_cache correctness." + }, + { + "left": "issue:43550", + "right": "issue:43819", + "accept": false, + "reason": "Both are runtime failures, but the code paths are unrelated: compile/SDPA vs DAC.from_latents/forward mismatch." + }, + { + "left": "issue:43716", + "right": "issue:44008", + "accept": false, + "reason": "Separate model bugs: image preprocessor/model dtype mismatch vs Gemma3n variable name collision in forward()." + }, + { + "left": "issue:44062", + "right": "issue:44488", + "accept": false, + "reason": "Tokenizer-related, but not the same bug: AddedToken special-kwarg duplication vs loading a specific tokenizer model." + }, + { + "left": "issue:43322", + "right": "issue:45685", + "accept": false, + "reason": "Completely different failure modes: Llava Next load segfault vs MPS int histogram unsupported in moe code." + }, + { + "left": "issue:43335", + "right": "issue:43756", + "accept": false, + "reason": "Both are config/model bugs, but one is sparse-layer creation and the other is RoPE-layer dropping; different underlying changes." + }, + { + "left": "issue:44186", + "right": "issue:44898", + "accept": false, + "reason": "Different components and symptoms: LayoutLMv2 tokenizer NER/padding crash vs Perceiver image classification with non-default resolution." + }, + { + "left": "issue:43653", + "right": "issue:43819", + "accept": false, + "reason": "Not the same issue: BigBirdTokenizer special-token registration vs DAC latent/forward mismatch." + }, + { + "left": "issue:43950", + "right": "issue:45987", + "accept": false, + "reason": "Both mention contract/regression style problems, but one is non-persistent buffer corruption and the other is cache sequence-length return type." + }, + { + "left": "issue:43531", + "right": "issue:43643", + "accept": false, + "reason": "Different regression targets: Qwen3-MoE sliding_window behavior vs trust_remote_code config field loss." + }, + { + "left": "issue:44843", + "right": "issue:45081", + "accept": false, + "reason": "Same helper name, but distinct bugs: unconditional model_info() network access vs backend_tokenizer attribute crash." + }, + { + "left": "issue:43643", + "right": "issue:44877", + "accept": false, + "reason": "Both involve config loading, but one is missing fields with trust_remote_code and the other is strict config rejection for granite_speech." + }, + { + "left": "issue:43540", + "right": "issue:45381", + "accept": false, + "reason": "Different model families and failures: Qwen3OmniMoe video input ValueError vs qwen2.5-vl vision_position_ids issue." + }, + { + "left": "issue:30064", + "right": "issue:36331", + "accept": false, + "reason": "Unrelated domains: image processor segmentation-map handling vs CustomTrainer.compute_loss signature error." + }, + { + "left": "issue:44448", + "right": "issue:44671", + "accept": false, + "reason": "Both are output regressions in v5, but different models and behaviors: Pegasus generation vs CamemBERT masked LM predictions." + }, + { + "left": "issue:43421", + "right": "issue:44811", + "accept": false, + "reason": "Tokenizer/post-processing related, but different bugs: runtime special-token updates vs Whisper batch_decode skip_special_tokens handling." + }, + { + "left": "issue:43782", + "right": "issue:44977", + "accept": false, + "reason": "Different Qwen variants and failures: weight_only loading error vs flash-attention generation correctness." + }, + { + "left": "issue:41720", + "right": "issue:44991", + "accept": false, + "reason": "Completely unrelated: Qwen3 auto device mapping CUDA assert vs tokenizer loading failure for a RoBERTa model." + }, + { + "left": "issue:43577", + "right": "issue:43818", + "accept": false, + "reason": "Both are multimodal model reports, but one is dtype persistence in BLIP2 and the other is Llava Next architecture/weights problem." + }, + { + "left": "issue:44821", + "right": "issue:45084", + "accept": false, + "reason": "Different failures: loading AutoImageProcessor from URL vs template compilation error." + }, + { + "left": "issue:39692", + "right": "issue:45834", + "accept": false, + "reason": "Documentation/example errors for SigLIP2 are unrelated to Kosmos2.5 OCR index handling." + }, + { + "left": "issue:45584", + "right": "issue:45834", + "accept": false, + "reason": "Whisper empty-transcription generation bug and Kosmos2.5 OCR index error are unrelated code paths." + }, + { + "left": "issue:44186", + "right": "issue:44442", + "accept": false, + "reason": "Both are tokenizer problems, but for different tokenizers and crashes; not a single underlying bug." + }, + { + "left": "issue:44561", + "right": "issue:45020", + "accept": false, + "reason": "Related only at a high level (v5 remote_code breakage), but these are not the same concrete bug or fix." + }, + { + "left": "issue:43644", + "right": "issue:43749", + "accept": false, + "reason": "Both are transformer loading regressions, but one corrupts non-persistent buffers while the other breaks FSDP CPU RAM efficient loading." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 16, + "review_unit_count": 49, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-426", + "nodes": [ + "issue:28218", + "issue:29159", + "issue:34634", + "issue:36010", + "issue:36331", + "issue:42175", + "issue:43262", + "issue:43329", + "issue:43526", + "issue:43577", + "issue:43582", + "issue:43606", + "issue:43650", + "issue:43653", + "issue:43697", + "issue:43720", + "issue:43742", + "issue:43749", + "issue:43782", + "issue:43819", + "issue:43825", + "issue:43827", + "issue:43906", + "issue:43994", + "issue:44060", + "issue:44291", + "issue:44368", + "issue:44479", + "issue:44485", + "issue:44509", + "issue:44554", + "issue:44610", + "issue:44671", + "issue:44805", + "issue:44857", + "issue:44912", + "issue:44936", + "issue:45020", + "issue:45072", + "issue:45081", + "issue:45084", + "issue:45325", + "issue:45406", + "issue:45538", + "issue:45600", + "issue:45701", + "issue:45834" + ], + "soft_pairs": [ + "issue:43606|issue:45072", + "issue:44479|issue:45834", + "issue:43825|issue:43994", + "issue:34634|issue:36010", + "issue:43577|issue:43582", + "issue:44805|issue:44936", + "issue:44485|issue:44509", + "issue:44671|issue:45701", + "issue:43825|issue:43827", + "issue:28218|issue:29159", + "issue:44610|issue:45072", + "issue:43720|issue:45072", + "issue:43782|issue:45325", + "issue:42175|issue:45081", + "issue:43749|issue:44291", + "issue:45081|issue:45406", + "issue:43582|issue:44857", + "issue:43650|issue:43906", + "issue:43262|issue:43819", + "issue:43526|issue:43697", + "issue:44060|issue:44368", + "issue:34634|issue:36331", + "issue:43653|issue:44912", + "issue:43742|issue:45020", + "issue:45084|issue:45538", + "issue:43526|issue:44554", + "issue:43329|issue:45600" + ], + "prepared_review_unit_hash": "9c4125658a2713a99f4f9b48866c41c27b296475a5938e2a2f4a0534b6994bfc", + "component_budget": { + "node_count": 426, + "item_count": 426, + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 27, + "serialized_chars": 35505, + "estimated_input_tokens": 8877, + "estimated_eval_tokens": 18010 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a grab bag of unrelated Transformers bug reports across tokenizers, model loading, processors, and v5 regressions. The only clear duplicate-like pair is the tokenizer spacing/inconsistency reports; a couple of others are thematically related but not the same concrete bug.", + "confidence": 0.71, + "canonical_issue_reason": "issue:45020 is the broadest and most representative report in the set: it covers recent-version breakage for models using `remote_code`, which matches the cluster\u2019s general pattern of version-regression/loading failures and has the strongest overall signal among the issues.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45020 is the best single issue to anchor the cluster because it is broad, well-discussed, and captures the most general regression theme; the rest are narrower point bugs.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43606", + "right": "issue:45072", + "accept": false, + "reason": "Both are model-loading/runtime mismatches, but they affect different subsystems and different concrete failure modes (CPU offload device mismatch vs dtype mismatch)." + }, + { + "left": "issue:44479", + "right": "issue:45834", + "accept": false, + "reason": "Both involve multimodal/video-style inputs, but the model families and bugs differ: a v5 video-input regression versus an OCR long-input index error." + }, + { + "left": "issue:43825", + "right": "issue:43994", + "accept": false, + "reason": "One is a wrong error message about removed translation pipelines; the other is incorrect outputs from AutoModel/pipeline on SigLIP. Same broad area, not the same bug." + }, + { + "left": "issue:34634", + "right": "issue:36010", + "accept": false, + "reason": "Completely different problems: BarkProcessor voice preset handling versus a GenerationMixin import error." + }, + { + "left": "issue:43577", + "right": "issue:43582", + "accept": false, + "reason": "Different code paths and symptoms: Blip2 dtype propagation versus an Apple Silicon allocator warmup TypeError." + }, + { + "left": "issue:44805", + "right": "issue:44936", + "accept": false, + "reason": "These are unrelated failures: a tensor mask shape mismatch versus trainer.evaluate() failing after training." + }, + { + "left": "issue:44485", + "right": "issue:44509", + "accept": false, + "reason": "One is a GLM-5 RoPE implementation discussion; the other is a docs cleanup issue about removed pipeline tasks." + }, + { + "left": "issue:44671", + "right": "issue:45701", + "accept": false, + "reason": "One is a CamemBERT masked-LM regression; the other is a generic tokenization-version change report. Not the same underlying bug." + }, + { + "left": "issue:43825", + "right": "issue:43827", + "accept": false, + "reason": "Both are about pipeline removals in v5, but one is runtime/error-message behavior and the other is documentation references; too broad to treat as one bug." + }, + { + "left": "issue:28218", + "right": "issue:29159", + "accept": true, + "reason": "Both point to the same tokenizer added-token inconsistency, especially around spacing/slow-vs-fast behavior, and look plausibly fixable by one tokenizer patch." + }, + { + "left": "issue:44610", + "right": "issue:45072", + "accept": false, + "reason": "Different issues: processor image-size mismatch versus dtype mismatch in inference." + }, + { + "left": "issue:43720", + "right": "issue:45072", + "accept": false, + "reason": "Packed-weight loading in BitNet and dtype mismatches in SwitchTransformers/TimmWrapper are separate bugs." + }, + { + "left": "issue:43782", + "right": "issue:45325", + "accept": false, + "reason": "Different Qwen-VL failures: from_pretrained weight_only loading versus RoPE/position-id scaling." + }, + { + "left": "issue:42175", + "right": "issue:45081", + "accept": false, + "reason": "Unrelated: packaging/backend dependency selection versus a tokenizer regex patch crash." + }, + { + "left": "issue:43749", + "right": "issue:44291", + "accept": false, + "reason": "Both involve loading paths, but one is FSDP CPU RAM-efficient loading and the other is init_empty_weights argument handling; not the same bug." + }, + { + "left": "issue:45081", + "right": "issue:45406", + "accept": false, + "reason": "Both are AttributeErrors during loading, but they arise from different objects and different processor/tokenizer code paths." + }, + { + "left": "issue:43582", + "right": "issue:44857", + "accept": false, + "reason": "Different runtime environments and failures: Apple Silicon allocator warmup TypeError versus CUDA float16 AMP loss crash." + }, + { + "left": "issue:43650", + "right": "issue:43906", + "accept": false, + "reason": "The first is a vague data request; the second is a reproduction issue for another bug. No duplicate signal." + }, + { + "left": "issue:43262", + "right": "issue:43819", + "accept": false, + "reason": "Audio chat-template sampling-rate defaults and DAC.from_latents/STE mismatch are unrelated." + }, + { + "left": "issue:43526", + "right": "issue:43697", + "accept": false, + "reason": "Different model families and failure modes: Beit label reduction bug versus RTDetrV2 output differences in v5." + }, + { + "left": "issue:44060", + "right": "issue:44368", + "accept": false, + "reason": "Both mention tied-weight warnings in Qwen models, but the titles suggest different manifestations and model variants; not enough to call the same concrete bug." + }, + { + "left": "issue:34634", + "right": "issue:36331", + "accept": false, + "reason": "Bark processor voice_preset and CustomTrainer.compute_loss signature changes are unrelated." + }, + { + "left": "issue:43653", + "right": "issue:44912", + "accept": false, + "reason": "Tokenizer special-token registration and MXFP4 quantization fallback are completely different issues." + }, + { + "left": "issue:43742", + "right": "issue:45020", + "accept": false, + "reason": "A specific key error while loading one model is not the same concrete issue as the broader remote_code regression report." + }, + { + "left": "issue:45084", + "right": "issue:45538", + "accept": false, + "reason": "Template compilation failure and an extreme `model_max_length` setting are unrelated." + }, + { + "left": "issue:43526", + "right": "issue:44554", + "accept": false, + "reason": "Beit label reduction and an MPS attention correctness issue are separate bugs." + }, + { + "left": "issue:43329", + "right": "issue:45600", + "accept": false, + "reason": "One is a multimodal-token-counting video-branch bug; the other is stale auto_mappings config references. Different code paths." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The review is conservative and grounded in the packet. The cluster summary correctly frames most items as unrelated or only loosely thematically similar, and the one accepted pair (issue:28218 / issue:29159) is plausibly the same tokenizer bug from the titles alone. The canonical issue choice of issue:45020 is also reasonable as the broadest representative of the regression/loading-failure theme. I did not see an overstatement of duplicate evidence in the soft-edge verdicts." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 17, + "review_unit_count": 49, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-426", + "nodes": [ + "issue:34567", + "issue:36010", + "issue:36331", + "issue:38175", + "issue:39692", + "issue:40990", + "issue:42490", + "issue:43257", + "issue:43262", + "issue:43329", + "issue:43450", + "issue:43452", + "issue:43454", + "issue:43531", + "issue:43550", + "issue:43577", + "issue:43606", + "issue:43638", + "issue:43643", + "issue:43645", + "issue:43696", + "issue:43720", + "issue:43746", + "issue:43749", + "issue:43819", + "issue:43854", + "issue:43927", + "issue:43931", + "issue:44387", + "issue:44479", + "issue:44568", + "issue:44589", + "issue:44610", + "issue:44617", + "issue:44756", + "issue:44811", + "issue:44849", + "issue:44912", + "issue:44936", + "issue:44977", + "issue:44991", + "issue:45042", + "issue:45356", + "issue:45459", + "issue:45701", + "issue:45759", + "issue:45803", + "issue:45859" + ], + "soft_pairs": [ + "issue:44589|issue:45042", + "issue:43746|issue:44811", + "issue:43452|issue:44991", + "issue:43638|issue:44936", + "issue:44387|issue:44617", + "issue:45459|issue:45803", + "issue:36010|issue:43645", + "issue:43819|issue:43927", + "issue:43550|issue:44912", + "issue:43854|issue:44977", + "issue:43257|issue:43329", + "issue:39692|issue:44912", + "issue:43720|issue:44610", + "issue:38175|issue:44610", + "issue:36010|issue:36331", + "issue:43531|issue:45042", + "issue:43454|issue:43577", + "issue:43257|issue:43696", + "issue:43262|issue:44811", + "issue:42490|issue:43257", + "issue:43329|issue:44479", + "issue:43643|issue:45759", + "issue:43749|issue:44756", + "issue:43329|issue:43450", + "issue:34567|issue:40990", + "issue:43606|issue:45859", + "issue:44568|issue:45356", + "issue:43257|issue:43931", + "issue:44849|issue:45701" + ], + "prepared_review_unit_hash": "84dba4b135a6b0e21decee58db714d8cf58a6e7292e4cc5680e664d20ecfd4c8", + "component_budget": { + "node_count": 426, + "item_count": 426, + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36615, + "estimated_input_tokens": 9154, + "estimated_eval_tokens": 18564 + }, + "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, tokenizer, multimodal, training, or memory bugs. I found no convincing duplicate core; the closest theme is broad tokenizer/regression reports, but even those are different code paths and failure modes.", + "confidence": 0.33, + "canonical_issue_reason": "Issue 45701 is the broadest representative item here: it describes a version-dependent tokenization regression with multiple downstream reports, so it best captures the widest cross-cutting symptom in an otherwise mixed cluster.", + "canonical_pr_reason": null, + "best_issue_reason": "45701 is the best single issue candidate because it is broad, externally corroborated, and more representative of the cluster than the many highly specific model- or feature-specific bugs.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44589", + "right": "issue:45042", + "accept": false, + "reason": "Float8 storage lookup failure and image processor torchvision dependency are different subsystems and different failure modes." + }, + { + "left": "issue:43746", + "right": "issue:44811", + "accept": false, + "reason": "Local checkpoint/PEFT loading bug and Whisper batch_decode special-token handling are unrelated code paths." + }, + { + "left": "issue:43452", + "right": "issue:44991", + "accept": false, + "reason": "Both involve loading, but one is gguf_file handling for AutoTokenizer/AutoModel and the other is a specific tokenizer regression for one model." + }, + { + "left": "issue:43638", + "right": "issue:44936", + "accept": false, + "reason": "Deepspeed ZeRO3 index error and trainer.evaluate() failing after train are distinct trainer bugs." + }, + { + "left": "issue:44387", + "right": "issue:44617", + "accept": false, + "reason": "Both mention OOM, but the model, trigger, and underlying cause are different." + }, + { + "left": "issue:45459", + "right": "issue:45803", + "accept": false, + "reason": "Both concern exception handling, but they affect different functions and different tokenizer error paths." + }, + { + "left": "issue:36010", + "right": "issue:43645", + "accept": false, + "reason": "GenerationMixin import failure and Jupyter custom-model initialization regression are unrelated." + }, + { + "left": "issue:43819", + "right": "issue:43927", + "accept": false, + "reason": "DAC from_latents STE mismatch and DiaConfig token-ID loss on save/load are different model/config issues." + }, + { + "left": "issue:43550", + "right": "issue:44912", + "accept": false, + "reason": "torch.compile+SDPA failure and MXFP4 quantization fallback are separate inference/loading problems." + }, + { + "left": "issue:43854", + "right": "issue:44977", + "accept": false, + "reason": "Both are Qwen-family reports, but one is a loading/test issue and the other is flash-attention generation behavior." + }, + { + "left": "issue:43257", + "right": "issue:43329", + "accept": false, + "reason": "Qwen3 MoE weight conversion under accelerate/deepspeed and multimodal video token counting are different bugs." + }, + { + "left": "issue:39692", + "right": "issue:44912", + "accept": false, + "reason": "Documentation example errors and quantization fallback/OOM behavior are unrelated." + }, + { + "left": "issue:43720", + "right": "issue:44610", + "accept": false, + "reason": "Packed-weight loading in accelerate and processor image-size mismatch are not the same defect." + }, + { + "left": "issue:38175", + "right": "issue:44610", + "accept": false, + "reason": "Unexpected zero probabilities and input-resolution mismatch are different model issues." + }, + { + "left": "issue:36010", + "right": "issue:36331", + "accept": false, + "reason": "Importing GenerationMixin and custom compute_loss kwarg compatibility are unrelated." + }, + { + "left": "issue:43531", + "right": "issue:45042", + "accept": false, + "reason": "Qwen3-MoE sliding_window behavior and PIL backend torchvision dependency are unrelated." + }, + { + "left": "issue:43454", + "right": "issue:43577", + "accept": false, + "reason": "Weight-tying failure in AyaVision and dtype propagation failure in Blip2 are different initialization bugs." + }, + { + "left": "issue:43257", + "right": "issue:43696", + "accept": false, + "reason": "Qwen3 MoE loading/conversion and GPT-oss OOM are unrelated symptoms and fixes." + }, + { + "left": "issue:43262", + "right": "issue:44811", + "accept": false, + "reason": "Audio chat-template sampling-rate default and Whisper batch_decode special-token handling are different processor behaviors." + }, + { + "left": "issue:42490", + "right": "issue:43257", + "accept": false, + "reason": "Local-path trailing-slash warning and Qwen3 MoE loading are unrelated." + }, + { + "left": "issue:43329", + "right": "issue:44479", + "accept": false, + "reason": "Both are video/VLM related, but one is an undefined-variable bug in token counting and the other is a broader model-specific regression; not enough evidence they are the same fix." + }, + { + "left": "issue:43643", + "right": "issue:45759", + "accept": false, + "reason": "Missing config fields with trust_remote_code and unwrapping text_config for composite multimodal configs are different config-loading bugs." + }, + { + "left": "issue:43749", + "right": "issue:44756", + "accept": false, + "reason": "FSDP CPU RAM efficient loading and disabling mmap on Strix Halo are different memory-management problems." + }, + { + "left": "issue:43329", + "right": "issue:43450", + "accept": false, + "reason": "Video token-count helper bug and batched video processor shape bug are different layers of the stack." + }, + { + "left": "issue:34567", + "right": "issue:40990", + "accept": false, + "reason": "Trainer token-count tracking and GPT-oss perplexity are unrelated." + }, + { + "left": "issue:43606", + "right": "issue:45859", + "accept": false, + "reason": "CPU offload device mismatch and rotary embedding CPU-offload incompatibility are related only by offload context, not the same code-path bug." + }, + { + "left": "issue:44568", + "right": "issue:45356", + "accept": false, + "reason": "Tokenizer special-token handling for mdeberta and Kimi codec handling are different tokenizer regressions." + }, + { + "left": "issue:43257", + "right": "issue:43931", + "accept": false, + "reason": "Qwen3 MoE accelerate/deepspeed conversion failure and Qwen3-VL weight-shape mismatch are different loading failures." + }, + { + "left": "issue:44849", + "right": "issue:45701", + "accept": false, + "reason": "Qwen3.5 hidden-states bug and broad tokenization-version regression are different APIs and failure modes." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 18, + "review_unit_count": 49, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-426", + "nodes": [ + "issue:24643", + "issue:28282", + "issue:29127", + "issue:29159", + "issue:33357", + "issue:34567", + "issue:34689", + "issue:38175", + "issue:41720", + "issue:42175", + "issue:42915", + "issue:43257", + "issue:43317", + "issue:43404", + "issue:43452", + "issue:43540", + "issue:43577", + "issue:43582", + "issue:43697", + "issue:43742", + "issue:43749", + "issue:43756", + "issue:43782", + "issue:43824", + "issue:43906", + "issue:43927", + "issue:44062", + "issue:44291", + "issue:44479", + "issue:44488", + "issue:44560", + "issue:44568", + "issue:44821", + "issue:44898", + "issue:44912", + "issue:44938", + "issue:44987", + "issue:44991", + "issue:45042", + "issue:45072", + "issue:45137", + "issue:45357", + "issue:45381", + "issue:45405", + "issue:45488", + "issue:45538", + "issue:45907", + "issue:45987" + ], + "soft_pairs": [ + "issue:44568|issue:45488", + "issue:29159|issue:34567", + "issue:44291|issue:45042", + "issue:42915|issue:43742", + "issue:42915|issue:44821", + "issue:44987|issue:45538", + "issue:43697|issue:45357", + "issue:41720|issue:43540", + "issue:42915|issue:45538", + "issue:43452|issue:43582", + "issue:43927|issue:44912", + "issue:44479|issue:45381", + "issue:33357|issue:34567", + "issue:42175|issue:43824", + "issue:43540|issue:44479", + "issue:43317|issue:45538", + "issue:43540|issue:44560", + "issue:43257|issue:43756", + "issue:44987|issue:45907", + "issue:44062|issue:44991", + "issue:24643|issue:29159", + "issue:45137|issue:45405", + "issue:43577|issue:45072", + "issue:28282|issue:34689", + "issue:43742|issue:44291", + "issue:42175|issue:44488", + "issue:42175|issue:44938", + "issue:42915|issue:43749", + "issue:43906|issue:45381", + "issue:28282|issue:29127", + "issue:43404|issue:45987", + "issue:38175|issue:44898", + "issue:43582|issue:43782" + ], + "prepared_review_unit_hash": "0ac98a271c4a3f700608e1a5113758e90d14895c0fe2aec50f95700577aac904", + "component_budget": { + "node_count": 426, + "item_count": 426, + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 33, + "serialized_chars": 37643, + "estimated_input_tokens": 9411, + "estimated_eval_tokens": 19078 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These are mostly unrelated issues across different subsystems. Only the Qwen2.5-VL / Qwen3 VL video-input regression pair looks like the same underlying bug; the rest are false-positive similarity matches.", + "confidence": 0.86, + "canonical_issue_reason": "Issue 44479 is the best umbrella representative here: it describes a concrete 5.3.0 video-input regression across multiple Qwen VL variants, and it cleanly covers the one true duplicate-like pair in this set.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 44479 is the strongest canonical candidate because it is broad, version-specific, and well-scoped to the shared Qwen video-input regression path.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44568", + "right": "issue:45488", + "accept": false, + "reason": "Both are tokenizer-related, but they describe different bugs on different models and code paths." + }, + { + "left": "issue:29159", + "right": "issue:34567", + "accept": false, + "reason": "One is tokenizer inconsistency; the other is a TrainerState token-count tracking issue." + }, + { + "left": "issue:44291", + "right": "issue:45042", + "accept": false, + "reason": "Unexpected init-empty-weights argument handling vs image processor torchvision dependency are unrelated." + }, + { + "left": "issue:42915", + "right": "issue:43742", + "accept": false, + "reason": "Qwen3MoE FP8 failure and MobileLLM key error are different model-loading problems." + }, + { + "left": "issue:42915", + "right": "issue:44821", + "accept": false, + "reason": "FineGrainedFP8Config failure is unrelated to loading AutoImageProcessor from URL." + }, + { + "left": "issue:44987", + "right": "issue:45538", + "accept": false, + "reason": "Tokenizer/model-loading failure for one repo does not match the CLIPTokenizer model_max_length bug." + }, + { + "left": "issue:43697", + "right": "issue:45357", + "accept": false, + "reason": "Different regressions: one changes RTDetr outputs, the other saves incorrect Qwen3.5 visual encoder keys." + }, + { + "left": "issue:41720", + "right": "issue:43540", + "accept": false, + "reason": "CUDA assert during auto device mapping is a different failure from Qwen3OmniMoe video input parsing." + }, + { + "left": "issue:42915", + "right": "issue:45538", + "accept": false, + "reason": "FP8 config loading failure and CLIPTokenizer max-length behavior are unrelated." + }, + { + "left": "issue:43452", + "right": "issue:43582", + "accept": false, + "reason": "GGUF loading/tokenizer issues and AppleSilicon allocator warmup TypeError are different code paths." + }, + { + "left": "issue:43927", + "right": "issue:44912", + "accept": false, + "reason": "DiaConfig save/load token-ID loss is unrelated to MXFP4 quantization falling back to bf16." + }, + { + "left": "issue:44479", + "right": "issue:45381", + "accept": true, + "reason": "Both point to the same 5.3.0 Qwen2.5-VL video-input regression; 45381 is a narrower symptom of the broader bug in 44479." + }, + { + "left": "issue:33357", + "right": "issue:34567", + "accept": false, + "reason": "MacOS bus error on a CLIP community model is unrelated to TrainerState token counting." + }, + { + "left": "issue:42175", + "right": "issue:43824", + "accept": false, + "reason": "Backend packaging dependency mismatch and missing Qwen2.5-VL symbol are different issues." + }, + { + "left": "issue:43540", + "right": "issue:44479", + "accept": false, + "reason": "Different Qwen video bugs: one is a ValueError processing video inputs, the other is a broader v5.3.0 regression report." + }, + { + "left": "issue:43317", + "right": "issue:45538", + "accept": false, + "reason": "Device-map/offload dequantization failure is unrelated to CLIPTokenizer max-length behavior." + }, + { + "left": "issue:43540", + "right": "issue:44560", + "accept": false, + "reason": "Both involve video inputs, but they are different models and symptoms." + }, + { + "left": "issue:43257", + "right": "issue:43756", + "accept": false, + "reason": "Qwen3 MoE weight-conversion failure is unrelated to Smollm3 RoPE-layer dropping." + }, + { + "left": "issue:44987", + "right": "issue:45907", + "accept": false, + "reason": "Model loading regression and list_repo_files revision forwarding are unrelated." + }, + { + "left": "issue:44062", + "right": "issue:44991", + "accept": false, + "reason": "AddedToken special-flag duplication and tokenizer loading failure for EMBEDDIA/est-roberta are not the same bug." + }, + { + "left": "issue:24643", + "right": "issue:29159", + "accept": false, + "reason": "DeepSpeed training weight shape error is unrelated to tokenizer fast/slow inconsistency." + }, + { + "left": "issue:45137", + "right": "issue:45405", + "accept": false, + "reason": "DeepSpeed ZeRO3 deque error and PEFT minimum-version bump are unrelated." + }, + { + "left": "issue:43577", + "right": "issue:45072", + "accept": false, + "reason": "Both mention dtype, but they affect different model families and different loading/inference paths." + }, + { + "left": "issue:28282", + "right": "issue:34689", + "accept": false, + "reason": "Missing PyTorch import error is not the same as Llama 3.2 Vision model-loading breakage." + }, + { + "left": "issue:43742", + "right": "issue:44291", + "accept": false, + "reason": "MobileLLM key error and init_empty_weights argument incompatibility are different failures." + }, + { + "left": "issue:42175", + "right": "issue:44488", + "accept": false, + "reason": "Packaging/backend selection issue and model loading failure for cjvt/sleng-bert are unrelated." + }, + { + "left": "issue:42175", + "right": "issue:44938", + "accept": false, + "reason": "Packaging backend contents and Python 3.14 load failure are different problems." + }, + { + "left": "issue:42915", + "right": "issue:43749", + "accept": false, + "reason": "Qwen3MoE FP8 failure is unrelated to FSDP CPU RAM efficient loading." + }, + { + "left": "issue:43906", + "right": "issue:45381", + "accept": false, + "reason": "An isolated reproduction of another issue is not the same as the Qwen2.5-VL video-position-id bug." + }, + { + "left": "issue:28282", + "right": "issue:29127", + "accept": false, + "reason": "AutoModel missing PyTorch and LayoutLMv3 box-validation messaging are unrelated." + }, + { + "left": "issue:43404", + "right": "issue:45987", + "accept": false, + "reason": "Mistral3 lm_head tying bug and StaticCache sequence-length contract bug are different code paths." + }, + { + "left": "issue:38175", + "right": "issue:44898", + "accept": false, + "reason": "Siglip2 zero-probability outputs and Perceiver resize failures are different image-model bugs." + }, + { + "left": "issue:43582", + "right": "issue:43782", + "accept": false, + "reason": "AppleSilicon allocator TypeError and Qwen3VL weight_only load error are unrelated." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet: most pairs are clearly cross-subsystem mismatches, and the single accepted Qwen video-input pair is the only plausible duplicate-like match. The soft-edge verdicts are conservative overall." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 19, + "review_unit_count": 49, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-426", + "nodes": [ + "issue:24643", + "issue:29127", + "issue:34567", + "issue:36331", + "issue:41762", + "issue:43452", + "issue:43493", + "issue:43526", + "issue:43531", + "issue:43540", + "issue:43643", + "issue:43688", + "issue:43697", + "issue:43720", + "issue:43742", + "issue:43749", + "issue:43756", + "issue:43994", + "issue:44062", + "issue:44291", + "issue:44488", + "issue:44514", + "issue:44589", + "issue:44625", + "issue:44737", + "issue:44792", + "issue:44821", + "issue:44877", + "issue:44898", + "issue:44913", + "issue:44936", + "issue:44964", + "issue:45042", + "issue:45081", + "issue:45216", + "issue:45325", + "issue:45357", + "issue:45372", + "issue:45375", + "issue:45381", + "issue:45405", + "issue:45479", + "issue:45588", + "issue:45600", + "issue:45823", + "issue:45920", + "issue:46121", + "issue:46170" + ], + "soft_pairs": [ + "issue:44737|issue:45588", + "issue:43540|issue:45823", + "issue:43526|issue:45479", + "issue:43493|issue:43994", + "issue:44291|issue:44821", + "issue:43756|issue:45325", + "issue:34567|issue:36331", + "issue:44913|issue:46121", + "issue:44062|issue:45081", + "issue:44792|issue:44964", + "issue:45042|issue:45081", + "issue:45357|issue:45920", + "issue:43452|issue:44488", + "issue:43742|issue:44589", + "issue:43531|issue:44821", + "issue:43643|issue:45375", + "issue:45405|issue:45600", + "issue:43697|issue:45216", + "issue:44625|issue:44877", + "issue:43720|issue:44898", + "issue:24643|issue:29127", + "issue:44514|issue:45381", + "issue:45372|issue:46170", + "issue:41762|issue:44936", + "issue:43688|issue:43749" + ], + "prepared_review_unit_hash": "2e7b5e427da48904ec8df9e9f1cd6ff18a2294f8ecf6757c7f10be2e50765bb8", + "component_budget": { + "node_count": 426, + "item_count": 426, + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 25, + "serialized_chars": 35649, + "estimated_input_tokens": 8913, + "estimated_eval_tokens": 18082 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This set is mostly a grab bag of unrelated bugs across training, config loading, tokenization, multimodal processors, and model-specific regressions. None of the soft edges is strong enough to merge as the same underlying defect, so I would not consolidate any of them.", + "confidence": 0.88, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44737", + "right": "issue:45588", + "accept": false, + "reason": "Different code paths and symptoms: XLNet CPU-bound positional encoding vs a flash-attention AttributeError in sink-less models." + }, + { + "left": "issue:43540", + "right": "issue:45823", + "accept": false, + "reason": "Different models and failures: Qwen3OmniMoe video input processing vs Gemma4 PLE device placement during forward." + }, + { + "left": "issue:43526", + "right": "issue:45479", + "accept": false, + "reason": "One is an image processor label-reduction bug; the other is a sequence-classification loss issue with num_labels=1." + }, + { + "left": "issue:43493", + "right": "issue:43994", + "accept": false, + "reason": "Both involve SigLIP2, but one is a HF-vs-JAX implementation discrepancy and the other is a specific AutoModel/pipeline symptom; not clearly the same bug." + }, + { + "left": "issue:44291", + "right": "issue:44821", + "accept": false, + "reason": "Unrelated failures: init_empty_weights/_is_hf_initialized handling vs loading AutoImageProcessor from URL." + }, + { + "left": "issue:43756", + "right": "issue:45325", + "accept": false, + "reason": "Both mention RoPE, but the bugs are model-specific and different: dropped RoPE layers vs wrong temporal position scaling." + }, + { + "left": "issue:34567", + "right": "issue:36331", + "accept": false, + "reason": "TrainerState token counting and CustomTrainer.compute_loss signature regression are different issues." + }, + { + "left": "issue:44913", + "right": "issue:46121", + "accept": false, + "reason": "Both touch rope config handling, but one is silent default reversion on reload and the other is a TypeError in validation for JSON-loaded lists." + }, + { + "left": "issue:44062", + "right": "issue:45081", + "accept": false, + "reason": "Different tokenizer bugs: AddedToken special-kwarg duplication vs backend_tokenizer missing during Mistral regex patching." + }, + { + "left": "issue:44792", + "right": "issue:44964", + "accept": false, + "reason": "Different models and failure modes: janus image-generation test failure vs Phi-4 multimodal load failure." + }, + { + "left": "issue:45042", + "right": "issue:45081", + "accept": false, + "reason": "Unrelated subsystems: PIL image processors requiring torchvision vs a Mistral tokenizer regex patch crash." + }, + { + "left": "issue:45357", + "right": "issue:45920", + "accept": false, + "reason": "Different regression classes: Qwen3.5 save_pretrained key handling vs AutoTokenizer returning wrong token IDs across several models." + }, + { + "left": "issue:43452", + "right": "issue:44488", + "accept": false, + "reason": "Both are loading problems, but one is a gguf_file API break in auto loaders and the other is a specific model load failure; too broad to treat as one bug." + }, + { + "left": "issue:43742", + "right": "issue:44589", + "accept": false, + "reason": "Different load-time errors: MobileLLM key error vs missing Float8 storage object." + }, + { + "left": "issue:43531", + "right": "issue:44821", + "accept": false, + "reason": "Qwen3-MoE sliding_window behavior is unrelated to AutoImageProcessor URL loading." + }, + { + "left": "issue:43643", + "right": "issue:45375", + "accept": false, + "reason": "Both concern missing config fields, but one is generic AutoConfig remote-code loading and the other is a specific Qwen3.5 vision config field dropped by @strict." + }, + { + "left": "issue:45405", + "right": "issue:45600", + "accept": false, + "reason": "Unrelated maintenance issues: PEFT minimum-version bump vs stale auto_mappings references." + }, + { + "left": "issue:43697", + "right": "issue:45216", + "accept": false, + "reason": "Different problems: RTDetrV2 output mismatch in v5 vs Qwen3.5 save_pretrained checkpoint corruption." + }, + { + "left": "issue:44625", + "right": "issue:44877", + "accept": false, + "reason": "Related only at a very high level; one is num_labels propagation between configs, the other is strict config loading for granite_speech." + }, + { + "left": "issue:43720", + "right": "issue:44898", + "accept": false, + "reason": "Different model families and code paths: BitNet packed-weight loading vs Perceiver interpolation with non-default resolution." + }, + { + "left": "issue:24643", + "right": "issue:29127", + "accept": false, + "reason": "Completely different reports: DeepSpeed training weight-shape runtime error vs LayoutLMv3 error-message clarity." + }, + { + "left": "issue:44514", + "right": "issue:45381", + "accept": false, + "reason": "Same Qwen2.5-VL family, but one is a batched chat-template crash and the other is wrong video vision_position_ids; adjacent but not the same bug." + }, + { + "left": "issue:45372", + "right": "issue:46170", + "accept": false, + "reason": "Gemma 4 processor loading via mistral_common and Windows ROCm torch.distributed import issues are unrelated." + }, + { + "left": "issue:41762", + "right": "issue:44936", + "accept": false, + "reason": "Different trainer/model flows: Gemma3 ZeRO-3 load-time index error vs trainer.evaluate() failing after train()." + }, + { + "left": "issue:43688", + "right": "issue:43749", + "accept": false, + "reason": "Both involve training/loading infrastructure, but auxiliary-loss normalization and FSDP CPU RAM-efficient loading are distinct bugs." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 20, + "review_unit_count": 49, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-426", + "nodes": [ + "issue:28282", + "issue:29159", + "issue:30064", + "issue:32090", + "issue:33357", + "issue:33666", + "issue:35141", + "issue:38175", + "issue:41720", + "issue:43295", + "issue:43452", + "issue:43531", + "issue:43550", + "issue:43577", + "issue:43646", + "issue:43653", + "issue:43688", + "issue:43697", + "issue:43742", + "issue:43749", + "issue:43819", + "issue:43906", + "issue:43927", + "issue:44361", + "issue:44442", + "issue:44464", + "issue:44466", + "issue:44484", + "issue:44493", + "issue:44514", + "issue:44568", + "issue:44704", + "issue:44743", + "issue:44811", + "issue:44843", + "issue:44936", + "issue:44977", + "issue:44991", + "issue:45005", + "issue:45290", + "issue:45356", + "issue:45440", + "issue:45491", + "issue:45701", + "issue:45704", + "issue:45823", + "issue:46002", + "issue:46050" + ], + "soft_pairs": [ + "issue:44514|issue:45290", + "issue:43452|issue:44843", + "issue:44493|issue:46050", + "issue:43452|issue:43577", + "issue:43819|issue:44811", + "issue:44466|issue:45005", + "issue:44936|issue:45823", + "issue:30064|issue:33666", + "issue:44484|issue:44704", + "issue:43295|issue:45356", + "issue:44464|issue:44743", + "issue:43653|issue:44361", + "issue:44361|issue:44442", + "issue:28282|issue:32090", + "issue:41720|issue:43906", + "issue:28282|issue:33357", + "issue:38175|issue:43295", + "issue:28282|issue:35141", + "issue:45491|issue:46002", + "issue:43653|issue:44568", + "issue:43577|issue:43697", + "issue:43688|issue:45704", + "issue:44843|issue:44991", + "issue:43688|issue:45440", + "issue:43646|issue:43749", + "issue:29159|issue:33357", + "issue:43688|issue:45701", + "issue:43531|issue:43742", + "issue:43653|issue:43927", + "issue:44464|issue:44977", + "issue:43550|issue:43927" + ], + "prepared_review_unit_hash": "0ea78acd5e1a55eab854a1d5a9719e69f92f4cfa96ff4d54d723f37425939f87", + "component_budget": { + "node_count": 426, + "item_count": 426, + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 37025, + "estimated_input_tokens": 9257, + "estimated_eval_tokens": 18770 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is heterogeneous and does not contain a clear duplicate set. The strongest topical overlap is around tokenizer/load-path regressions, but the pairwise evidence is mostly too weak or points to different concrete failure modes.", + "confidence": 0.32, + "canonical_issue_reason": "issue:43452 is the broadest representative of the recurring from_pretrained/tokenizer-loading regression theme, and it\u2019s the best fit for a mixed cluster with no single clean duplicate family.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43452 is the most central and broadly representative issue in the cluster, even though the cluster overall is not cohesive enough to support a tighter canonical duplicate.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44514", + "right": "issue:45290", + "accept": false, + "reason": "Both are apply_chat_template crashes, but the triggers differ: batched input with padding=False vs assistant messages with tool calls and no content." + }, + { + "left": "issue:43452", + "right": "issue:44843", + "accept": false, + "reason": "Same loader area, but different concrete bugs: gguf_file support vs unconditional model_info() in offline regex patching." + }, + { + "left": "issue:44493", + "right": "issue:46050", + "accept": false, + "reason": "Different symptoms and paths: unexpected position-id keys after save/load vs NaNs after sinq quantization round-trip." + }, + { + "left": "issue:43452", + "right": "issue:43577", + "accept": false, + "reason": "Different bugs: gguf_file/from_pretrained loading breakage vs BLIP2 dtype not propagating to qformer." + }, + { + "left": "issue:43819", + "right": "issue:44811", + "accept": false, + "reason": "Different subsystems: DAC latent reconstruction mismatch vs Whisper batch_decode skip_special_tokens handling." + }, + { + "left": "issue:44466", + "right": "issue:45005", + "accept": false, + "reason": "Both affect generation behavior, but one is compiled chunked generation and the other is tied-weights serialization." + }, + { + "left": "issue:44936", + "right": "issue:45823", + "accept": false, + "reason": "Different concrete problems: trainer lifecycle failure vs Gemma4 device mismatch during forward." + }, + { + "left": "issue:30064", + "right": "issue:33666", + "accept": false, + "reason": "Different tasks and failure modes: void segmentation-map processing vs Qwen2-VL multi-GPU training." + }, + { + "left": "issue:44484", + "right": "issue:44704", + "accept": false, + "reason": "Not the same bug: shard-size default question vs AutoProcessor not forwarding kwargs to cached_file." + }, + { + "left": "issue:43295", + "right": "issue:45356", + "accept": false, + "reason": "Related tokenizer regressions, but different code paths and symptoms: processor.tokenizer/image passing vs Kimi-K2.5 codec/warning handling." + }, + { + "left": "issue:44464", + "right": "issue:44743", + "accept": false, + "reason": "Different generation bugs: compiled-forward chunked output inconsistency vs recurrent-state reset with cache." + }, + { + "left": "issue:43653", + "right": "issue:44361", + "accept": false, + "reason": "Different tokenizer-specific bugs: BigBird special-token decode issue vs MLukeTokenizer AttributeError on tasks." + }, + { + "left": "issue:44361", + "right": "issue:44442", + "accept": false, + "reason": "Different failures: MLukeTokenizer task error vs AutoTokenizer loading FastSpeech2ConformerTokenizer." + }, + { + "left": "issue:28282", + "right": "issue:32090", + "accept": false, + "reason": "Unrelated: missing PyTorch ImportError vs distributed broadcast TypeError on NoneType." + }, + { + "left": "issue:41720", + "right": "issue:43906", + "accept": false, + "reason": "43906 is only an isolated reproduction reference; not enough evidence it is the same cudaErrorAssert bug." + }, + { + "left": "issue:28282", + "right": "issue:33357", + "accept": false, + "reason": "Different issues: missing PyTorch dependency vs MacOS bus error on a community CLIP model." + }, + { + "left": "issue:38175", + "right": "issue:43295", + "accept": false, + "reason": "Different model-level bugs: zero probabilities in SigLIP2 vs processor.tokenizer regression." + }, + { + "left": "issue:28282", + "right": "issue:35141", + "accept": false, + "reason": "Completely different problems: missing PyTorch dependency vs output embedding reinitialization after resize." + }, + { + "left": "issue:45491", + "right": "issue:46002", + "accept": false, + "reason": "Both mention NaNs, but the mechanisms differ: sliding-window/all-padding Gemma3 embeddings vs classification-head NaNs after checkout/init." + }, + { + "left": "issue:43653", + "right": "issue:44568", + "accept": false, + "reason": "Both involve special tokens, but they are different tokenizer behaviors and models." + }, + { + "left": "issue:43577", + "right": "issue:43697", + "accept": false, + "reason": "Different bugs: BLIP2 dtype propagation vs RTDetrV2 output divergence in v5." + }, + { + "left": "issue:43688", + "right": "issue:45704", + "accept": false, + "reason": "Different issues: MoE auxiliary-loss normalization vs T5 apex.FusedRMSNorm memory leak." + }, + { + "left": "issue:44843", + "right": "issue:44991", + "accept": false, + "reason": "Both are tokenizer-loading regressions, but the concrete causes differ." + }, + { + "left": "issue:43688", + "right": "issue:45440", + "accept": false, + "reason": "Different correctness bugs: MoE auxiliary-loss normalization vs DeepseekV3MoE divergence from remote implementation." + }, + { + "left": "issue:43646", + "right": "issue:43749", + "accept": false, + "reason": "Different scopes: custom model initialization breakage in v5 vs FSDP CPU RAM efficient loading regression." + }, + { + "left": "issue:29159", + "right": "issue:33357", + "accept": false, + "reason": "Different failures: slow/fast tokenizer inconsistency vs MacOS bus error on CLIP model load." + }, + { + "left": "issue:43688", + "right": "issue:45701", + "accept": false, + "reason": "Too broad and not the same bug: specific MoE auxiliary-loss normalization vs generic version-related tokenization changes." + }, + { + "left": "issue:43531", + "right": "issue:43742", + "accept": false, + "reason": "Different models and failures: Qwen3-MoE sliding_window issue vs MobileLLM-125M key error on load." + }, + { + "left": "issue:43653", + "right": "issue:43927", + "accept": false, + "reason": "Different problems: BigBird special-token registration/decode vs DiaConfig losing custom token IDs after save/load." + }, + { + "left": "issue:44464", + "right": "issue:44977", + "accept": false, + "reason": "Different generation bugs: compiled-forward chunked generation inconsistency vs Qwen3.5 flash-attention failure." + }, + { + "left": "issue:43550", + "right": "issue:43927", + "accept": false, + "reason": "Different code paths: torch.compile + SDPA failure vs DiaConfig token-ID persistence bug." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "The cluster summary is broadly grounded, but one soft-edge verdict is too strong: issue:44514 vs issue:45290 are both apply_chat_template crashes, yet the packet only supports a loose topical similarity, not a clear same-bug duplicate. The rest of the rejects are conservative and well-aligned with the titles, but because the lone accepted pair overstates the evidence, I would reject this evaluation." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 21, + "review_unit_count": 49, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-426", + "nodes": [ + "issue:28218", + "issue:32090", + "issue:33357", + "issue:33666", + "issue:34567", + "issue:36010", + "issue:36331", + "issue:41720", + "issue:43122", + "issue:43262", + "issue:43278", + "issue:43295", + "issue:43452", + "issue:43493", + "issue:43504", + "issue:43525", + "issue:43550", + "issue:43572", + "issue:43577", + "issue:43606", + "issue:43643", + "issue:43688", + "issue:43720", + "issue:43742", + "issue:43756", + "issue:43825", + "issue:43906", + "issue:43927", + "issue:44361", + "issue:44451", + "issue:44464", + "issue:44466", + "issue:44479", + "issue:44554", + "issue:44561", + "issue:44568", + "issue:44610", + "issue:44625", + "issue:44821", + "issue:44849", + "issue:44913", + "issue:45357", + "issue:45588", + "issue:45812", + "issue:45823", + "issue:46002", + "issue:46050" + ], + "soft_pairs": [ + "issue:43643|issue:44561", + "issue:44479|issue:44625", + "issue:43550|issue:44361", + "issue:41720|issue:44464", + "issue:43122|issue:43906", + "issue:43577|issue:45588", + "issue:43577|issue:44554", + "issue:32090|issue:34567", + "issue:43504|issue:43720", + "issue:43504|issue:43550", + "issue:43504|issue:44610", + "issue:44466|issue:46050", + "issue:32090|issue:33666", + "issue:43742|issue:44821", + "issue:33666|issue:36010", + "issue:45823|issue:46050", + "issue:43452|issue:44451", + "issue:43720|issue:43927", + "issue:43606|issue:43927", + "issue:43278|issue:43688", + "issue:43262|issue:43295", + "issue:44849|issue:45357", + "issue:43525|issue:43572", + "issue:33666|issue:36331", + "issue:44625|issue:46002", + "issue:43756|issue:44913", + "issue:43720|issue:44361", + "issue:43493|issue:43825", + "issue:44568|issue:45812", + "issue:28218|issue:33357" + ], + "prepared_review_unit_hash": "1eec4a1d6d8ce9cebbdc05b0be572d114944f5cd5fb12c13400e6469ce31024d", + "component_budget": { + "node_count": 426, + "item_count": 426, + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 30, + "serialized_chars": 36216, + "estimated_input_tokens": 9054, + "estimated_eval_tokens": 18364 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is very heterogeneous: the candidate edges mostly connect unrelated regression reports across tokenization, loading, training, device placement, and model-specific behavior. I would not merge any of the soft pairs as duplicates.", + "confidence": 0.93, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43643", + "right": "issue:44561", + "accept": false, + "reason": "Both involve trust_remote_code, but they describe different failures: missing fields in AutoConfig return values vs. removal of is_torch_fx_available breaking remote-code models." + }, + { + "left": "issue:44479", + "right": "issue:44625", + "accept": false, + "reason": "One is a video-input regression for Qwen VL models; the other is num_labels propagation from core config to text config. Different code paths." + }, + { + "left": "issue:43550", + "right": "issue:44361", + "accept": false, + "reason": "Bamba torch.compile/SDPA failure and MLukeTokenizer AttributeError are unrelated bugs." + }, + { + "left": "issue:41720", + "right": "issue:44464", + "accept": false, + "reason": "cudaErrorAssert with auto device mapping is not the same as inconsistent chunked generation under compiled forward." + }, + { + "left": "issue:43122", + "right": "issue:43906", + "accept": false, + "reason": "One reports tokenization changes across versions; the other is just an isolated reproduction reference for a separate upstream issue." + }, + { + "left": "issue:43577", + "right": "issue:45588", + "accept": false, + "reason": "Blip2 dtype staying float32 and flash_attention crashing on s_aux=None are different failures." + }, + { + "left": "issue:43577", + "right": "issue:44554", + "accept": false, + "reason": "A dtype propagation bug is unrelated to the MPS attention correctness issue with differing value/query head dims." + }, + { + "left": "issue:32090", + "right": "issue:34567", + "accept": false, + "reason": "Trainer broadcast NoneType error and num_input_tokens_seen not updating are separate Trainer issues." + }, + { + "left": "issue:43504", + "right": "issue:43720", + "accept": false, + "reason": "Legacy preset loading for Beit and BitNet packed-weight loading during accelerate are different load paths and failure modes." + }, + { + "left": "issue:43504", + "right": "issue:43550", + "accept": false, + "reason": "Beit legacy-field loading bug is unrelated to Bamba torch.compile SDPA failure." + }, + { + "left": "issue:43504", + "right": "issue:44610", + "accept": false, + "reason": "Loading a pretrained Beit preset and OmDet processor producing the wrong image size are unrelated." + }, + { + "left": "issue:44466", + "right": "issue:46050", + "accept": false, + "reason": "lm_head.weight serialization depending on device and sinq quantized save/load breakage are both save/load themed but not the same bug." + }, + { + "left": "issue:32090", + "right": "issue:33666", + "accept": false, + "reason": "Trainer GPU broadcast TypeError is unrelated to Qwen2-VL multi-GPU training behavior." + }, + { + "left": "issue:43742", + "right": "issue:44821", + "accept": false, + "reason": "A key error loading MobileLLM and an AutoImageProcessor URL-loading failure are different loader problems." + }, + { + "left": "issue:33666", + "right": "issue:36010", + "accept": false, + "reason": "Qwen2-VL multi-GPU training and a GenerationMixin import error do not share the same underlying cause." + }, + { + "left": "issue:45823", + "right": "issue:46050", + "accept": false, + "reason": "Gemma4 device_map auto forward mismatch is unrelated to quantized save/load corruption." + }, + { + "left": "issue:43452", + "right": "issue:44451", + "accept": false, + "reason": "gguf_file breaking AutoTokenizer/AutoModel loading and ScandiBERT load failure are distinct model-loading issues." + }, + { + "left": "issue:43720", + "right": "issue:43927", + "accept": false, + "reason": "BitNet packed-weight unpacking during load and DiaConfig losing custom token IDs on save/load are unrelated." + }, + { + "left": "issue:43606", + "right": "issue:43927", + "accept": false, + "reason": "CPU offload device mismatch for bark-small is not the same as DiaConfig save/load token ID loss." + }, + { + "left": "issue:43278", + "right": "issue:43688", + "accept": false, + "reason": "Embedding dtype changing on eval and auxiliary loss normalization are separate concerns." + }, + { + "left": "issue:43262", + "right": "issue:43295", + "accept": false, + "reason": "Audio chat-template sampling-rate default and processor.tokenizer/image passing regression are different processor bugs." + }, + { + "left": "issue:44849", + "right": "issue:45357", + "accept": false, + "reason": "Qwen3.5 output_hidden_states bug and incorrect visual encoder keys in save_pretrained are unrelated." + }, + { + "left": "issue:43525", + "right": "issue:43572", + "accept": false, + "reason": "Both mention missing pad token config fields, but they affect different model families and different attribute names; not the same concrete bug." + }, + { + "left": "issue:44625", + "right": "issue:46002", + "accept": false, + "reason": "num_labels propagation failure is not the same as NaNs in classification heads after checkout/init." + }, + { + "left": "issue:43756", + "right": "issue:44913", + "accept": false, + "reason": "Smollm3 RoPE-layer dropping and GPTNeoX rotary_pct resetting on reload are different rotary/config issues." + }, + { + "left": "issue:43720", + "right": "issue:44361", + "accept": false, + "reason": "BitNet load failure and MLukeTokenizer AttributeError are unrelated." + }, + { + "left": "issue:43493", + "right": "issue:43825", + "accept": false, + "reason": "SigLIP2 HF/JAX discrepancy and pipeline translation-task error messaging are completely different problems." + }, + { + "left": "issue:44568", + "right": "issue:45812", + "accept": false, + "reason": "These are both tokenizer regressions, but for different models and symptoms: missing BOS/EOS vs wrong token IDs." + }, + { + "left": "issue:28218", + "right": "issue:33357", + "accept": false, + "reason": "Tokenizer extra-space behavior and a MacOS bus error on CLIP are 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": 1, + "component_count": 17, + "review_unit_index": 22, + "review_unit_count": 49, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-426", + "nodes": [ + "issue:33357", + "issue:33666", + "issue:39692", + "issue:43257", + "issue:43278", + "issue:43295", + "issue:43299", + "issue:43334", + "issue:43493", + "issue:43504", + "issue:43540", + "issue:43572", + "issue:43575", + "issue:43606", + "issue:43646", + "issue:43653", + "issue:43701", + "issue:43756", + "issue:43824", + "issue:43844", + "issue:43873", + "issue:43927", + "issue:43994", + "issue:44112", + "issue:44291", + "issue:44361", + "issue:44423", + "issue:44464", + "issue:44466", + "issue:44488", + "issue:44493", + "issue:44589", + "issue:44734", + "issue:44792", + "issue:44843", + "issue:44964", + "issue:44977", + "issue:45083", + "issue:45200", + "issue:45229", + "issue:45405", + "issue:45520", + "issue:45584", + "issue:45759", + "issue:45803", + "issue:45823", + "issue:45987", + "issue:46050" + ], + "soft_pairs": [ + "issue:44964|issue:45405", + "issue:39692|issue:43493", + "issue:43540|issue:44977", + "issue:43334|issue:43572", + "issue:43493|issue:43844", + "issue:45200|issue:45229", + "issue:44488|issue:44843", + "issue:43278|issue:43844", + "issue:43257|issue:43299", + "issue:43756|issue:45083", + "issue:45803|issue:45987", + "issue:43334|issue:45759", + "issue:39692|issue:43295", + "issue:43646|issue:44466", + "issue:43824|issue:44589", + "issue:45229|issue:45823", + "issue:33357|issue:33666", + "issue:44589|issue:45520", + "issue:43504|issue:43606", + "issue:43572|issue:44291", + "issue:43927|issue:44361", + "issue:44464|issue:45584", + "issue:43606|issue:44112", + "issue:43504|issue:43653", + "issue:43575|issue:45405", + "issue:43278|issue:43701", + "issue:44792|issue:45405", + "issue:43994|issue:44493", + "issue:44423|issue:44734", + "issue:43873|issue:46050" + ], + "prepared_review_unit_hash": "c3b704b2219eb3d8926b2705f2b89670ba916edc7b6739f4eecdceb8fd64653b", + "component_budget": { + "node_count": 426, + "item_count": 426, + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 36850, + "estimated_input_tokens": 9213, + "estimated_eval_tokens": 18682 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is heterogeneous: the soft pairs mostly connect unrelated model-loading, config, quantization, offload, and CI issues rather than true duplicates. I would not merge any of the suggested soft edges; if forced to pick a representative issue, Gemma4 multi-GPU OOM is the strongest anchor.", + "confidence": 0.78, + "canonical_issue_reason": "issue:45229 is the best canonical anchor because it has the clearest concrete runtime failure, the most discussion, and broad user impact compared with the more niche config/docs regressions.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45229 is the most suitable overall representative for this cluster; the other items are mostly isolated model-specific or release-specific bugs, not one underlying duplicate.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44964", + "right": "issue:45405", + "accept": false, + "reason": "Phi-4 multimodal loading regression vs an unreleased PEFT version bump; different problems." + }, + { + "left": "issue:39692", + "right": "issue:43493", + "accept": false, + "reason": "Both are SigLIP2-related, but one is a docs/example mismatch and the other is an implementation discrepancy with JAX." + }, + { + "left": "issue:43540", + "right": "issue:44977", + "accept": false, + "reason": "Qwen3 video-input ValueError vs Qwen3.5 flash-attention generation behavior; not the same bug." + }, + { + "left": "issue:43334", + "right": "issue:43572", + "accept": false, + "reason": "Both mention missing pad-token fields after 5.0, but they affect different model configs and code paths." + }, + { + "left": "issue:43493", + "right": "issue:43844", + "accept": false, + "reason": "SigLIP2 implementation mismatch vs ZeRO-3 gradient growth; unrelated failure modes." + }, + { + "left": "issue:45200", + "right": "issue:45229", + "accept": false, + "reason": "Text-only fine-tuning token defaults vs Gemma4 multi-GPU inference OOM; different issues." + }, + { + "left": "issue:44488", + "right": "issue:44843", + "accept": false, + "reason": "Model loading failure vs offline hub lookup regression; not the same underlying bug." + }, + { + "left": "issue:43278", + "right": "issue:43844", + "accept": false, + "reason": "Embedding dtype drift vs DeepSpeed gradient anomaly; different code paths." + }, + { + "left": "issue:43257", + "right": "issue:43299", + "accept": false, + "reason": "Both are Qwen3 MOE-related, but one is accelerate/deepspeed conversion and the other is a dev loading regression for Qwen3VL." + }, + { + "left": "issue:43756", + "right": "issue:45083", + "accept": false, + "reason": "Smollm3 RoPE-layer bug vs qwen3_omni_moe feature-length helper behavior; unrelated models." + }, + { + "left": "issue:45803", + "right": "issue:45987", + "accept": false, + "reason": "Bare except in Fuyu tensor conversion vs StaticCache return-type mismatch; unrelated." + }, + { + "left": "issue:43334", + "right": "issue:45759", + "accept": false, + "reason": "Different Qwen multimodal config bugs: missing pad_token_id vs text_config unwrapping in from_config." + }, + { + "left": "issue:39692", + "right": "issue:43295", + "accept": false, + "reason": "Docs/example mismatch in SigLIP2 vs processor.tokenizer regression; different bugs." + }, + { + "left": "issue:43646", + "right": "issue:44466", + "accept": false, + "reason": "Custom model initialization breakage vs lm_head.weight serialization inconsistency; not the same code-path." + }, + { + "left": "issue:43824", + "right": "issue:44589", + "accept": false, + "reason": "Qwen2.5-VL import failure vs float8 storage lookup error; unrelated." + }, + { + "left": "issue:45229", + "right": "issue:45823", + "accept": false, + "reason": "Both are Gemma4, but one is multi-GPU inference OOM and the other is PLE device mismatch in forward." + }, + { + "left": "issue:33357", + "right": "issue:33666", + "accept": false, + "reason": "MacOS CLIP bus error vs Qwen2-VL multi-GPU training; unrelated." + }, + { + "left": "issue:44589", + "right": "issue:45520", + "accept": false, + "reason": "Float8 storage object error vs Python 3.13 flash_attn import key error; different subsystem." + }, + { + "left": "issue:43504", + "right": "issue:43606", + "accept": false, + "reason": "Legacy field loading issue in Beit vs CPU-offload device mismatch in Bark; unrelated." + }, + { + "left": "issue:43572", + "right": "issue:44291", + "accept": false, + "reason": "Missing pad_token_idx vs unexpected _is_hf_initialized arg; both are release regressions but not the same bug." + }, + { + "left": "issue:43927", + "right": "issue:44361", + "accept": false, + "reason": "DiaConfig token persistence loss vs MLukeTokenizer AttributeError; distinct failures." + }, + { + "left": "issue:44464", + "right": "issue:45584", + "accept": false, + "reason": "Compiled-forward chunked generation inconsistency vs Whisper empty transcription failure after token alignment." + }, + { + "left": "issue:43606", + "right": "issue:44112", + "accept": false, + "reason": "Both are CI/device-related, but they concern different models and failure mechanisms." + }, + { + "left": "issue:43504", + "right": "issue:43653", + "accept": false, + "reason": "Beit legacy-field loading bug vs BigBirdTokenizer special-token registration issue; unrelated." + }, + { + "left": "issue:43575", + "right": "issue:45405", + "accept": false, + "reason": "Qwen2 tp OOM vs an unreleased PEFT version bump; unrelated." + }, + { + "left": "issue:43278", + "right": "issue:43701", + "accept": false, + "reason": "Embedding dtype regression vs resume_from_checkpoint key mismatch; different code paths." + }, + { + "left": "issue:44792", + "right": "issue:45405", + "accept": false, + "reason": "Janus image-generation test failure vs PEFT version pin; unrelated." + }, + { + "left": "issue:43994", + "right": "issue:44493", + "accept": false, + "reason": "SigLIP2 output discrepancy vs unexpected position-id key warnings; different bugs." + }, + { + "left": "issue:44423", + "right": "issue:44734", + "accept": false, + "reason": "Serve crashes in multimodal batching vs KV-cache continuation tensor indexing; both are serve bugs, but not the same issue." + }, + { + "left": "issue:43873", + "right": "issue:46050", + "accept": false, + "reason": "Quantization/offloading behavior vs sinq save/load breakage; related theme but different concrete failures." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 23, + "review_unit_count": 49, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-426", + "nodes": [ + "issue:15354", + "issue:28218", + "issue:29159", + "issue:29942", + "issue:30333", + "issue:32090", + "issue:34567", + "issue:37428", + "issue:38175", + "issue:42490", + "issue:43232", + "issue:43295", + "issue:43450", + "issue:43493", + "issue:43540", + "issue:43572", + "issue:43582", + "issue:43696", + "issue:43701", + "issue:43824", + "issue:43844", + "issue:43873", + "issue:43957", + "issue:44315", + "issue:44466", + "issue:44521", + "issue:44554", + "issue:44610", + "issue:44849", + "issue:44964", + "issue:44977", + "issue:44987", + "issue:45072", + "issue:45084", + "issue:45127", + "issue:45198", + "issue:45200", + "issue:45290", + "issue:45325", + "issue:45356", + "issue:45406", + "issue:45446", + "issue:45520", + "issue:45588", + "issue:45823", + "issue:46050", + "issue:46121" + ], + "soft_pairs": [ + "issue:29942|issue:30333", + "issue:45406|issue:46121", + "issue:45588|issue:46050", + "issue:42490|issue:43696", + "issue:44521|issue:45290", + "issue:43844|issue:44964", + "issue:28218|issue:34567", + "issue:29159|issue:32090", + "issue:45084|issue:45520", + "issue:43582|issue:44977", + "issue:44554|issue:45446", + "issue:43540|issue:45200", + "issue:15354|issue:38175", + "issue:43232|issue:44849", + "issue:37428|issue:43957", + "issue:45072|issue:45198", + "issue:45200|issue:45823", + "issue:43493|issue:43701", + "issue:44315|issue:44849", + "issue:43572|issue:44987", + "issue:43295|issue:44610", + "issue:43295|issue:43450", + "issue:44554|issue:45356", + "issue:43572|issue:43957", + "issue:37428|issue:43824", + "issue:45200|issue:45325", + "issue:43295|issue:43873", + "issue:44466|issue:45127" + ], + "prepared_review_unit_hash": "97ce52e651462554b0e837d4dc2101d414e731db7eae550a3faccab980acef60", + "component_budget": { + "node_count": 426, + "item_count": 426, + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 28, + "serialized_chars": 35685, + "estimated_input_tokens": 8922, + "estimated_eval_tokens": 18100 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "No strong duplicate cluster here; the items are mostly unrelated bugs/regressions across different features. I\u2019d treat issue 43295 as the best representative issue for triage, but not as a duplicate hub.", + "confidence": 0.96, + "canonical_issue_reason": "Issue 43295 is the most concrete, actionable regression report in the set, with a narrow failure mode and clear reproduction context.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 43295 is the strongest overall triage candidate because it describes a specific regression path and is easier to act on than the more vague or subsystem-broad reports.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:29942", + "right": "issue:30333", + "accept": false, + "reason": "Flash Attention CI failures and MLFlow job-status reporting are unrelated bugs in different subsystems." + }, + { + "left": "issue:45406", + "right": "issue:46121", + "accept": false, + "reason": "One is a Gemma4 processor attribute crash; the other is rope-parameter validation with JSON-loaded lists. Different code paths." + }, + { + "left": "issue:45588", + "right": "issue:46050", + "accept": false, + "reason": "Flash-attention sink handling and sinq save/load corruption are distinct quantization/attention issues." + }, + { + "left": "issue:42490", + "right": "issue:43696", + "accept": false, + "reason": "Local-path import warnings and CUDA OOM are not the same underlying problem." + }, + { + "left": "issue:44521", + "right": "issue:45290", + "accept": false, + "reason": "Both involve apply_chat_template, but one is zero assistant masks for multimodal input and the other is a crash on tool-call messages; different failures." + }, + { + "left": "issue:43844", + "right": "issue:44964", + "accept": false, + "reason": "ZeRO-3 gradient growth and model-loading import failure are unrelated." + }, + { + "left": "issue:28218", + "right": "issue:34567", + "accept": false, + "reason": "Tokenizer spacing after added tokens is unrelated to TrainerState token-count tracking." + }, + { + "left": "issue:29159", + "right": "issue:32090", + "accept": false, + "reason": "Slow/fast tokenizer inconsistency and a Trainer NoneType broadcast error do not share the same bug." + }, + { + "left": "issue:45084", + "right": "issue:45520", + "accept": false, + "reason": "Template-node compilation errors and a Python 3.13 flash_attn import KeyError are different problems." + }, + { + "left": "issue:43582", + "right": "issue:44977", + "accept": false, + "reason": "Apple Silicon allocator TypeError and Qwen3.5 flash-attention generation bugs are not the same issue." + }, + { + "left": "issue:44554", + "right": "issue:45446", + "accept": false, + "reason": "MPS attention correctness and a flex_attention import version check are unrelated." + }, + { + "left": "issue:43540", + "right": "issue:45200", + "accept": false, + "reason": "Video input validation in Qwen3OmniMoe and Gemma4 text-only mm_token_type_ids handling are different bugs." + }, + { + "left": "issue:15354", + "right": "issue:38175", + "accept": false, + "reason": "TorchScript GeneratorExp export failure and SigLIP2 zero-probability behavior are separate issues." + }, + { + "left": "issue:43232", + "right": "issue:44849", + "accept": false, + "reason": "Generation kwargs update after sync_gpus and Qwen3.5 hidden-state behavior are different code paths." + }, + { + "left": "issue:37428", + "right": "issue:43957", + "accept": false, + "reason": "Missing flash-support import and meta-device model loading breakage are unrelated." + }, + { + "left": "issue:45072", + "right": "issue:45198", + "accept": false, + "reason": "BF16 dtype mismatches in inference and Wav2Vec2 save/tokenization failures are not the same bug." + }, + { + "left": "issue:45200", + "right": "issue:45823", + "accept": false, + "reason": "Both are Gemma4-related, but one is token-type defaults and the other is a device-map forward mismatch; not the same underlying defect." + }, + { + "left": "issue:43493", + "right": "issue:43701", + "accept": false, + "reason": "SigLIP2 implementation discrepancy and resume_from_checkpoint key mismatch are unrelated." + }, + { + "left": "issue:44315", + "right": "issue:44849", + "accept": false, + "reason": "Liger Kernel application in model_init and Qwen3.5 hidden-state behavior do not share a concrete bug." + }, + { + "left": "issue:43572", + "right": "issue:44987", + "accept": false, + "reason": "StableLm pad_token_idx regression and loading physical-intelligence/fast on newer transformers are different failures." + }, + { + "left": "issue:43295", + "right": "issue:44610", + "accept": false, + "reason": "Both involve processor behavior, but one is processor.tokenizer/image access regression and the other is a 640px-vs-224px processor/model size mismatch." + }, + { + "left": "issue:43295", + "right": "issue:43450", + "accept": false, + "reason": "Processor/tokenizer regression and batched video output shape bug are distinct." + }, + { + "left": "issue:44554", + "right": "issue:45356", + "accept": false, + "reason": "MPS attention correctness and Kimi tokenizer regression are separate issues." + }, + { + "left": "issue:43572", + "right": "issue:43957", + "accept": false, + "reason": "Pad token config regression and meta-device loading breakage are unrelated." + }, + { + "left": "issue:37428", + "right": "issue:43824", + "accept": false, + "reason": "Importing a missing flash-attention helper and missing Qwen2.5-VL class export are different import errors." + }, + { + "left": "issue:45200", + "right": "issue:45325", + "accept": false, + "reason": "Gemma4 token-type defaults and Qwen2.5-VL rope index scaling are unrelated multimodal bugs." + }, + { + "left": "issue:43295", + "right": "issue:43873", + "accept": false, + "reason": "Processor/tokenizer regression and quantization offloading behavior are not the same bug." + }, + { + "left": "issue:44466", + "right": "issue:45127", + "accept": false, + "reason": "lm_head.weight serialization inconsistency and LoRA merge collapse with extended vocab are different weight-handling issues." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 24, + "review_unit_count": 49, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-426", + "nodes": [ + "issue:28218", + "issue:35141", + "issue:37428", + "issue:38175", + "issue:42490", + "issue:42915", + "issue:43262", + "issue:43278", + "issue:43334", + "issue:43335", + "issue:43388", + "issue:43425", + "issue:43450", + "issue:43493", + "issue:43525", + "issue:43526", + "issue:43550", + "issue:43575", + "issue:43606", + "issue:43653", + "issue:43819", + "issue:43824", + "issue:43827", + "issue:43844", + "issue:43856", + "issue:43901", + "issue:43975", + "issue:44112", + "issue:44206", + "issue:44442", + "issue:44514", + "issue:44521", + "issue:44610", + "issue:44792", + "issue:44857", + "issue:44898", + "issue:45083", + "issue:45127", + "issue:45198", + "issue:45245", + "issue:45290", + "issue:45381", + "issue:45412", + "issue:45520", + "issue:45563", + "issue:45584", + "issue:45593", + "issue:46170" + ], + "soft_pairs": [ + "issue:44898|issue:45412", + "issue:43550|issue:43844", + "issue:43334|issue:43525", + "issue:44792|issue:45563", + "issue:43824|issue:43975", + "issue:42915|issue:43824", + "issue:43388|issue:43526", + "issue:43450|issue:43575", + "issue:42490|issue:43493", + "issue:43278|issue:43856", + "issue:43975|issue:45520", + "issue:28218|issue:35141", + "issue:43653|issue:44112", + "issue:38175|issue:43493", + "issue:44442|issue:45198", + "issue:44514|issue:44521", + "issue:45520|issue:46170", + "issue:38175|issue:45412", + "issue:44206|issue:45584", + "issue:43827|issue:43901", + "issue:43262|issue:43450", + "issue:43819|issue:44112", + "issue:43819|issue:45083", + "issue:44610|issue:44857", + "issue:37428|issue:43425", + "issue:43606|issue:45083", + "issue:43335|issue:43550", + "issue:44857|issue:45593", + "issue:45290|issue:45381", + "issue:45127|issue:45245" + ], + "prepared_review_unit_hash": "9278750a47ab5b50c09f75f696d438b801a26c18f744c2d5e754f71c0cb12001", + "component_budget": { + "node_count": 426, + "item_count": 426, + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 30, + "serialized_chars": 36847, + "estimated_input_tokens": 9212, + "estimated_eval_tokens": 18680 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "Mostly unrelated issues clustered by broad subsystem vocabulary. The strongest duplicate-like pair is the SigLIP2 discrepancy report and the zero-probability symptom report, but a few other title-level near-matches exist (notably the two pad_token_id AttributeErrors and the two apply_chat_template reports) without enough evidence here to merge them.", + "confidence": 0.71, + "canonical_issue_reason": "issue:43493 is the best anchor for the SigLIP2 bug family because it describes the broader HF-vs-JAX implementation mismatch; issue:38175 looks like a symptom of that same underlying discrepancy.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43493 is the strongest global issue candidate because it states the underlying SigLIP2 implementation mismatch more generally than the symptom report in issue:38175.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44898", + "right": "issue:45412", + "accept": false, + "reason": "Different models and failure modes: Perceiver resolution handling vs RT-DETR memory cleanup." + }, + { + "left": "issue:43550", + "right": "issue:43844", + "accept": false, + "reason": "Both are training/runtime failures, but the concrete code paths differ: Bamba torch.compile/SDPA vs ZeRO-3 gradient growth." + }, + { + "left": "issue:43334", + "right": "issue:43525", + "accept": false, + "reason": "Same missing-attribute shape, but different config classes and model families; not enough to treat as one bug." + }, + { + "left": "issue:44792", + "right": "issue:45563", + "accept": false, + "reason": "Janus image-generation test failure vs a stale paged-generate warning; not the same underlying issue." + }, + { + "left": "issue:43824", + "right": "issue:43975", + "accept": false, + "reason": "ImportError for a missing class vs a detokenization regression; no shared concrete code path is evident." + }, + { + "left": "issue:42915", + "right": "issue:43824", + "accept": false, + "reason": "Different models and symptoms: FineGrainedFP8Config failure vs Qwen2_5_VL import failure." + }, + { + "left": "issue:43388", + "right": "issue:43526", + "accept": false, + "reason": "Both involve labels, but one is gather_for_metrics batching and the other is BeitImageProcessorFast label reduction." + }, + { + "left": "issue:43450", + "right": "issue:43575", + "accept": false, + "reason": "Video batch-shape bug and Qwen2-57B OOM are unrelated." + }, + { + "left": "issue:42490", + "right": "issue:43493", + "accept": false, + "reason": "Local-path import warning vs SigLIP2 implementation discrepancy; no concrete overlap from the titles alone." + }, + { + "left": "issue:43278", + "right": "issue:43856", + "accept": false, + "reason": "Embedding dtype drift in eval and MoE training memory usage are different bugs." + }, + { + "left": "issue:43975", + "right": "issue:45520", + "accept": false, + "reason": "Detokenization regression and a Python 3.13 flash_attn import KeyError are unrelated." + }, + { + "left": "issue:28218", + "right": "issue:35141", + "accept": false, + "reason": "Tokenizer spacing after added tokens and output embedding reinitialization after resize are distinct issues." + }, + { + "left": "issue:43653", + "right": "issue:44112", + "accept": false, + "reason": "Tokenizer special-token registration bug vs GraniteSpeech CI device-override test failure." + }, + { + "left": "issue:38175", + "right": "issue:43493", + "accept": true, + "reason": "Both are SigLIP2 output-discrepancy reports, and the broader HF-vs-JAX mismatch plausibly explains the zero-probability symptom." + }, + { + "left": "issue:44442", + "right": "issue:45198", + "accept": false, + "reason": "FastSpeech2Conformer tokenizer loading and Wav2Vec2 save_pretrained/tokenization failures are different issues." + }, + { + "left": "issue:44514", + "right": "issue:44521", + "accept": false, + "reason": "Both are apply_chat_template reports, but one is a batched padding=False crash and the other is all-zero assistant masks; not enough to merge." + }, + { + "left": "issue:45520", + "right": "issue:46170", + "accept": false, + "reason": "Different platform/import failures: flash_attn on Python 3.13 vs Windows ROCm torch.distributed dependency issues." + }, + { + "left": "issue:38175", + "right": "issue:45412", + "accept": false, + "reason": "SigLIP2 zero-probability symptom and an RT-DETR memory leak are unrelated." + }, + { + "left": "issue:44206", + "right": "issue:45584", + "accept": false, + "reason": "Unsupported center argument in LasrFeatureExtractor is unrelated to Whisper empty-transcription handling." + }, + { + "left": "issue:43827", + "right": "issue:43901", + "accept": false, + "reason": "Both are docs issues, but they reference different APIs and different pages." + }, + { + "left": "issue:43262", + "right": "issue:43450", + "accept": false, + "reason": "Audio chat-template sampling-rate default and video processor batched-shape bug are unrelated." + }, + { + "left": "issue:43819", + "right": "issue:44112", + "accept": false, + "reason": "DAC latent reconstruction mismatch vs GraniteSpeech CI device override are unrelated." + }, + { + "left": "issue:43819", + "right": "issue:45083", + "accept": false, + "reason": "Different code paths: DAC latent conversion vs qwen3_omni_moe helper length calculation." + }, + { + "left": "issue:44610", + "right": "issue:44857", + "accept": false, + "reason": "Processor output resolution mismatch and LwDetrImageLoss AMP/CUDA crash are different bugs." + }, + { + "left": "issue:37428", + "right": "issue:43425", + "accept": false, + "reason": "Missing flash-attention import and Torch 2.10 incompatibility are not the same underlying problem." + }, + { + "left": "issue:43606", + "right": "issue:45083", + "accept": false, + "reason": "CPU offload device mismatch in bark-small and qwen3_omni_moe length helper behavior are unrelated." + }, + { + "left": "issue:43335", + "right": "issue:43550", + "accept": false, + "reason": "SwitchTransformers sparse-layer construction and Bamba torch.compile/SDPA failure do not share a concrete code path." + }, + { + "left": "issue:44857", + "right": "issue:45593", + "accept": false, + "reason": "AMP/CUDA crash in LwDetrImageLoss vs D-FINE auxiliary-loss behavior are unrelated." + }, + { + "left": "issue:45290", + "right": "issue:45381", + "accept": false, + "reason": "Tool-call assistant message crash and Qwen2.5-VL video vision_position_ids bug are different chat-template/video problems." + }, + { + "left": "issue:45127", + "right": "issue:45245", + "accept": false, + "reason": "LoRA merge collapse with extended vocab and a categories-limit runtime error are unrelated." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "The SigLIP2 pair looks plausibly related, but the writeup overstates the evidence by claiming it is the only pair that could be the same underlying issue. From the packet titles alone, there are other plausible near-matches (e.g. the two pad_token_id AttributeErrors and the two apply_chat_template issues), so the summary should be framed more cautiously. The soft-edge rejections are mostly conservative, but the exclusivity claim is too strong for the available evidence." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 25, + "review_unit_count": 49, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-426", + "nodes": [ + "issue:28218", + "issue:29127", + "issue:29942", + "issue:32090", + "issue:34689", + "issue:35141", + "issue:37428", + "issue:38175", + "issue:39692", + "issue:40444", + "issue:42490", + "issue:43278", + "issue:43335", + "issue:43504", + "issue:43526", + "issue:43550", + "issue:43606", + "issue:43611", + "issue:43653", + "issue:43701", + "issue:43720", + "issue:43819", + "issue:43825", + "issue:43854", + "issue:43931", + "issue:44112", + "issue:44206", + "issue:44292", + "issue:44351", + "issue:44360", + "issue:44448", + "issue:44493", + "issue:44561", + "issue:44743", + "issue:44779", + "issue:44855", + "issue:44857", + "issue:44977", + "issue:45072", + "issue:45083", + "issue:45200", + "issue:45229", + "issue:45310", + "issue:45478", + "issue:45520", + "issue:45588", + "issue:45823", + "issue:45859" + ], + "soft_pairs": [ + "issue:44448|issue:44743", + "issue:29942|issue:32090", + "issue:37428|issue:44351", + "issue:45200|issue:45588", + "issue:44206|issue:45823", + "issue:43611|issue:43931", + "issue:43278|issue:43825", + "issue:43335|issue:43653", + "issue:43550|issue:44112", + "issue:43819|issue:44493", + "issue:44855|issue:45083", + "issue:43335|issue:43504", + "issue:29127|issue:32090", + "issue:45588|issue:45859", + "issue:43335|issue:45083", + "issue:39692|issue:43526", + "issue:45310|issue:45478", + "issue:39692|issue:44493", + "issue:44292|issue:45520", + "issue:43720|issue:44112", + "issue:44448|issue:44779", + "issue:43335|issue:45072", + "issue:44360|issue:44561", + "issue:43335|issue:44112", + "issue:42490|issue:43854", + "issue:44977|issue:45229", + "issue:38175|issue:44493", + "issue:34689|issue:35141", + "issue:43606|issue:44857", + "issue:44448|issue:44977", + "issue:44857|issue:45072", + "issue:40444|issue:43526", + "issue:28218|issue:32090", + "issue:43701|issue:45083" + ], + "prepared_review_unit_hash": "b3261adc3d5b19a34802607b43742c17fc6869dd6ff152e6c6f60ca84aa5387e", + "component_budget": { + "node_count": 426, + "item_count": 426, + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 34, + "serialized_chars": 37900, + "estimated_input_tokens": 9475, + "estimated_eval_tokens": 19206 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These are mostly unrelated Transformers bug reports across different models, loaders, and execution modes. None of the provided soft edges look like the same underlying bug, so I would not merge any of them. The only clear duplicate in the set is 45310/45478, which appear to be the same Qwen3.5 MoE from_pretrained regression; 45310 is the better canonical anchor.", + "confidence": 0.89, + "canonical_issue_reason": "issue:45310 is the older, slightly fuller report of the exact duplicate pair with issue:45478 (same Qwen3.5 MoE from_pretrained error).", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45310 is the strongest canonical issue in the set because it has an exact duplicate and a concrete, reproducible loading failure; it is also the earliest report.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44448", + "right": "issue:44743", + "accept": false, + "reason": "Different bugs: Pegasus output drift vs Qwen3.5 recurrent-state reset in cached generation." + }, + { + "left": "issue:29942", + "right": "issue:32090", + "accept": false, + "reason": "Unrelated failures: FlashAttention 2 test regressions vs Trainer `_gpu_broadcast_one` NoneType error." + }, + { + "left": "issue:37428", + "right": "issue:44351", + "accept": false, + "reason": "Both are import errors, but for different missing symbols and different breakage points." + }, + { + "left": "issue:45200", + "right": "issue:45588", + "accept": false, + "reason": "Same broad Gemma4 area, but one is mm_token_type_ids defaults and the other is a flash-attention `s_aux=None` crash." + }, + { + "left": "issue:44206", + "right": "issue:45823", + "accept": false, + "reason": "Different code paths: unsupported feature-extractor arg vs device-map forward mismatch." + }, + { + "left": "issue:43611", + "right": "issue:43931", + "accept": false, + "reason": "Both are model-loading issues, but one is `base_model_prefix` handling and the other is a Qwen3-VL shape mismatch." + }, + { + "left": "issue:43278", + "right": "issue:43825", + "accept": false, + "reason": "Different regressions: dtype change during evaluate vs a pipeline error-message/deprecation issue." + }, + { + "left": "issue:43335", + "right": "issue:43653", + "accept": false, + "reason": "Different components and symptoms: SwitchTransformers config bug vs BigBirdTokenizer special-token registration." + }, + { + "left": "issue:43550", + "right": "issue:44112", + "accept": false, + "reason": "Different failures: Bamba torch.compile/SDPA behavior vs a GraniteSpeech CI device-override test." + }, + { + "left": "issue:43819", + "right": "issue:44493", + "accept": false, + "reason": "Different problems: DAC latent/STE mismatch vs unexpected key handling for position ids." + }, + { + "left": "issue:44855", + "right": "issue:45083", + "accept": false, + "reason": "Different issues: Python 3.13 import parsing error vs qwen3_omni_moe feature-length helper behavior." + }, + { + "left": "issue:43335", + "right": "issue:43504", + "accept": false, + "reason": "Different models and failures: SwitchTransformers sparse-layer creation vs BEiT legacy-field loading." + }, + { + "left": "issue:29127", + "right": "issue:32090", + "accept": false, + "reason": "Error-message clarity request vs a runtime NoneType broadcast failure; not the same bug." + }, + { + "left": "issue:45588", + "right": "issue:45859", + "accept": false, + "reason": "Both involve flash attention/offload, but they fail in different places with different root causes." + }, + { + "left": "issue:43335", + "right": "issue:45083", + "accept": false, + "reason": "SwitchTransformers config creation bug vs qwen3_omni_moe output-length helper behavior." + }, + { + "left": "issue:39692", + "right": "issue:43526", + "accept": false, + "reason": "Different areas: SigLIP2 docs/model mismatch vs BeitImageProcessorFast label reduction." + }, + { + "left": "issue:39692", + "right": "issue:44493", + "accept": false, + "reason": "Docs/example errors are unrelated to the position-id key regression." + }, + { + "left": "issue:44292", + "right": "issue:45520", + "accept": false, + "reason": "Different failures: Qwen-3-8B-NVFP4 runtime issue vs Python 3.13 flash_attn import KeyError." + }, + { + "left": "issue:43720", + "right": "issue:44112", + "accept": false, + "reason": "BitNet packed-weight loading bug vs stale CI device-override test; different underlying problems." + }, + { + "left": "issue:44448", + "right": "issue:44779", + "accept": false, + "reason": "Pegasus v4/v5 output difference vs Deepseek tokenizer regression; different subsystems." + }, + { + "left": "issue:43335", + "right": "issue:45072", + "accept": false, + "reason": "Same model family, but one is config construction and the other is bfloat16 inference dtype mismatch." + }, + { + "left": "issue:44360", + "right": "issue:44561", + "accept": false, + "reason": "DSA indexer missing ReLU is unrelated to the `is_torch_fx_available` v5 compatibility break." + }, + { + "left": "issue:43335", + "right": "issue:44112", + "accept": false, + "reason": "SwitchTransformers config bug and GraniteSpeech CI device-override failure are not the same issue." + }, + { + "left": "issue:42490", + "right": "issue:43854", + "accept": false, + "reason": "Local-path import warning vs model-loading failure in unit tests; no shared root cause." + }, + { + "left": "issue:44977", + "right": "issue:45229", + "accept": false, + "reason": "Flash-attention generation bug vs CUDA OOM during multi-GPU inference; different failure modes." + }, + { + "left": "issue:38175", + "right": "issue:44493", + "accept": false, + "reason": "SigLIP2 zero-probability output bug is unrelated to the position-id key warning/regression." + }, + { + "left": "issue:34689", + "right": "issue:35141", + "accept": false, + "reason": "Different regressions: Llama 3.2 Vision load break vs output embedding reinitialization after resize." + }, + { + "left": "issue:43606", + "right": "issue:44857", + "accept": false, + "reason": "CPU offload device mismatch vs float16 AMP crash in LwDetrImageLoss; unrelated code paths." + }, + { + "left": "issue:44448", + "right": "issue:44977", + "accept": false, + "reason": "Both are generation regressions, but on different models and different attention stacks." + }, + { + "left": "issue:44857", + "right": "issue:45072", + "accept": false, + "reason": "AMP loss crash vs dtype mismatch in a different model family; not the same underlying bug." + }, + { + "left": "issue:40444", + "right": "issue:43526", + "accept": false, + "reason": "Multi-image IterableDataset fine-tuning failure is unrelated to BeitImageProcessorFast `reduce_labels`." + }, + { + "left": "issue:28218", + "right": "issue:32090", + "accept": false, + "reason": "Tokenizer added-token spacing bug and Trainer broadcast error are unrelated." + }, + { + "left": "issue:43701", + "right": "issue:45083", + "accept": false, + "reason": "Checkpoint-key mismatch vs qwen3_omni_moe helper behavior; no shared 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": 1, + "component_count": 17, + "review_unit_index": 26, + "review_unit_count": 49, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-426", + "nodes": [ + "issue:10000", + "issue:15354", + "issue:25251", + "issue:29942", + "issue:32090", + "issue:33357", + "issue:35141", + "issue:37428", + "issue:41720", + "issue:42490", + "issue:43278", + "issue:43329", + "issue:43381", + "issue:43412", + "issue:43550", + "issue:43575", + "issue:43653", + "issue:43723", + "issue:43873", + "issue:43927", + "issue:43976", + "issue:44448", + "issue:44462", + "issue:44466", + "issue:44554", + "issue:44561", + "issue:44568", + "issue:44743", + "issue:44779", + "issue:44857", + "issue:44977", + "issue:44987", + "issue:45083", + "issue:45198", + "issue:45229", + "issue:45245", + "issue:45372", + "issue:45406", + "issue:45440", + "issue:45459", + "issue:45491", + "issue:45563", + "issue:45588", + "issue:45698", + "issue:45712", + "issue:45803", + "issue:45823" + ], + "soft_pairs": [ + "issue:43723|issue:44554", + "issue:43550|issue:45083", + "issue:32090|issue:33357", + "issue:29942|issue:35141", + "issue:37428|issue:44987", + "issue:44561|issue:44779", + "issue:43873|issue:45229", + "issue:43927|issue:45698", + "issue:44462|issue:44561", + "issue:41720|issue:45229", + "issue:44977|issue:45563", + "issue:44568|issue:45491", + "issue:43278|issue:43329", + "issue:10000|issue:15354", + "issue:42490|issue:43575", + "issue:45372|issue:45588", + "issue:43412|issue:45588", + "issue:45198|issue:45803", + "issue:44857|issue:45588", + "issue:43976|issue:45440", + "issue:45372|issue:45406", + "issue:45459|issue:45712", + "issue:43278|issue:43381", + "issue:45588|issue:45823", + "issue:44448|issue:44466", + "issue:43653|issue:44743", + "issue:45245|issue:45563", + "issue:25251|issue:32090" + ], + "prepared_review_unit_hash": "7eba210ddc744ebe434eb4989bdfeb06105cef80be654fc8075af137464da2bb", + "component_budget": { + "node_count": 426, + "item_count": 426, + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 28, + "serialized_chars": 35784, + "estimated_input_tokens": 8946, + "estimated_eval_tokens": 18148 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This is not a coherent duplicate cluster: the items span unrelated tokenizer, attention, export, loading, memory, and generation bugs. None of the soft pairs look like the same concrete underlying issue closely enough to merge.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43723", + "right": "issue:44554", + "accept": false, + "reason": "Tokenizer loading regression and MPS attention correctness bug are unrelated code paths." + }, + { + "left": "issue:43550", + "right": "issue:45083", + "accept": false, + "reason": "Torch.compile/SDPA failure for Bamba is unrelated to the qwen3_omni_moe feature-length helper bug." + }, + { + "left": "issue:32090", + "right": "issue:33357", + "accept": false, + "reason": "Trainer broadcast TypeError and MacOS bus error in CLIP model loading are different failures." + }, + { + "left": "issue:29942", + "right": "issue:35141", + "accept": false, + "reason": "Flash Attention 2 test failures and token-embedding reinitialization are separate subsystems." + }, + { + "left": "issue:37428", + "right": "issue:44987", + "accept": false, + "reason": "Missing flash-attention import symbol is not the same as a generic model-loading failure for physical-intelligence/fast." + }, + { + "left": "issue:44561", + "right": "issue:44779", + "accept": false, + "reason": "FX helper removal breaking trust_remote_code and Deepseek tokenizer regression are different bugs." + }, + { + "left": "issue:43873", + "right": "issue:45229", + "accept": false, + "reason": "Quantization/offloading behavior and Gemma4 multi-GPU OOM are related only broadly by memory usage, not the same bug." + }, + { + "left": "issue:43927", + "right": "issue:45698", + "accept": false, + "reason": "Custom token IDs being lost on save/load and loading the wrong custom module after save_pretrained are distinct serialization problems." + }, + { + "left": "issue:44462", + "right": "issue:44561", + "accept": false, + "reason": "AutoTokenizer ignoring tokenizer.json is unrelated to the FX-availability API removal regression." + }, + { + "left": "issue:41720", + "right": "issue:45229", + "accept": false, + "reason": "Qwen3 auto-device-map CUDA assert and Gemma4 multi-GPU OOM are different inference failures." + }, + { + "left": "issue:44977", + "right": "issue:45563", + "accept": false, + "reason": "Flash-attention generation bug and stale num_return_sequences warning do not share the same failure mode." + }, + { + "left": "issue:44568", + "right": "issue:45491", + "accept": false, + "reason": "Tokenizer BOS/EOS omission and Gemma3 NaN embeddings are unrelated." + }, + { + "left": "issue:43278", + "right": "issue:43329", + "accept": false, + "reason": "Embedding dtype drift in eval and undefined multimodal-token helpers in the video branch are separate bugs." + }, + { + "left": "issue:10000", + "right": "issue:15354", + "accept": false, + "reason": "German DistilBertModel issue and torch.jit GeneratorExp export failure are unrelated." + }, + { + "left": "issue:42490", + "right": "issue:43575", + "accept": false, + "reason": "Trailing-slash import warning and Qwen2 tensor-parallel OOM are not the same bug." + }, + { + "left": "issue:45372", + "right": "issue:45588", + "accept": false, + "reason": "Gemma4 processor import dependency failure and flash_attention.py s_aux=None crash are different code paths." + }, + { + "left": "issue:43412", + "right": "issue:45588", + "accept": false, + "reason": "Executorch export issues for gemma3n and sink-less flash-attention AttributeError are unrelated." + }, + { + "left": "issue:45198", + "right": "issue:45803", + "accept": false, + "reason": "Wav2Vec2 save/tokenization failure and Fuyu bare-except masking errors are different bugs." + }, + { + "left": "issue:44857", + "right": "issue:45588", + "accept": false, + "reason": "CUDA AMP crash in LwDetrImageLoss is unrelated to the flash_attention s_aux=None AttributeError." + }, + { + "left": "issue:43976", + "right": "issue:45440", + "accept": false, + "reason": "Python version compatibility issue and DeepSeekV3MoE divergence are unrelated." + }, + { + "left": "issue:45372", + "right": "issue:45406", + "accept": false, + "reason": "Gemma4 processor import error and Gemma4Processor missing _tokenizer in serve are related by model family only, not the same underlying defect." + }, + { + "left": "issue:45459", + "right": "issue:45712", + "accept": false, + "reason": "Tokenizer protobuf exception handling and leftover dummy classes are separate housekeeping bugs." + }, + { + "left": "issue:43278", + "right": "issue:43381", + "accept": false, + "reason": "Eval dtype change and gradient checkpointing in eval mode are unrelated." + }, + { + "left": "issue:45588", + "right": "issue:45823", + "accept": false, + "reason": "Flash-attention s_aux=None crash and Gemma4 PLE device mismatch are different forward-pass bugs." + }, + { + "left": "issue:44448", + "right": "issue:44466", + "accept": false, + "reason": "Pegasus v4/v5 output divergence and lm_head serialization inconsistency are separate regressions." + }, + { + "left": "issue:43653", + "right": "issue:44743", + "accept": false, + "reason": "BigBirdTokenizer special-token registration and Qwen3 recurrent-state reset are unrelated." + }, + { + "left": "issue:45245", + "right": "issue:45563", + "accept": false, + "reason": "2^24 category limit error and stale generate warning are not the same issue." + }, + { + "left": "issue:25251", + "right": "issue:32090", + "accept": false, + "reason": "Pipeline top_k nesting behavior and Trainer GPU broadcast TypeError 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": 1, + "component_count": 17, + "review_unit_index": 27, + "review_unit_count": 49, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-426", + "nodes": [ + "issue:29127", + "issue:29942", + "issue:30064", + "issue:32090", + "issue:33290", + "issue:33357", + "issue:34567", + "issue:36010", + "issue:39692", + "issue:42490", + "issue:42915", + "issue:42994", + "issue:43065", + "issue:43066", + "issue:43257", + "issue:43278", + "issue:43381", + "issue:43421", + "issue:43526", + "issue:43611", + "issue:43653", + "issue:43701", + "issue:43756", + "issue:43761", + "issue:43824", + "issue:43873", + "issue:43994", + "issue:44186", + "issue:44206", + "issue:44265", + "issue:44279", + "issue:44361", + "issue:44442", + "issue:44462", + "issue:44479", + "issue:44610", + "issue:44871", + "issue:44977", + "issue:45198", + "issue:45372", + "issue:45412", + "issue:45440", + "issue:45459", + "issue:45491", + "issue:45593", + "issue:45834", + "issue:45859", + "issue:46121" + ], + "soft_pairs": [ + "issue:44206|issue:44479", + "issue:43065|issue:43994", + "issue:45372|issue:45459", + "issue:44186|issue:45198", + "issue:44871|issue:44977", + "issue:43065|issue:45412", + "issue:43611|issue:46121", + "issue:44361|issue:45198", + "issue:42915|issue:42994", + "issue:42490|issue:43873", + "issue:32090|issue:33290", + "issue:33357|issue:39692", + "issue:43066|issue:44462", + "issue:29942|issue:30064", + "issue:29127|issue:34567", + "issue:43994|issue:45412", + "issue:44265|issue:45198", + "issue:43257|issue:43278", + "issue:43756|issue:45440", + "issue:43421|issue:45491", + "issue:43653|issue:44442", + "issue:43381|issue:43701", + "issue:43824|issue:44279", + "issue:44610|issue:45412", + "issue:45440|issue:45593", + "issue:29127|issue:33357", + "issue:29942|issue:36010", + "issue:43526|issue:45834", + "issue:43761|issue:45859" + ], + "prepared_review_unit_hash": "796b20125a79070af577c3e58d35d8d42e82efe2a0bf52a9b8fea327e45922c0", + "component_budget": { + "node_count": 426, + "item_count": 426, + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36334, + "estimated_input_tokens": 9084, + "estimated_eval_tokens": 18424 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is highly heterogeneous: the items span tokenizers, trainer state, model loading, quantization, vision processors, export, and memory/offload bugs. I do not see any pair that clearly describes the same underlying bug or fix, so all soft edges are rejected. If a representative issue is needed, issue 44462 is the strongest general loader-path bug among the set.", + "confidence": 0.95, + "canonical_issue_reason": "No single canonical duplicate emerges; the issues are unrelated across many subsystems and failure modes.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:44462 is a clear, user-facing, reproducible loading bug with broad impact in a common path, so it is the best representative issue in this mixed set.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44206", + "right": "issue:44479", + "accept": false, + "reason": "Different regressions in different code paths: feature extractor argument handling vs multimodal video-input behavior." + }, + { + "left": "issue:43065", + "right": "issue:43994", + "accept": false, + "reason": "Unrelated bugs: a dummy Conv2d in Sam3PixelDecoder vs incorrect SigLIP2 outputs with AutoModel/pipeline." + }, + { + "left": "issue:45372", + "right": "issue:45459", + "accept": false, + "reason": "Different tokenizer-loading failures: Gemma 4 processor import issue vs protobuf-missing decode error masking." + }, + { + "left": "issue:44186", + "right": "issue:45198", + "accept": false, + "reason": "Different model families and failures: LayoutLMv2 tokenizer NER/padding crash vs Wav2Vec2 save/tokenization failure." + }, + { + "left": "issue:44871", + "right": "issue:44977", + "accept": false, + "reason": "Different bugs: eos_token_id config mismatch vs flash-attention generation regression." + }, + { + "left": "issue:43065", + "right": "issue:45412", + "accept": false, + "reason": "Different problem classes: model layer placeholder vs RT-DETR memory release/GC issue." + }, + { + "left": "issue:43611", + "right": "issue:46121", + "accept": false, + "reason": "Different loading/config bugs: base_model_prefix breakage vs rope-validation TypeError on JSON-loaded lists." + }, + { + "left": "issue:44361", + "right": "issue:45198", + "accept": false, + "reason": "Different tokenizers and failure modes: MLuke task AttributeError vs Wav2Vec2 save/tokenization failure." + }, + { + "left": "issue:42915", + "right": "issue:42994", + "accept": false, + "reason": "Different quantization issues: FineGrainedFP8Config failure vs quantized model saving." + }, + { + "left": "issue:42490", + "right": "issue:43873", + "accept": false, + "reason": "Different paths: local-path slash import warning vs quantization/offloading behavior." + }, + { + "left": "issue:32090", + "right": "issue:33290", + "accept": false, + "reason": "Different failures: NoneType broadcast error in Trainer vs OOM with Adafactor and DeepSpeed." + }, + { + "left": "issue:33357", + "right": "issue:39692", + "accept": false, + "reason": "Different scope: MacOS bus error for a pretrained CLIP model vs documentation example/model mismatch and quantization errors for SigLIP2." + }, + { + "left": "issue:43066", + "right": "issue:44462", + "accept": false, + "reason": "Both involve tokenizer loading, but the concrete bugs differ: decoder type in v5 vs AutoTokenizer ignoring repository tokenizer.json." + }, + { + "left": "issue:29942", + "right": "issue:30064", + "accept": false, + "reason": "Unrelated test failure vs image processor void-segmentation-map handling." + }, + { + "left": "issue:29127", + "right": "issue:34567", + "accept": false, + "reason": "Different concerns: clearer error messaging for LayoutLMv3 boxes vs TrainerState token-count tracking." + }, + { + "left": "issue:43994", + "right": "issue:45412", + "accept": false, + "reason": "Different issues: SigLIP2 bad outputs vs RT-DETR memory leak." + }, + { + "left": "issue:44265", + "right": "issue:45198", + "accept": false, + "reason": "Different code paths: torch.export compatibility with torch_compilable_check vs Wav2Vec2 save/tokenization." + }, + { + "left": "issue:43257", + "right": "issue:43278", + "accept": false, + "reason": "Different failures: Qwen3 MoE weights not converting under accelerate+deepspeed vs dtype drift between train and eval." + }, + { + "left": "issue:43756", + "right": "issue:45440", + "accept": false, + "reason": "Different model-specific bugs: Smollm3 RoPE layer mismatch vs DeepseekV3MoE divergence from remote implementation." + }, + { + "left": "issue:43421", + "right": "issue:45491", + "accept": false, + "reason": "Different subsystems: runtime special-token post-processing vs Gemma3 NaN embeddings with sliding-window attention." + }, + { + "left": "issue:43653", + "right": "issue:44442", + "accept": false, + "reason": "Different tokenizer problems: BigBird mask-token registration vs AutoTokenizer loading FastSpeech2ConformerTokenizer." + }, + { + "left": "issue:43381", + "right": "issue:43701", + "accept": false, + "reason": "Different behaviors: gradient checkpointing in eval mode vs resume_from_checkpoint key mismatch." + }, + { + "left": "issue:43824", + "right": "issue:44279", + "accept": false, + "reason": "Different import/dependency issues: missing Qwen2.5-VL symbol vs generic transformers dependency problem." + }, + { + "left": "issue:44610", + "right": "issue:45412", + "accept": false, + "reason": "Different bugs: processor output size mismatch vs model memory not being released." + }, + { + "left": "issue:45440", + "right": "issue:45593", + "accept": false, + "reason": "Different model behavior: DeepseekV3MoE implementation divergence vs D-FINE auxiliary-loss omission." + }, + { + "left": "issue:29127", + "right": "issue:33357", + "accept": false, + "reason": "Different domains and failures: LayoutLMv3 error clarity vs MacOS bus error for CLIP." + }, + { + "left": "issue:29942", + "right": "issue:36010", + "accept": false, + "reason": "Different issues: Flash Attention 2 test failures vs GenerationMixin import error." + }, + { + "left": "issue:43526", + "right": "issue:45834", + "accept": false, + "reason": "Different processor bugs: BeitImageProcessorFast reduce_labels behavior vs Kosmos2.5 long-OCR index error." + }, + { + "left": "issue:43761", + "right": "issue:45859", + "accept": false, + "reason": "Different regressions: CLIPVisionModel hidden_states not returned vs Qwen3_5Moe CPU offload incompatibility." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 28, + "review_unit_count": 49, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-426", + "nodes": [ + "issue:29127", + "issue:33290", + "issue:34567", + "issue:34689", + "issue:35141", + "issue:42490", + "issue:42915", + "issue:43232", + "issue:43262", + "issue:43334", + "issue:43381", + "issue:43388", + "issue:43441", + "issue:43526", + "issue:43531", + "issue:43575", + "issue:43646", + "issue:43701", + "issue:43723", + "issue:43761", + "issue:43931", + "issue:44062", + "issue:44077", + "issue:44079", + "issue:44186", + "issue:44206", + "issue:44387", + "issue:44442", + "issue:44464", + "issue:44479", + "issue:44625", + "issue:44743", + "issue:44811", + "issue:44857", + "issue:44898", + "issue:44977", + "issue:45003", + "issue:45072", + "issue:45103", + "issue:45245", + "issue:45290", + "issue:45372", + "issue:45440", + "issue:45479", + "issue:45491", + "issue:45584", + "issue:45588", + "issue:45907" + ], + "soft_pairs": [ + "issue:45103|issue:45372", + "issue:43526|issue:44186", + "issue:44625|issue:45584", + "issue:29127|issue:35141", + "issue:43334|issue:43531", + "issue:44442|issue:45072", + "issue:42915|issue:43931", + "issue:44977|issue:45245", + "issue:43526|issue:44857", + "issue:42490|issue:43381", + "issue:44464|issue:45491", + "issue:43575|issue:43701", + "issue:43262|issue:44206", + "issue:43575|issue:44387", + "issue:44743|issue:44811", + "issue:43646|issue:44077", + "issue:44062|issue:45907", + "issue:43441|issue:45588", + "issue:43761|issue:44079", + "issue:43723|issue:45479", + "issue:45003|issue:45440", + "issue:44479|issue:45290", + "issue:43232|issue:43388", + "issue:44442|issue:44898", + "issue:33290|issue:34567", + "issue:33290|issue:34689" + ], + "prepared_review_unit_hash": "0c003826f099711a1df0ae44136d52b8a3bc9871b6cf8afef872dbaf053ab1e7", + "component_budget": { + "node_count": 426, + "item_count": 426, + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 26, + "serialized_chars": 35690, + "estimated_input_tokens": 8923, + "estimated_eval_tokens": 18102 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is heterogeneous: the soft pairs mostly connect unrelated bugs that only share broad themes like model loading, generation, or tokenizer behavior. I would not merge any of the proposed soft edges.", + "confidence": 0.66, + "canonical_issue_reason": "If one issue must represent the cluster, issue 45584 is the strongest standalone tracker: it is open, has inbound references, and describes a concrete reproducible runtime failure. That said, the items here are not true duplicates of one another.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 45584 is the best single issue to keep as a representative because it has clear user impact and the most discussion/inbound references among the open issues. It still should not absorb the others as duplicates.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45103", + "right": "issue:45372", + "accept": false, + "reason": "Different failures in different loading paths: docstring parsing with future annotations vs Gemma processor import dependency breakage." + }, + { + "left": "issue:43526", + "right": "issue:44186", + "accept": false, + "reason": "Different tokenizer/image-processor bugs affecting different models and inputs; no same concrete code path." + }, + { + "left": "issue:44625", + "right": "issue:45584", + "accept": false, + "reason": "Qwen3.5 config propagation bug vs Whisper empty-transcription generation failure are unrelated." + }, + { + "left": "issue:29127", + "right": "issue:35141", + "accept": false, + "reason": "One is a LayoutLMv3 error-message clarity issue, the other is an embedding reinitialization bug during resizing." + }, + { + "left": "issue:43334", + "right": "issue:43531", + "accept": false, + "reason": "Missing pad_token_id in Qwen3-VL text config is unrelated to Qwen3-MoE sliding-window behavior." + }, + { + "left": "issue:44442", + "right": "issue:45072", + "accept": false, + "reason": "AutoTokenizer loading failure and bfloat16 dtype mismatch are different bugs in different subsystems." + }, + { + "left": "issue:42915", + "right": "issue:43931", + "accept": false, + "reason": "Different Qwen model families and different failure modes: FP8 config handling vs weight-shape mismatch." + }, + { + "left": "issue:44977", + "right": "issue:45245", + "accept": false, + "reason": "Flash-attention generation regression is unrelated to the category-count runtime error." + }, + { + "left": "issue:43526", + "right": "issue:44857", + "accept": false, + "reason": "BeitImageProcessorFast label reduction bug and LwDetr AMP crash are distinct issues." + }, + { + "left": "issue:42490", + "right": "issue:43381", + "accept": false, + "reason": "Local-path import warning and gradient-checkpointing eval-mode restriction are not the same bug." + }, + { + "left": "issue:44464", + "right": "issue:45491", + "accept": false, + "reason": "Compiled-forward chunked generation inconsistency is separate from Gemma3 sliding-window NaN embeddings." + }, + { + "left": "issue:43575", + "right": "issue:43701", + "accept": false, + "reason": "Tensor-parallel OOM on model load and resume-from-checkpoint key mismatch are unrelated." + }, + { + "left": "issue:43262", + "right": "issue:44206", + "accept": false, + "reason": "Audio chat-template sampling-rate defaulting and LaserFeatureExtractor center-arg crash are different code paths." + }, + { + "left": "issue:43575", + "right": "issue:44387", + "accept": false, + "reason": "Both mention OOM, but one is tp load failure and the other is int4 quantization memory growth; not the same underlying bug." + }, + { + "left": "issue:44743", + "right": "issue:44811", + "accept": false, + "reason": "Recurrent-state cache reset in Qwen3.5 is unrelated to Whisper batch_decode skip_special_tokens behavior." + }, + { + "left": "issue:43646", + "right": "issue:44077", + "accept": false, + "reason": "General Transformers 5.0 custom-init breakage is broader than the specific patchtsmixer post_init policy issue." + }, + { + "left": "issue:44062", + "right": "issue:45907", + "accept": false, + "reason": "AddedToken special-flag collision and list_repo_files revision forwarding are unrelated." + }, + { + "left": "issue:43441", + "right": "issue:45588", + "accept": false, + "reason": "Different flash-attention crashes in different models/code paths; not mergeable as one fix." + }, + { + "left": "issue:43761", + "right": "issue:44079", + "accept": false, + "reason": "Missing hidden_states output in CLIPVisionModel is not the same as ModelOutput key assignment for previously-None keys." + }, + { + "left": "issue:43723", + "right": "issue:45479", + "accept": false, + "reason": "Tokenizer loading failure and degenerate zero-loss classification setup are unrelated." + }, + { + "left": "issue:45003", + "right": "issue:45440", + "accept": false, + "reason": "Unsafe sys.modules access in modeling_utils is unrelated to DeepseekV3MoE parity with remote implementation." + }, + { + "left": "issue:44479", + "right": "issue:45290", + "accept": false, + "reason": "Video-input regression in Qwen VL models and chat-template crash on tool-call assistant messages are different bugs." + }, + { + "left": "issue:43232", + "right": "issue:43388", + "accept": false, + "reason": "Generation bookkeeping after sync_gpus is unrelated to gather_for_metrics dropping tuple labels." + }, + { + "left": "issue:44442", + "right": "issue:44898", + "accept": false, + "reason": "Tokenizer loading failure is unrelated to Perceiver interpolation at non-default resolution." + }, + { + "left": "issue:33290", + "right": "issue:34567", + "accept": false, + "reason": "AdaFactor/deepspeed OOM and TrainerState token counter not updating are unrelated." + }, + { + "left": "issue:33290", + "right": "issue:34689", + "accept": false, + "reason": "Optimizer memory issue and Llama 3.2 Vision loading regression do not share 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": 1, + "component_count": 17, + "review_unit_index": 29, + "review_unit_count": 49, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-426", + "nodes": [ + "issue:25251", + "issue:28282", + "issue:33290", + "issue:34689", + "issue:35141", + "issue:36010", + "issue:36331", + "issue:42490", + "issue:42907", + "issue:43065", + "issue:43295", + "issue:43316", + "issue:43381", + "issue:43404", + "issue:43441", + "issue:43493", + "issue:43525", + "issue:43540", + "issue:43611", + "issue:43643", + "issue:43716", + "issue:43720", + "issue:43749", + "issue:43761", + "issue:43825", + "issue:43827", + "issue:43873", + "issue:43931", + "issue:44077", + "issue:44162", + "issue:44188", + "issue:44190", + "issue:44442", + "issue:44479", + "issue:44496", + "issue:44514", + "issue:44617", + "issue:44743", + "issue:44792", + "issue:44855", + "issue:44871", + "issue:44877", + "issue:44913", + "issue:45137", + "issue:45216", + "issue:45698", + "issue:45759", + "issue:45803" + ], + "soft_pairs": [ + "issue:43316|issue:43525", + "issue:42490|issue:44743", + "issue:33290|issue:35141", + "issue:43381|issue:43873", + "issue:44442|issue:45803", + "issue:43761|issue:44188", + "issue:34689|issue:43611", + "issue:43720|issue:44442", + "issue:44077|issue:44479", + "issue:43931|issue:45216", + "issue:43493|issue:43827", + "issue:42907|issue:45698", + "issue:34689|issue:36010", + "issue:43316|issue:45759", + "issue:33290|issue:36331", + "issue:43749|issue:43931", + "issue:43873|issue:44871", + "issue:43643|issue:44913", + "issue:44190|issue:44496", + "issue:44855|issue:45137", + "issue:25251|issue:28282", + "issue:43540|issue:44514", + "issue:44162|issue:44792", + "issue:43295|issue:43825", + "issue:43065|issue:44617", + "issue:43316|issue:44877", + "issue:43316|issue:43716", + "issue:43404|issue:43441" + ], + "prepared_review_unit_hash": "e12d8853e66d89f46dd445453e31cbe7e618e1abaa47f7ef3785c0bf68e8e667", + "component_budget": { + "node_count": 426, + "item_count": 426, + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 28, + "serialized_chars": 36197, + "estimated_input_tokens": 9050, + "estimated_eval_tokens": 18356 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is heterogeneous: most items are unrelated model-loading, config, tokenizer, or multimodal regressions that only share broad Transformers vocabulary. I do not see any pair that is the same underlying bug/change, so no soft edges should be merged.", + "confidence": 0.41, + "canonical_issue_reason": "Issue 43611 is the broadest representative of the set because it describes a general v5 model-loading regression, but the cluster as a whole is too mixed to have a strong true duplicate nucleus.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 43611 is the best single issue to anchor on because it is broad and systemic, whereas most others are narrower model-, config-, or processor-specific regressions.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43316", + "right": "issue:43525", + "accept": false, + "reason": "Both are config-related, but they affect different models and different fields; not the same bug." + }, + { + "left": "issue:42490", + "right": "issue:44743", + "accept": false, + "reason": "A local-path import warning and a recurrent-state cache bug are unrelated." + }, + { + "left": "issue:33290", + "right": "issue:35141", + "accept": false, + "reason": "Deepspeed/Adafactor OOM and embedding reinitialization in post_init are different code paths." + }, + { + "left": "issue:43381", + "right": "issue:43873", + "accept": false, + "reason": "Gradient-checkpointing eval-mode behavior and quantization offloading are separate issues." + }, + { + "left": "issue:44442", + "right": "issue:45803", + "accept": false, + "reason": "Tokenizer-loading failure and a bare-except bug in tensor conversion are not the same defect." + }, + { + "left": "issue:43761", + "right": "issue:44188", + "accept": false, + "reason": "CLIP hidden_states regression and attention-kernel divergence under torch.compile are unrelated." + }, + { + "left": "issue:34689", + "right": "issue:43611", + "accept": false, + "reason": "Both are loading regressions, but for different causes and model families; not mergeable as one fix." + }, + { + "left": "issue:43720", + "right": "issue:44442", + "accept": false, + "reason": "BitNet packed-weight loading and tokenizer class loading are unrelated." + }, + { + "left": "issue:44077", + "right": "issue:44479", + "accept": false, + "reason": "A model post_init policy issue and a video-input regression are different bugs." + }, + { + "left": "issue:43931", + "right": "issue:45216", + "accept": false, + "reason": "Shape mismatch on load and incorrect save_pretrained checkpoint contents are different failure modes." + }, + { + "left": "issue:43493", + "right": "issue:43827", + "accept": false, + "reason": "SigLIP2 implementation mismatch and outdated docs referencing pipeline are unrelated." + }, + { + "left": "issue:42907", + "right": "issue:45698", + "accept": false, + "reason": "Both touch save/load flows, but one is dequantized-model saving and the other is wrong custom-module loading after save_pretrained." + }, + { + "left": "issue:34689", + "right": "issue:36010", + "accept": false, + "reason": "A model-loading regression and a missing GenerationMixin import are not the same issue." + }, + { + "left": "issue:43316", + "right": "issue:45759", + "accept": false, + "reason": "Both involve config APIs, but they concern different models and different handling of text_config." + }, + { + "left": "issue:33290", + "right": "issue:36331", + "accept": false, + "reason": "DeepSpeed memory use and a custom trainer signature change are unrelated." + }, + { + "left": "issue:43749", + "right": "issue:43931", + "accept": false, + "reason": "FSDP CPU RAM efficient loading and Qwen3-VL shape mismatch are different loading bugs." + }, + { + "left": "issue:43873", + "right": "issue:44871", + "accept": false, + "reason": "Quantization offloading and eos_token_id config inconsistency do not share a concrete fix." + }, + { + "left": "issue:43643", + "right": "issue:44913", + "accept": false, + "reason": "Missing fields with trust_remote_code and GPTNeoX rotary_pct not persisting are different config-serialization problems." + }, + { + "left": "issue:44190", + "right": "issue:44496", + "accept": false, + "reason": "Local dataset loading in a training script and an unrecognized model config are unrelated." + }, + { + "left": "issue:44855", + "right": "issue:45137", + "accept": false, + "reason": "Python 3.13 import parsing and DeepSpeed ZeRO3 deque underflow are unrelated." + }, + { + "left": "issue:25251", + "right": "issue:28282", + "accept": false, + "reason": "Pipeline top_k shape change and missing PyTorch ImportError are unrelated." + }, + { + "left": "issue:43540", + "right": "issue:44514", + "accept": false, + "reason": "Both are Qwen multimodal processing bugs, but one is video-input handling and the other is batched chat-template padding behavior." + }, + { + "left": "issue:44162", + "right": "issue:44792", + "accept": false, + "reason": "ESM2 breakage and a janus image-generation test failure are unrelated." + }, + { + "left": "issue:43295", + "right": "issue:43825", + "accept": false, + "reason": "Processor/tokenizer regression and pipeline error-message wording are different issues." + }, + { + "left": "issue:43065", + "right": "issue:44617", + "accept": false, + "reason": "A dummy Conv2d in Sam3PixelDecoder and Sam3Video OOM are not the same bug." + }, + { + "left": "issue:43316", + "right": "issue:44877", + "accept": false, + "reason": "Both are config-related, but they affect different model families and different loading logic." + }, + { + "left": "issue:43316", + "right": "issue:43716", + "accept": false, + "reason": "Gemma3 config API inconsistency and Mistral-3 dtype mismatch are unrelated." + }, + { + "left": "issue:43404", + "right": "issue:43441", + "accept": false, + "reason": "One is lm_head weight tying, the other is FlashAttention failure; different code paths." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 30, + "review_unit_count": 49, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-426", + "nodes": [ + "issue:16998", + "issue:22355", + "issue:32090", + "issue:35141", + "issue:41950", + "issue:42617", + "issue:43295", + "issue:43388", + "issue:43452", + "issue:43475", + "issue:43504", + "issue:43525", + "issue:43531", + "issue:43550", + "issue:43575", + "issue:43606", + "issue:43646", + "issue:43673", + "issue:43688", + "issue:43701", + "issue:43742", + "issue:43761", + "issue:43906", + "issue:43931", + "issue:44060", + "issue:44077", + "issue:44242", + "issue:44464", + "issue:44479", + "issue:44496", + "issue:44534", + "issue:44617", + "issue:44734", + "issue:44743", + "issue:44792", + "issue:44871", + "issue:44913", + "issue:44977", + "issue:45127", + "issue:45200", + "issue:45290", + "issue:45406", + "issue:45563", + "issue:45584", + "issue:45698", + "issue:45803", + "issue:46050", + "issue:46170" + ], + "soft_pairs": [ + "issue:43550|issue:43701", + "issue:16998|issue:22355", + "issue:45803|issue:46170", + "issue:43606|issue:43701", + "issue:44077|issue:44534", + "issue:44871|issue:45584", + "issue:43673|issue:44743", + "issue:42617|issue:43575", + "issue:41950|issue:43475", + "issue:43575|issue:43742", + "issue:43295|issue:44871", + "issue:43388|issue:43688", + "issue:43761|issue:44479", + "issue:43452|issue:44496", + "issue:45127|issue:45200", + "issue:44871|issue:46050", + "issue:43646|issue:44913", + "issue:43504|issue:45698", + "issue:32090|issue:35141", + "issue:44734|issue:45406", + "issue:44617|issue:45584", + "issue:45563|issue:45584", + "issue:43525|issue:43531", + "issue:43761|issue:44242", + "issue:43931|issue:44060", + "issue:44792|issue:44977", + "issue:43906|issue:45290", + "issue:44479|issue:44734", + "issue:44464|issue:45563" + ], + "prepared_review_unit_hash": "fe936d6a3630a5c35d3ffbb0a03c2135a82934e771528c21ba14614057e613ff", + "component_budget": { + "node_count": 426, + "item_count": 426, + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 36443, + "estimated_input_tokens": 9111, + "estimated_eval_tokens": 18478 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is mostly a set of unrelated Transformers bug reports with superficial title similarity around loading, generation, and model-specific regressions. None of the soft pairs look like the same underlying issue, so all candidate duplicate edges are rejected.", + "confidence": 0.91, + "canonical_issue_reason": "issue:45698 is the best anchor in this set: it is open, has substantial discussion, and describes a concrete reproducible save/load regression affecting custom modules.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45698 is the most suitable representative issue because it is broad, actionable, and centered on a specific loading regression rather than a model-specific one-off failure.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43550", + "right": "issue:43701", + "accept": false, + "reason": "Different failures: Bamba torch.compile/SDPA runtime bug vs resume_from_checkpoint key mismatch." + }, + { + "left": "issue:16998", + "right": "issue:22355", + "accept": false, + "reason": "Unrelated topics: model_max_length question vs missing transformers.onnx import." + }, + { + "left": "issue:45803", + "right": "issue:46170", + "accept": false, + "reason": "Different subsystems: Fuyu tensor conversion exception handling vs Windows/ROCm distributed import failure." + }, + { + "left": "issue:43606", + "right": "issue:43701", + "accept": false, + "reason": "CPU offload device mismatch is unrelated to checkpoint key mismatches." + }, + { + "left": "issue:44077", + "right": "issue:44534", + "accept": false, + "reason": "One is a post_init API restriction, the other is non-persistent buffer initialization corruption." + }, + { + "left": "issue:44871", + "right": "issue:45584", + "accept": false, + "reason": "Gemma EOS config mismatch and Whisper empty-transcription crash are different bugs." + }, + { + "left": "issue:43673", + "right": "issue:44743", + "accept": false, + "reason": "Both involve cache semantics, but the models and concrete failure modes differ materially." + }, + { + "left": "issue:42617", + "right": "issue:43575", + "accept": false, + "reason": "3d_parallel.py execution error vs tensor-parallel OOM while loading a Qwen model." + }, + { + "left": "issue:41950", + "right": "issue:43475", + "accept": false, + "reason": "Pipeline processor lookup bug is unrelated to SAM3 vision encoder missing attribute." + }, + { + "left": "issue:43575", + "right": "issue:43742", + "accept": false, + "reason": "OOM on tp load is unrelated to a MobileLLM config/key loading error." + }, + { + "left": "issue:43295", + "right": "issue:44871", + "accept": false, + "reason": "Custom processor/tokenizer regression and Gemma EOS list-vs-scalar mismatch are distinct issues." + }, + { + "left": "issue:43388", + "right": "issue:43688", + "accept": false, + "reason": "Metric gathering label truncation is unrelated to auxiliary loss normalization." + }, + { + "left": "issue:43761", + "right": "issue:44479", + "accept": false, + "reason": "CLIPVision hidden_states bug and Qwen VL video-input regression affect different code paths." + }, + { + "left": "issue:43452", + "right": "issue:44496", + "accept": false, + "reason": "Both are loading-related, but gguf_file handling and unrecognized model_type are different root causes." + }, + { + "left": "issue:45127", + "right": "issue:45200", + "accept": false, + "reason": "LoRA/tied-embeddings collapse is unrelated to Gemma 4 mm_token_type_ids defaults." + }, + { + "left": "issue:44871", + "right": "issue:46050", + "accept": false, + "reason": "EOS token config mismatch is unrelated to sinq save/load quantization breakage." + }, + { + "left": "issue:43646", + "right": "issue:44913", + "accept": false, + "reason": "General custom model init regression is not the same as GPTNeoX rotary_pct persistence on reload." + }, + { + "left": "issue:43504", + "right": "issue:45698", + "accept": false, + "reason": "Both involve loading, but BEiT legacy-field preset loading and wrong custom-module selection are different bugs." + }, + { + "left": "issue:32090", + "right": "issue:35141", + "accept": false, + "reason": "Trainer gpu_broadcast NoneType error is unrelated to embedding reinitialization in post_init." + }, + { + "left": "issue:44734", + "right": "issue:45406", + "accept": false, + "reason": "Serve KV-cache indexing crash and Gemma4Processor missing _tokenizer are unrelated." + }, + { + "left": "issue:44617", + "right": "issue:45584", + "accept": false, + "reason": "CUDA OOM in Sam3Video is unrelated to Whisper empty transcription after align_special_tokens." + }, + { + "left": "issue:45563", + "right": "issue:45584", + "accept": false, + "reason": "A stale warning about num_return_sequences is not the same as a transcription crash." + }, + { + "left": "issue:43525", + "right": "issue:43531", + "accept": false, + "reason": "Missing pad_token_id and sliding_window behavior in different models are not the same defect." + }, + { + "left": "issue:43761", + "right": "issue:44242", + "accept": false, + "reason": "Vision hidden_states regression is unrelated to load balancing loss being omitted." + }, + { + "left": "issue:43931", + "right": "issue:44060", + "accept": false, + "reason": "Qwen3-VL weight shape mismatch and Qwen3-Next tied-weights warning are different model-loading issues." + }, + { + "left": "issue:44792", + "right": "issue:44977", + "accept": false, + "reason": "Janus image-generation test failure is unrelated to Qwen3.5 flash-attention generation issues." + }, + { + "left": "issue:43906", + "right": "issue:45290", + "accept": false, + "reason": "An isolated repro of another issue is not the same as the assistant-tool-call crash in apply_chat_template." + }, + { + "left": "issue:44479", + "right": "issue:44734", + "accept": false, + "reason": "Video input regression for Qwen VL models and serve KV-cache continuation indexing are distinct problems." + }, + { + "left": "issue:44464", + "right": "issue:45563", + "accept": false, + "reason": "Chunked generation inconsistency with compiled forward is unrelated to a stale warning on num_return_sequences." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 31, + "review_unit_count": 49, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-426", + "nodes": [ + "issue:35141", + "issue:36246", + "issue:41628", + "issue:41720", + "issue:42490", + "issue:43065", + "issue:43116", + "issue:43335", + "issue:43479", + "issue:43540", + "issue:43550", + "issue:43644", + "issue:43650", + "issue:43827", + "issue:43976", + "issue:43994", + "issue:44077", + "issue:44112", + "issue:44190", + "issue:44220", + "issue:44462", + "issue:44521", + "issue:44534", + "issue:44610", + "issue:44734", + "issue:44737", + "issue:44792", + "issue:44855", + "issue:44877", + "issue:44913", + "issue:44933", + "issue:44938", + "issue:44964", + "issue:45127", + "issue:45245", + "issue:45276", + "issue:45357", + "issue:45362", + "issue:45375", + "issue:45381", + "issue:45542", + "issue:45584", + "issue:45588", + "issue:45600", + "issue:45803", + "issue:45907", + "issue:46082", + "issue:46121" + ], + "soft_pairs": [ + "issue:44190|issue:44964", + "issue:44534|issue:44913", + "issue:43827|issue:43994", + "issue:44792|issue:45600", + "issue:44521|issue:45381", + "issue:41720|issue:45245", + "issue:44855|issue:45588", + "issue:43065|issue:44610", + "issue:44933|issue:45362", + "issue:35141|issue:42490", + "issue:44077|issue:44877", + "issue:43550|issue:44190", + "issue:43335|issue:44855", + "issue:43644|issue:44913", + "issue:43976|issue:44855", + "issue:45127|issue:45357", + "issue:44737|issue:45803", + "issue:43540|issue:44734", + "issue:44462|issue:45907", + "issue:44938|issue:45542", + "issue:44112|issue:45600", + "issue:43335|issue:44190", + "issue:46082|issue:46121", + "issue:43116|issue:44190", + "issue:45276|issue:45584", + "issue:43479|issue:45375", + "issue:36246|issue:41628", + "issue:43650|issue:44220" + ], + "prepared_review_unit_hash": "e5791d792266579718867401756dce094cc096600cdfcab5a8d5ee5e82ba224b", + "component_budget": { + "node_count": 426, + "item_count": 426, + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 28, + "serialized_chars": 36181, + "estimated_input_tokens": 9046, + "estimated_eval_tokens": 18348 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is highly heterogeneous: the soft pairs span unrelated bugs in imports, configs, multimodal models, CI, and docs. I would not merge any of the suggested soft pairs as duplicates. If a single issue must represent the cluster, issue 44521 is the most active and broadly scoped, but the set is not really cohesive.", + "confidence": 0.9, + "canonical_issue_reason": "Issue 44521 is the strongest representative by activity and breadth (open, many discussion entries, multiple inbound refs), though the overall cluster is not semantically unified.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 44521 is the best duplicate-triage anchor among this set because it has the most ongoing discussion and cross-references; the other issues are mostly one-off, narrow reports in unrelated areas.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44190", + "right": "issue:44964", + "accept": false, + "reason": "Local dataset loading in an example script vs a model import failure; different code paths and symptoms." + }, + { + "left": "issue:44534", + "right": "issue:44913", + "accept": false, + "reason": "Both involve configuration/state, but one is about non-persistent buffers being filled with junk and the other about a config field not surviving reload." + }, + { + "left": "issue:43827", + "right": "issue:43994", + "accept": false, + "reason": "Docs still mentioning pipeline() vs incorrect model outputs from AutoModel/pipeline; unrelated bug classes." + }, + { + "left": "issue:44792", + "right": "issue:45600", + "accept": false, + "reason": "Janus image-generation test failure vs auto_mappings referencing removed Sam3LiteText configs; different models and failures." + }, + { + "left": "issue:44521", + "right": "issue:45381", + "accept": false, + "reason": "Both are multimodal, but one is all-zero assistant masks in chat templating and the other is wrong vision_position_ids for Qwen2.5-VL video input." + }, + { + "left": "issue:41720", + "right": "issue:45245", + "accept": false, + "reason": "CUDA assert during Qwen3 auto device mapping is unrelated to the category-count limit error." + }, + { + "left": "issue:44855", + "right": "issue:45588", + "accept": false, + "reason": "Python 3.13 import parse failure in DebertaV2 vs flash_attention crash on s_aux=None; distinct root causes." + }, + { + "left": "issue:43065", + "right": "issue:44610", + "accept": false, + "reason": "Dummy Conv2d in Sam3PixelDecoder vs processor output size mismatch in OmDet-Turbo; different subsystems." + }, + { + "left": "issue:44933", + "right": "issue:45362", + "accept": false, + "reason": "Missing import from image_utils vs Qwen3.5 chat crash; unrelated issues." + }, + { + "left": "issue:35141", + "right": "issue:42490", + "accept": false, + "reason": "Embedding reinit after resize_token_embeddings vs import warning for paths ending with a slash; no shared code path." + }, + { + "left": "issue:44077", + "right": "issue:44877", + "accept": false, + "reason": "Optional post_init support vs strict config loading for granite_speech; related only at a very broad config level." + }, + { + "left": "issue:43550", + "right": "issue:44190", + "accept": false, + "reason": "torch.compile/SDPA failure in Bamba vs local dataset loading in an example script; unrelated." + }, + { + "left": "issue:43335", + "right": "issue:44855", + "accept": false, + "reason": "SwitchTransformers sparse-layer construction bug vs Python 3.13 indentation/import error; different areas entirely." + }, + { + "left": "issue:43644", + "right": "issue:44913", + "accept": false, + "reason": "Non-persistent buffers initialized with junk vs rotary_pct not persisting on reload; both serialization-adjacent but not the same bug." + }, + { + "left": "issue:43976", + "right": "issue:44855", + "accept": false, + "reason": "General Python-version compatibility issue vs a specific decorator/comment parsing failure; not the same underlying defect." + }, + { + "left": "issue:45127", + "right": "issue:45357", + "accept": false, + "reason": "LoRA merge collapse with extended vocab vs incorrect visual encoder keys on save_pretrained; both Qwen-related but different code paths." + }, + { + "left": "issue:44737", + "right": "issue:45803", + "accept": false, + "reason": "XLNet device placement bug vs bare except swallowing errors in FuyuBatchFeature; unrelated." + }, + { + "left": "issue:43540", + "right": "issue:44734", + "accept": false, + "reason": "Qwen3OmniMoe video-input ValueError vs transformers serve KV-cache continuation crash; different execution paths." + }, + { + "left": "issue:44462", + "right": "issue:45907", + "accept": false, + "reason": "AutoTokenizer ignoring tokenizer.json vs list_repo_files failing to forward revision; different Hub API behaviors." + }, + { + "left": "issue:44938", + "right": "issue:45542", + "accept": false, + "reason": "Python 3.14 loading failure vs TensorBoard/TensorFlow backend mismatch; environment-related but distinct." + }, + { + "left": "issue:44112", + "right": "issue:45600", + "accept": false, + "reason": "GraniteSpeech CI device override flake vs Sam3LiteText auto-mapping breakage; no shared bug." + }, + { + "left": "issue:43335", + "right": "issue:44190", + "accept": false, + "reason": "SwitchTransformers config bug vs example script dataset loading issue; unrelated." + }, + { + "left": "issue:46082", + "right": "issue:46121", + "accept": false, + "reason": "Llama head_dim validation vs rope param validation on JSON-loaded lists; both config bugs, but different validation logic and symptoms." + }, + { + "left": "issue:43116", + "right": "issue:44190", + "accept": false, + "reason": "Multi-label classification example returns empty results vs local dataset loading failure; separate example-script bugs." + }, + { + "left": "issue:45276", + "right": "issue:45584", + "accept": false, + "reason": "Gemma4 resize_token_embeddings not updating outputs vs Whisper empty-transcription generation failure; unrelated." + }, + { + "left": "issue:43479", + "right": "issue:45375", + "accept": false, + "reason": "Phi4MultimodalConfig defaulting None fields vs missing deepstack_visual_indexes in Qwen3_5MoeVisionConfig; both config issues, but not the same one." + }, + { + "left": "issue:36246", + "right": "issue:41628", + "accept": false, + "reason": "Two different import errors for different symbols; same broad symptom, but not the same missing export." + }, + { + "left": "issue:43650", + "right": "issue:44220", + "accept": false, + "reason": "A non-informative placeholder issue vs an audio feature extraction bug; no duplicate evidence." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 32, + "review_unit_count": 49, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-426", + "nodes": [ + "issue:36246", + "issue:38175", + "issue:38617", + "issue:41950", + "issue:42175", + "issue:42490", + "issue:42907", + "issue:43065", + "issue:43295", + "issue:43299", + "issue:43335", + "issue:43404", + "issue:43441", + "issue:43526", + "issue:43531", + "issue:43643", + "issue:43673", + "issue:43824", + "issue:43854", + "issue:43901", + "issue:43976", + "issue:44016", + "issue:44062", + "issue:44112", + "issue:44186", + "issue:44220", + "issue:44246", + "issue:44279", + "issue:44336", + "issue:44496", + "issue:44561", + "issue:44683", + "issue:44734", + "issue:44933", + "issue:44964", + "issue:44987", + "issue:45003", + "issue:45042", + "issue:45216", + "issue:45245", + "issue:45276", + "issue:45335", + "issue:45446", + "issue:45584", + "issue:45593", + "issue:45600", + "issue:45632" + ], + "soft_pairs": [ + "issue:43295|issue:45276", + "issue:45245|issue:45584", + "issue:41950|issue:44016", + "issue:42175|issue:44279", + "issue:36246|issue:38617", + "issue:44186|issue:44734", + "issue:44683|issue:45446", + "issue:44246|issue:45003", + "issue:44016|issue:44220", + "issue:43404|issue:45216", + "issue:38175|issue:43065", + "issue:44734|issue:45632", + "issue:43643|issue:44496", + "issue:44964|issue:45600", + "issue:44561|issue:45632", + "issue:43854|issue:44964", + "issue:44112|issue:44964", + "issue:43976|issue:44987", + "issue:42907|issue:43441", + "issue:42490|issue:45632", + "issue:43531|issue:45600", + "issue:43295|issue:45335", + "issue:43526|issue:43901", + "issue:44496|issue:45042", + "issue:44016|issue:44062", + "issue:43673|issue:44336", + "issue:43824|issue:44933", + "issue:43335|issue:45593", + "issue:43299|issue:45216" + ], + "prepared_review_unit_hash": "b2be04d4bdff9f6e54de0ce8086bf879923d9db3ae803b3008e0d41c466f979f", + "component_budget": { + "node_count": 426, + "item_count": 426, + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 29, + "serialized_chars": 35785, + "estimated_input_tokens": 8947, + "estimated_eval_tokens": 18150 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This set is mostly heterogeneous: import regressions, model-specific loading bugs, tokenizer/pipeline issues, and unrelated docs/CI reports. None of the soft pairs look like true duplicates; the closest themes still fail the strict same-code-path test. The clearest representative issue is the Qwen2.5-VL import regression.", + "confidence": 0.93, + "canonical_issue_reason": "Issue 43824 is the most concrete, well-scoped representative: a specific top-level import regression with clear affected symbol and inbound references.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 43824 is the best single issue to anchor this cluster because it is explicit, reproducible, and broadly recognizable as a Transformers export/import regression.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43295", + "right": "issue:45276", + "accept": false, + "reason": "Both mention resize/tokenization-related regressions, but they affect different model stacks and different code paths." + }, + { + "left": "issue:45245", + "right": "issue:45584", + "accept": false, + "reason": "One is a category-cardinality runtime error; the other is a Whisper generation failure after token alignment. Not the same bug." + }, + { + "left": "issue:41950", + "right": "issue:44016", + "accept": false, + "reason": "A pipeline processor lookup bug and a notebook syntax error are unrelated." + }, + { + "left": "issue:42175", + "right": "issue:44279", + "accept": false, + "reason": "Dependency/backend packaging trouble vs a generic dependency issue; too broad and not the same concrete failure." + }, + { + "left": "issue:36246", + "right": "issue:38617", + "accept": false, + "reason": "Both are ImportErrors, but for different missing symbols in different modules and different fixes." + }, + { + "left": "issue:44186", + "right": "issue:44734", + "accept": false, + "reason": "Tokenizer padding/NER crashes and serve KV-cache continuation indexing are unrelated code paths." + }, + { + "left": "issue:44683", + "right": "issue:45446", + "accept": false, + "reason": "Both involve flex_attention, but one is a torch-2.9 compiled failure and the other is a version-check/import guard issue." + }, + { + "left": "issue:44246", + "right": "issue:45003", + "accept": false, + "reason": "Import slowness and unsafe sys.modules access are adjacent, but not clearly the same defect." + }, + { + "left": "issue:44016", + "right": "issue:44220", + "accept": false, + "reason": "Notebook syntax error vs audio feature extraction issue; unrelated." + }, + { + "left": "issue:43404", + "right": "issue:45216", + "accept": false, + "reason": "Different model families and different bugs: untied weights vs incorrect save_pretrained output." + }, + { + "left": "issue:38175", + "right": "issue:43065", + "accept": false, + "reason": "A probability anomaly in SIGLIP2 is unrelated to a dummy Conv2d in Sam3PixelDecoder." + }, + { + "left": "issue:44734", + "right": "issue:45632", + "accept": false, + "reason": "Serve KV-cache continuation bug and local trust_remote_code cache collision are different subsystems." + }, + { + "left": "issue:43643", + "right": "issue:44496", + "accept": false, + "reason": "One is missing fields with trust_remote_code; the other is an unrecognized model lacking model_type. Similar category, different bug." + }, + { + "left": "issue:44964", + "right": "issue:45600", + "accept": false, + "reason": "Both concern model loading, but one is a Phi-4 multimodal load failure and the other is stale auto_mappings references." + }, + { + "left": "issue:44561", + "right": "issue:45632", + "accept": false, + "reason": "Both mention trust_remote_code, but one is a removed API break and the other is a cache-path collision." + }, + { + "left": "issue:43854", + "right": "issue:44964", + "accept": false, + "reason": "Two different model-loading failures for different models; no shared concrete code path." + }, + { + "left": "issue:44112", + "right": "issue:44964", + "accept": false, + "reason": "A CI test flake in GraniteSpeech is unrelated to a multimodal model loading failure." + }, + { + "left": "issue:43976", + "right": "issue:44987", + "accept": false, + "reason": "Python compatibility and a model-load regression are not the same underlying issue." + }, + { + "left": "issue:42907", + "right": "issue:43441", + "accept": false, + "reason": "Both involve Ministral, but one is save/dequantization and the other is FlashAttention loading; different fixes." + }, + { + "left": "issue:42490", + "right": "issue:45632", + "accept": false, + "reason": "Both involve local paths, but one is a warning on trailing slash and the other is a leaf-directory cache collision." + }, + { + "left": "issue:43531", + "right": "issue:45600", + "accept": false, + "reason": "Qwen3-MoE sliding-window behavior and Sam3LiteText auto-mapping cleanup are unrelated." + }, + { + "left": "issue:43295", + "right": "issue:45335", + "accept": false, + "reason": "Both mention regression around embeddings/tokenizer APIs, but the affected models and failure modes differ." + }, + { + "left": "issue:43526", + "right": "issue:43901", + "accept": false, + "reason": "An image-processor label bug and a documentation mismatch in text classification are unrelated." + }, + { + "left": "issue:44496", + "right": "issue:45042", + "accept": false, + "reason": "Missing model_type in config and PIL image processor torchvision requirements are different load-time problems." + }, + { + "left": "issue:44016", + "right": "issue:44062", + "accept": false, + "reason": "Notebook syntax error and AddedToken argument collision are unrelated." + }, + { + "left": "issue:43673", + "right": "issue:44336", + "accept": false, + "reason": "Generation cache behavior and ANSI log formatting are entirely separate concerns." + }, + { + "left": "issue:43824", + "right": "issue:44933", + "accept": false, + "reason": "Both are import-related, but one is a missing top-level export and the other is a nonexistent image_utils import; not the same bug." + }, + { + "left": "issue:43335", + "right": "issue:45593", + "accept": false, + "reason": "SwitchTransformers config construction and D-FINE auxiliary-loss behavior are unrelated model-specific issues." + }, + { + "left": "issue:43299", + "right": "issue:45216", + "accept": false, + "reason": "Qwen3VL loading regression and Qwen3.5 save_pretrained regression are different failure modes." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 33, + "review_unit_count": 49, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-426", + "nodes": [ + "issue:38175", + "issue:39692", + "issue:41950", + "issue:43295", + "issue:43299", + "issue:43381", + "issue:43441", + "issue:43450", + "issue:43479", + "issue:43525", + "issue:43577", + "issue:43650", + "issue:43854", + "issue:43901", + "issue:43976", + "issue:44016", + "issue:44075", + "issue:44242", + "issue:44279", + "issue:44291", + "issue:44297", + "issue:44351", + "issue:44368", + "issue:44462", + "issue:44496", + "issue:44521", + "issue:44617", + "issue:44683", + "issue:44779", + "issue:45020", + "issue:45081", + "issue:45200", + "issue:45216", + "issue:45278", + "issue:45325", + "issue:45335", + "issue:45357", + "issue:45405", + "issue:45542", + "issue:45584", + "issue:45593", + "issue:45759", + "issue:45812", + "issue:45823", + "issue:45834", + "issue:45907", + "issue:46132" + ], + "soft_pairs": [ + "issue:44075|issue:44368", + "issue:43976|issue:44279", + "issue:43381|issue:45593", + "issue:43901|issue:44496", + "issue:44496|issue:45759", + "issue:45335|issue:45584", + "issue:43299|issue:45357", + "issue:43479|issue:43525", + "issue:44521|issue:45325", + "issue:44242|issue:45200", + "issue:44351|issue:45278", + "issue:41950|issue:45020", + "issue:41950|issue:45823", + "issue:43577|issue:45216", + "issue:44779|issue:45812", + "issue:44683|issue:45542", + "issue:43976|issue:45405", + "issue:38175|issue:44617", + "issue:44279|issue:45405", + "issue:39692|issue:43901", + "issue:43299|issue:43854", + "issue:43295|issue:44779", + "issue:43650|issue:46132", + "issue:44462|issue:45081", + "issue:44297|issue:44462", + "issue:45081|issue:45907", + "issue:43450|issue:44279", + "issue:44016|issue:44291", + "issue:41950|issue:43650", + "issue:43441|issue:45834" + ], + "prepared_review_unit_hash": "2a7d4e526d59f7d5f90c48ddc3ba6db3d68d076605ce6c32347da593e2366f2c", + "component_budget": { + "node_count": 426, + "item_count": 426, + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 30, + "serialized_chars": 36151, + "estimated_input_tokens": 9038, + "estimated_eval_tokens": 18332 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "No convincing duplicate cluster here: the soft pairs mostly connect unrelated bugs across different models, pipelines, tokenizers, and docs. I would not merge any of the candidate pairs.", + "confidence": 0.94, + "canonical_issue_reason": "issue:44521 is the strongest representative item in this set: it is active, has multiple inbound references, and describes a concrete reproducible multimodal bug.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:44521 is the best global issue candidate because it has the clearest ongoing impact and discussion footprint among otherwise unrelated reports.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44075", + "right": "issue:44368", + "accept": false, + "reason": "Different problems: optimizer argument handling vs a tie_word_embeddings warning in a Qwen fine-tuning workflow." + }, + { + "left": "issue:43976", + "right": "issue:44279", + "accept": false, + "reason": "One is a Python/version compatibility report; the other is a vague dependency issue. No shared code-path bug." + }, + { + "left": "issue:43381", + "right": "issue:45593", + "accept": false, + "reason": "Gradient checkpointing in eval mode is unrelated to D-FINE auxiliary-loss behavior when denoising is off." + }, + { + "left": "issue:43901", + "right": "issue:44496", + "accept": false, + "reason": "Docs mismatch for text classification is unrelated to model loading failing for a specific repository." + }, + { + "left": "issue:44496", + "right": "issue:45759", + "accept": false, + "reason": "Both involve model loading, but different mechanisms and configs: unrecognized model vs unwrapping text_config for multimodal Qwen configs." + }, + { + "left": "issue:45335", + "right": "issue:45584", + "accept": false, + "reason": "Unrelated bugs: embedding resize propagation in t5gemma vs Whisper empty-transcription generation after token alignment." + }, + { + "left": "issue:43299", + "right": "issue:45357", + "accept": false, + "reason": "Both are Qwen-related but distinct regressions affecting loading vs save_pretrained output structure." + }, + { + "left": "issue:43479", + "right": "issue:43525", + "accept": false, + "reason": "Different config construction bugs: missing default multimodal subconfigs vs missing pad_token_id on Llama4Config." + }, + { + "left": "issue:44521", + "right": "issue:45325", + "accept": false, + "reason": "Different multimodal chat-template/mask bug vs a Qwen2.5-VL rope index scaling issue." + }, + { + "left": "issue:44242", + "right": "issue:45200", + "accept": false, + "reason": "Load-balancing loss omission is a different training-loss bug from Gemma 4 mm_token_type_ids defaults." + }, + { + "left": "issue:44351", + "right": "issue:45278", + "accept": false, + "reason": "Importing HybridCache is a specific symbol regression; the other is a broad import-error report after upgrade." + }, + { + "left": "issue:41950", + "right": "issue:45020", + "accept": false, + "reason": "Video-classification processor lookup is unrelated to remote_code breakage." + }, + { + "left": "issue:41950", + "right": "issue:45823", + "accept": false, + "reason": "Different Gemma/PLE device-placement bug vs video-classification processor selection bug." + }, + { + "left": "issue:43577", + "right": "issue:45216", + "accept": false, + "reason": "Blip2 dtype propagation and Qwen3.5 save_pretrained corruption are separate code paths and regressions." + }, + { + "left": "issue:44779", + "right": "issue:45812", + "accept": false, + "reason": "Both are tokenizer regressions, but for different model families and likely different failure modes; too broad to merge." + }, + { + "left": "issue:44683", + "right": "issue:45542", + "accept": false, + "reason": "Compiled flex_attention on torch 2.9 is unrelated to a TensorFlow backend installation error." + }, + { + "left": "issue:43976", + "right": "issue:45405", + "accept": false, + "reason": "Python compatibility issue vs an unmet PEFT version requirement; not the same bug." + }, + { + "left": "issue:38175", + "right": "issue:44617", + "accept": false, + "reason": "Unexpected zero probabilities in SigLIP2 is unrelated to a Sam3Video CUDA OOM report." + }, + { + "left": "issue:44279", + "right": "issue:45405", + "accept": false, + "reason": "Generic dependency issue and a specific minimum-version bump problem are not the same underlying change." + }, + { + "left": "issue:39692", + "right": "issue:43901", + "accept": false, + "reason": "Both touch docs, but one is a SigLIP2 example with model/processor and quantization errors, the other is a text-classification API documentation mismatch." + }, + { + "left": "issue:43299", + "right": "issue:43854", + "accept": false, + "reason": "Qwen3VL loading regression and GLM-4.7-Flash test loading failure are unrelated model-specific issues." + }, + { + "left": "issue:43295", + "right": "issue:44779", + "accept": false, + "reason": "Processor/tokenizer API regression for custom code is unrelated to DeepSeek tokenizer output changes." + }, + { + "left": "issue:43650", + "right": "issue:46132", + "accept": false, + "reason": "Placeholder issue title and an AttentionInterface.register behavior change are clearly unrelated." + }, + { + "left": "issue:44462", + "right": "issue:45081", + "accept": false, + "reason": "AutoTokenizer ignoring tokenizer.json and a Mistral regex patch crash are different tokenizer-loading failures." + }, + { + "left": "issue:44297", + "right": "issue:44462", + "accept": false, + "reason": "Tokenizer save_pretrained metadata mismatch is a different side of the pipeline than AutoTokenizer ignoring tokenizer.json; not the same concrete bug." + }, + { + "left": "issue:45081", + "right": "issue:45907", + "accept": false, + "reason": "Tokenizer regex patching crash vs list_repo_files revision forwarding are unrelated." + }, + { + "left": "issue:43450", + "right": "issue:44279", + "accept": false, + "reason": "Video processor batch shape bug is unrelated to a generic dependency issue." + }, + { + "left": "issue:44016", + "right": "issue:44291", + "accept": false, + "reason": "Notebook syntax error and init_empty_weights argument compatibility are unrelated." + }, + { + "left": "issue:41950", + "right": "issue:43650", + "accept": false, + "reason": "Video-classification processor lookup bug is unrelated to the placeholder data request issue." + }, + { + "left": "issue:43441", + "right": "issue:45834", + "accept": false, + "reason": "Ministral FlashAttention failure and Kosmos2.5 OCR index error are different model-specific runtime bugs." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 34, + "review_unit_count": 49, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-426", + "nodes": [ + "issue:33290", + "issue:41628", + "issue:42175", + "issue:42994", + "issue:43065", + "issue:43262", + "issue:43317", + "issue:43441", + "issue:43450", + "issue:43525", + "issue:43643", + "issue:43650", + "issue:43653", + "issue:43673", + "issue:43723", + "issue:43761", + "issue:43784", + "issue:43824", + "issue:43881", + "issue:43901", + "issue:43976", + "issue:43994", + "issue:44016", + "issue:44162", + "issue:44246", + "issue:44261", + "issue:44263", + "issue:44297", + "issue:44361", + "issue:44466", + "issue:44521", + "issue:44617", + "issue:44623", + "issue:44671", + "issue:44683", + "issue:44749", + "issue:44779", + "issue:44898", + "issue:44987", + "issue:45276", + "issue:45290", + "issue:45397", + "issue:45542", + "issue:45593", + "issue:45600", + "issue:45656", + "issue:45712", + "issue:45823" + ], + "soft_pairs": [ + "issue:45600|issue:45712", + "issue:44466|issue:45276", + "issue:42994|issue:43317", + "issue:33290|issue:45656", + "issue:43650|issue:45290", + "issue:43653|issue:45593", + "issue:44623|issue:44749", + "issue:44623|issue:45397", + "issue:43065|issue:44898", + "issue:44261|issue:44263", + "issue:43441|issue:44683", + "issue:43723|issue:44162", + "issue:44016|issue:45823", + "issue:42175|issue:43976", + "issue:43262|issue:44521", + "issue:44361|issue:45593", + "issue:44671|issue:44779", + "issue:43901|issue:43994", + "issue:43673|issue:43761", + "issue:44987|issue:45542", + "issue:43784|issue:45542", + "issue:43824|issue:43976", + "issue:43525|issue:43881", + "issue:44297|issue:45397", + "issue:43450|issue:44617", + "issue:43525|issue:43643", + "issue:43901|issue:44246", + "issue:41628|issue:43976" + ], + "prepared_review_unit_hash": "0edc0dac4594e6e99dece354fb13aaadeb27303ceb9e430b7f1c5a0ff49af12f", + "component_budget": { + "node_count": 426, + "item_count": 426, + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 28, + "serialized_chars": 35862, + "estimated_input_tokens": 8966, + "estimated_eval_tokens": 18188 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These items are a mixed bag of unrelated Transformers issues across loading, tokenization, processors, models, docs, and build/runtime regressions. None of the soft-edge pairs look like the same concrete underlying bug/change.", + "confidence": 0.97, + "canonical_issue_reason": "issue:44521 is the best representative because it is a substantive user-facing bug with multiple references and broad multimodal/chat-template impact, unlike the more narrow or highly specific reports in the set.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:44521 has the strongest signal for reuse as a canonical issue: clear reproduction, active discussion, and broad downstream relevance.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45600", + "right": "issue:45712", + "accept": false, + "reason": "Both are CI/repo hygiene problems, but one is stale auto_mappings references and the other is leftover dummy classes; different files and failure modes." + }, + { + "left": "issue:44466", + "right": "issue:45276", + "accept": false, + "reason": "Both involve Gemma4, but one is serialization/tied-weights behavior while the other is resize_token_embeddings not updating embeddings." + }, + { + "left": "issue:42994", + "right": "issue:43317", + "accept": false, + "reason": "One is saving quantized models failing; the other is loading dequantized models with device_map/offload. Related area, different bug." + }, + { + "left": "issue:33290", + "right": "issue:45656", + "accept": false, + "reason": "Deepspeed is common, but adafactor OOM and optimizer.step being called twice are distinct optimizer/runtime issues." + }, + { + "left": "issue:43650", + "right": "issue:45290", + "accept": false, + "reason": "Completely different chat-template failures: one is a vague data-related issue, the other is a specific tool-call/content crash." + }, + { + "left": "issue:43653", + "right": "issue:45593", + "accept": false, + "reason": "BigBirdTokenizer special-token registration and D-FINE auxiliary-loss behavior are unrelated." + }, + { + "left": "issue:44623", + "right": "issue:44749", + "accept": false, + "reason": "One is missing files in processor.save_pretrained; the other is a performance regression during data filtering." + }, + { + "left": "issue:44623", + "right": "issue:45397", + "accept": false, + "reason": "Different areas: processor serialization vs Gemma-4 ZeRO-3 from_pretrained loading." + }, + { + "left": "issue:43065", + "right": "issue:44898", + "accept": false, + "reason": "Sam3PixelDecoder dummy conv and Perceiver non-default-resolution failure are separate vision-model bugs." + }, + { + "left": "issue:44261", + "right": "issue:44263", + "accept": false, + "reason": "MLA rms_norm_eps precision issue and GlmMoeDsaIndexer torch.split return values are unrelated." + }, + { + "left": "issue:43441", + "right": "issue:44683", + "accept": false, + "reason": "Both mention attention, but one is a model/backend incompatibility and the other is a torch 2.9 compiled flex_attention failure." + }, + { + "left": "issue:43723", + "right": "issue:44162", + "accept": false, + "reason": "Tokenizer loading and ESM2 model breakage are distinct failures." + }, + { + "left": "issue:44016", + "right": "issue:45823", + "accept": false, + "reason": "Notebook syntax error and Gemma4 device mismatch are unrelated." + }, + { + "left": "issue:42175", + "right": "issue:43976", + "accept": false, + "reason": "One is missing TensorFlow in the torch extra; the other is Python version compatibility for Transformers 5.1.0." + }, + { + "left": "issue:43262", + "right": "issue:44521", + "accept": false, + "reason": "Both involve apply_chat_template, but audio sampling-rate defaulting and multimodal assistant_masks are different code paths and bugs." + }, + { + "left": "issue:44361", + "right": "issue:45593", + "accept": false, + "reason": "MLukeTokenizer task AttributeError and D-FINE auxiliary-loss omission are unrelated." + }, + { + "left": "issue:44671", + "right": "issue:44779", + "accept": false, + "reason": "Both are v5 regressions with incorrect outputs, but they affect different models and tokenizer/model code paths." + }, + { + "left": "issue:43901", + "right": "issue:43994", + "accept": false, + "reason": "Docs mismatch for TextClassificationPipeline and bad SigLIP2 inference are not the same issue." + }, + { + "left": "issue:43673", + "right": "issue:43761", + "accept": false, + "reason": "Generation cache missing in chunked_prefill and CLIPVisionModel hidden_states regression are different model/runtime bugs." + }, + { + "left": "issue:44987", + "right": "issue:45542", + "accept": false, + "reason": "Loading a specific model failing and TensorFlow backend import issues are unrelated." + }, + { + "left": "issue:43784", + "right": "issue:45542", + "accept": false, + "reason": "Sentence-transformers import NameError and undefined tf backend from missing TensorFlow are different import problems." + }, + { + "left": "issue:43824", + "right": "issue:43976", + "accept": false, + "reason": "Missing Qwen2.5-VL export and Python compatibility regression do not share the same root cause." + }, + { + "left": "issue:43525", + "right": "issue:43881", + "accept": false, + "reason": "Llama4Config missing pad_token_id and glm-4v-9b loading failure are separate model-loading issues." + }, + { + "left": "issue:44297", + "right": "issue:45397", + "accept": false, + "reason": "Tokenizer save_pretrained metadata mismatch and Gemma-4 ZeRO-3 loading are unrelated." + }, + { + "left": "issue:43450", + "right": "issue:44617", + "accept": false, + "reason": "Batched video processor output shape and Sam3Video CUDA OOM are different problems." + }, + { + "left": "issue:43525", + "right": "issue:43643", + "accept": false, + "reason": "Both involve config fields, but one is a specific Llama4Config attribute omission and the other is a generic trust_remote_code config hydration issue." + }, + { + "left": "issue:43901", + "right": "issue:44246", + "accept": false, + "reason": "Docs/behavior mismatch in a pipeline and intermittent slow imports are unrelated." + }, + { + "left": "issue:41628", + "right": "issue:43976", + "accept": false, + "reason": "AutoImageProcessor import failure and Python version compatibility regression are 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": 1, + "component_count": 17, + "review_unit_index": 35, + "review_unit_count": 49, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-426", + "nodes": [ + "issue:36246", + "issue:41950", + "issue:42491", + "issue:42757", + "issue:42886", + "issue:43262", + "issue:43299", + "issue:43352", + "issue:43526", + "issue:43531", + "issue:43596", + "issue:43673", + "issue:43901", + "issue:44162", + "issue:44188", + "issue:44220", + "issue:44242", + "issue:44246", + "issue:44295", + "issue:44336", + "issue:44351", + "issue:44373", + "issue:44485", + "issue:44496", + "issue:44561", + "issue:44589", + "issue:44749", + "issue:44779", + "issue:44857", + "issue:44871", + "issue:44877", + "issue:45003", + "issue:45200", + "issue:45250", + "issue:45292", + "issue:45362", + "issue:45375", + "issue:45397", + "issue:45412", + "issue:45561", + "issue:45563", + "issue:45584", + "issue:45593", + "issue:45701", + "issue:45987", + "issue:46082", + "issue:46123", + "issue:46132" + ], + "soft_pairs": [ + "issue:44589|issue:45362", + "issue:44188|issue:44336", + "issue:45561|issue:45563", + "issue:42491|issue:43299", + "issue:44336|issue:44373", + "issue:44242|issue:45593", + "issue:44295|issue:45397", + "issue:44246|issue:44496", + "issue:41950|issue:44220", + "issue:43531|issue:45362", + "issue:42886|issue:46132", + "issue:43526|issue:45003", + "issue:36246|issue:42757", + "issue:36246|issue:44351", + "issue:44336|issue:44485", + "issue:45003|issue:45561", + "issue:44246|issue:44871", + "issue:43352|issue:45250", + "issue:43673|issue:44561", + "issue:36246|issue:43596", + "issue:44242|issue:44857", + "issue:44779|issue:45292", + "issue:44295|issue:44749", + "issue:44162|issue:45200", + "issue:43901|issue:45003", + "issue:45412|issue:46123", + "issue:44496|issue:45003", + "issue:44877|issue:45375", + "issue:45292|issue:45584", + "issue:45987|issue:46082", + "issue:43262|issue:44246", + "issue:45292|issue:45701" + ], + "prepared_review_unit_hash": "5b883c8ce00e74b0a36228c950cf9e8b0de0b36a0c46744bfffa258b55ef8857", + "component_budget": { + "node_count": 426, + "item_count": 426, + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 32, + "serialized_chars": 37098, + "estimated_input_tokens": 9275, + "estimated_eval_tokens": 18806 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster looks like a noisy semantic grouping of mostly unrelated issues. A few are thematically similar (import errors, Qwen/MoE regressions, config strictness), but none of the soft pairs look like the same concrete bug/change.", + "confidence": 0.82, + "canonical_issue_reason": "issue:44246 is the broadest and most discussion-heavy report in the set, so it is the least-bad anchor for a noisy cluster.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:44246 is the most generic/high-activity issue here, but the cluster still does not form a true duplicate family.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44589", + "right": "issue:45362", + "accept": false, + "reason": "Different failures: a missing storage import vs a Qwen3.5 crash in chat generation; same broad ML stack, not the same bug." + }, + { + "left": "issue:44188", + "right": "issue:44336", + "accept": false, + "reason": "Unrelated code paths: torch.compile attention-kernel divergence vs ANSI codes in loading_report." + }, + { + "left": "issue:45561", + "right": "issue:45563", + "accept": false, + "reason": "One is a pytest-xdist file race, the other a stale generation warning; no shared underlying defect." + }, + { + "left": "issue:42491", + "right": "issue:43299", + "accept": false, + "reason": "Both involve Qwen/MoE version regressions, but one is LoRA compatibility and the other is model loading; not the same concrete issue." + }, + { + "left": "issue:44336", + "right": "issue:44373", + "accept": false, + "reason": "Docs/formatting bug vs a position_ids docstring issue; unrelated." + }, + { + "left": "issue:44242", + "right": "issue:45593", + "accept": false, + "reason": "Both mention missing losses, but they are different models and different gating logic; not mergeable as one PR." + }, + { + "left": "issue:44295", + "right": "issue:45397", + "accept": false, + "reason": "A buffer/position_ids error and a ZeRO3 Gemma-4 load failure are different bugs." + }, + { + "left": "issue:44246", + "right": "issue:44496", + "accept": false, + "reason": "Import slowdown vs unrecognized model/config loading failure; no shared code path." + }, + { + "left": "issue:41950", + "right": "issue:44220", + "accept": false, + "reason": "Video pipeline processor selection and audio fbank extraction are different subsystems." + }, + { + "left": "issue:43531", + "right": "issue:45362", + "accept": false, + "reason": "Both are Qwen-related, but one is a sliding_window bug and the other is a chat crash; not the same defect." + }, + { + "left": "issue:42886", + "right": "issue:46132", + "accept": false, + "reason": "Offline cache loading and AttentionInterface.register behavior are unrelated." + }, + { + "left": "issue:43526", + "right": "issue:45003", + "accept": false, + "reason": "Beit label reduction and unsafe sys.modules access are separate implementation bugs." + }, + { + "left": "issue:36246", + "right": "issue:42757", + "accept": false, + "reason": "Both are import errors, but for different packages and missing symbols." + }, + { + "left": "issue:36246", + "right": "issue:44351", + "accept": false, + "reason": "Different missing exports: Qwen2_5_VLImageProcessor vs HybridCache." + }, + { + "left": "issue:44336", + "right": "issue:44485", + "accept": false, + "reason": "ANSI-code output formatting and GLM-5 RoPE implementation are unrelated." + }, + { + "left": "issue:45003", + "right": "issue:45561", + "accept": false, + "reason": "sys.modules access in modeling_utils vs a captured_info.txt race in tests are not the same bug." + }, + { + "left": "issue:44242", + "right": "issue:44871", + "accept": false, + "reason": "Load-balancing-loss handling and Gemma eos_token_id config mismatch are unrelated." + }, + { + "left": "issue:43352", + "right": "issue:45250", + "accept": false, + "reason": "A model-specific Flash Attention support gap is not the same as the generic Flash Attention topic." + }, + { + "left": "issue:43673", + "right": "issue:44561", + "accept": false, + "reason": "Generation cache regression during chunked prefill vs removal of is_torch_fx_available affecting trust_remote_code are different changes." + }, + { + "left": "issue:36246", + "right": "issue:43596", + "accept": false, + "reason": "ImportError for a missing processor symbol vs a DeepSpeed/Bert IndexError are unrelated." + }, + { + "left": "issue:44242", + "right": "issue:44857", + "accept": false, + "reason": "Both are training-loss failures, but on different models and different failure mechanisms." + }, + { + "left": "issue:44779", + "right": "issue:45292", + "accept": false, + "reason": "Tokenizer regression vs resize_token_embeddings not updating output embeddings are distinct bugs." + }, + { + "left": "issue:44295", + "right": "issue:44749", + "accept": false, + "reason": "A position_ids buffer error and a version-related performance slowdown are not the same issue." + }, + { + "left": "issue:44162", + "right": "issue:45200", + "accept": false, + "reason": "ESM2 breakage and Gemma-4 mm_token_type_ids defaults are separate model-specific problems." + }, + { + "left": "issue:43901", + "right": "issue:45003", + "accept": false, + "reason": "Documentation mismatch for return_all_scores does not match the sys.modules access bug." + }, + { + "left": "issue:45412", + "right": "issue:46123", + "accept": false, + "reason": "Memory leak on delete/GC and a final-partial-batch pipeline bug are unrelated." + }, + { + "left": "issue:44496", + "right": "issue:45003", + "accept": false, + "reason": "Model config recognition failure and unsafe sys.modules access are different code paths." + }, + { + "left": "issue:44877", + "right": "issue:45375", + "accept": false, + "reason": "Both touch strict config handling, but one rejects granite_speech loading while the other silently drops a missing field in a different config class; not the same concrete bug." + }, + { + "left": "issue:45292", + "right": "issue:45584", + "accept": false, + "reason": "Embedding resize behavior and Whisper empty-transcription failure are unrelated." + }, + { + "left": "issue:45987", + "right": "issue:46082", + "accept": false, + "reason": "StaticCache return type contract and LlamaConfig head_dim validation are different areas." + }, + { + "left": "issue:43262", + "right": "issue:44246", + "accept": false, + "reason": "Audio chat-template sampling-rate defaults and import-time slowness are unrelated." + }, + { + "left": "issue:45292", + "right": "issue:45701", + "accept": false, + "reason": "Embedding resizing behavior and version-dependent tokenization changes 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": 1, + "component_count": 17, + "review_unit_index": 36, + "review_unit_count": 49, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-426", + "nodes": [ + "issue:42886", + "issue:43065", + "issue:43232", + "issue:43257", + "issue:43408", + "issue:43475", + "issue:43540", + "issue:43646", + "issue:43653", + "issue:43673", + "issue:43931", + "issue:44060", + "issue:44077", + "issue:44162", + "issue:44297", + "issue:44336", + "issue:44361", + "issue:44368", + "issue:44466", + "issue:44479", + "issue:44530", + "issue:44623", + "issue:44734", + "issue:44749", + "issue:44869", + "issue:44871", + "issue:44877", + "issue:44933", + "issue:44977", + "issue:44987", + "issue:44995", + "issue:45137", + "issue:45200", + "issue:45229", + "issue:45278", + "issue:45335", + "issue:45356", + "issue:45375", + "issue:45397", + "issue:45459", + "issue:45563", + "issue:45588", + "issue:45600", + "issue:45656", + "issue:45759", + "issue:46123", + "issue:46132", + "issue:46170" + ], + "soft_pairs": [ + "issue:45375|issue:45759", + "issue:43257|issue:45137", + "issue:44987|issue:45459", + "issue:43232|issue:44336", + "issue:44995|issue:45200", + "issue:43475|issue:43540", + "issue:44479|issue:46170", + "issue:43065|issue:43408", + "issue:42886|issue:45229", + "issue:44466|issue:45335", + "issue:44060|issue:44077", + "issue:43653|issue:44162", + "issue:45397|issue:45656", + "issue:44995|issue:45588", + "issue:44368|issue:44977", + "issue:45137|issue:45600", + "issue:44368|issue:45563", + "issue:44877|issue:46170", + "issue:43673|issue:44734", + "issue:43931|issue:44368", + "issue:44530|issue:46170", + "issue:43065|issue:46123", + "issue:44933|issue:45459", + "issue:44869|issue:45356", + "issue:44297|issue:44623", + "issue:43232|issue:43673", + "issue:44361|issue:44749", + "issue:43646|issue:44162", + "issue:44871|issue:46132", + "issue:43673|issue:45278", + "issue:44623|issue:45656", + "issue:44749|issue:45656" + ], + "prepared_review_unit_hash": "9100cbc4a71d1794d0322bedf61bcae33d0f97b0bdfc41d2bbf13d49ac43aa3a", + "component_budget": { + "node_count": 426, + "item_count": 426, + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 32, + "serialized_chars": 37429, + "estimated_input_tokens": 9358, + "estimated_eval_tokens": 18972 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This set is mostly a collection of unrelated closed issues across tokenizer, generation, multimodal, DeepSpeed, and config-loading bugs. I do not see any soft pair that looks like the same concrete bug/change, so I would not merge any of them as duplicates.", + "confidence": 0.9, + "canonical_issue_reason": "No single issue is a true canonical duplicate anchor here; the set is heterogeneous. If forced to pick a representative, issue 44479 is the broadest single regression report, but it still does not subsume the rest.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 44479 is the most representative of the larger recurring regression theme in this set (multi-model Qwen/VL video-input breakage), but it is still only a representative issue, not a duplicate hub.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45375", + "right": "issue:45759", + "accept": false, + "reason": "Both involve Qwen 3.5/3.6 composite configs, but one is strict-config field dropping and the other is from_config text_config unwrapping; different bugs." + }, + { + "left": "issue:43257", + "right": "issue:45137", + "accept": false, + "reason": "Both mention DeepSpeed, but one is MoE weight conversion on load and the other is a ZeRO3 deque/index error; not the same code path." + }, + { + "left": "issue:44987", + "right": "issue:45459", + "accept": false, + "reason": "One is a model-loading regression for a specific repo/version, the other is tokenizer error masking when protobuf is missing." + }, + { + "left": "issue:43232", + "right": "issue:44336", + "accept": false, + "reason": "Generation kwargs after sync_gpus is unrelated to ANSI output in loading_report." + }, + { + "left": "issue:44995", + "right": "issue:45200", + "accept": false, + "reason": "Different models and failures: stale indexer cache on second forward vs missing mm_token_type_ids defaulting." + }, + { + "left": "issue:43475", + "right": "issue:43540", + "accept": false, + "reason": "Both are video/model issues, but one is a missing output attribute in SAM 3 and the other is a Qwen3OmniMoe video-processing ValueError." + }, + { + "left": "issue:44479", + "right": "issue:46170", + "accept": false, + "reason": "Video-input regression in Qwen models is unrelated to Windows/ROCm import and torch.distributed dependency handling." + }, + { + "left": "issue:43065", + "right": "issue:43408", + "accept": false, + "reason": "Both are SAM 3 related, but one is a dummy Conv2d implementation detail and the other is a model-type warning during instantiation." + }, + { + "left": "issue:42886", + "right": "issue:45229", + "accept": false, + "reason": "Tokenizer offline-cache loading is unrelated to Gemma4 multi-GPU OOM." + }, + { + "left": "issue:44466", + "right": "issue:45335", + "accept": false, + "reason": "Both touch tied embeddings, but one is serialization inconsistency and the other is resize_token_embeddings not propagating to decoder.embed_tokens." + }, + { + "left": "issue:44060", + "right": "issue:44077", + "accept": false, + "reason": "Different subsystems: Qwen3-Next tied-weights warning vs patchtsmixer post_init policy." + }, + { + "left": "issue:43653", + "right": "issue:44162", + "accept": false, + "reason": "BigBirdTokenizer special-token registration and ESM2 breakage are unrelated." + }, + { + "left": "issue:45397", + "right": "issue:45656", + "accept": false, + "reason": "Gemma-4 ZeRO3 loading and duplicate optimizer.step behavior are different DeepSpeed problems." + }, + { + "left": "issue:44995", + "right": "issue:45588", + "accept": false, + "reason": "Stale cache on second forward is unrelated to flash_attention s_aux=None handling." + }, + { + "left": "issue:44368", + "right": "issue:44977", + "accept": false, + "reason": "One is a tied-embeddings warning during fine-tuning, the other is a flash-attention generation regression." + }, + { + "left": "issue:45137", + "right": "issue:45600", + "accept": false, + "reason": "ZeRO3 deque error and auto_mappings broken CI are unrelated." + }, + { + "left": "issue:44368", + "right": "issue:45563", + "accept": false, + "reason": "Tied-embeddings warning is unrelated to the stale num_return_sequences warning." + }, + { + "left": "issue:44877", + "right": "issue:46170", + "accept": false, + "reason": "Strict config loading for granite_speech is unrelated to the Windows ROCm import error." + }, + { + "left": "issue:43673", + "right": "issue:44734", + "accept": false, + "reason": "Both concern generation/cache, but one is missing cache in chunked_prefill and the other is wrong tensor indexing in the serve API." + }, + { + "left": "issue:43931", + "right": "issue:44368", + "accept": false, + "reason": "Weight-shape mismatch loading Qwen3-VL is unrelated to a tied-embeddings warning." + }, + { + "left": "issue:44530", + "right": "issue:46170", + "accept": false, + "reason": "PagedAttentionCache invalid group type on Qwen3.5 is unrelated to the ROCm import/dependency issue." + }, + { + "left": "issue:43065", + "right": "issue:46123", + "accept": false, + "reason": "SAM 3 pixel decoder dummy layer and MaskGenerationPipeline partial-batch dropping are different bugs." + }, + { + "left": "issue:44933", + "right": "issue:45459", + "accept": false, + "reason": "Missing import from image_utils is unrelated to protobuf error masking in tokenizer decode." + }, + { + "left": "issue:44869", + "right": "issue:45356", + "accept": false, + "reason": "Both are tokenizer-adjacent, but they affect different models and code paths (Whisper timestamp decode vs Kimi-K2.5 codec handling)." + }, + { + "left": "issue:44297", + "right": "issue:44623", + "accept": false, + "reason": "tokenizer.save_pretrained mismatch and processor.save_pretrained missing files are separate serialization bugs." + }, + { + "left": "issue:43232", + "right": "issue:43673", + "accept": false, + "reason": "Different generation-cache failures: sync_gpus kwargs update vs chunked_prefill cache availability." + }, + { + "left": "issue:44361", + "right": "issue:44749", + "accept": false, + "reason": "MLukeTokenizer task AttributeError is unrelated to the Chinese report about performance slowing after upgrade." + }, + { + "left": "issue:43646", + "right": "issue:44162", + "accept": false, + "reason": "Custom model initialization breakage in Transformers 5.0.0 and ESM2 breakage are not the same issue." + }, + { + "left": "issue:44871", + "right": "issue:46132", + "accept": false, + "reason": "Gemma eos_token_id mismatch is unrelated to AttentionInterface.register behavior changes." + }, + { + "left": "issue:43673", + "right": "issue:45278", + "accept": false, + "reason": "Missing generation cache during chunked_prefill is not the same as general import errors after upgrade." + }, + { + "left": "issue:44623", + "right": "issue:45656", + "accept": false, + "reason": "Processor save_pretrained file omissions and duplicate optimizer.step calls are unrelated." + }, + { + "left": "issue:44749", + "right": "issue:45656", + "accept": false, + "reason": "Performance regression after upgrade and duplicate optimizer.step are different symptoms and code paths." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 37, + "review_unit_count": 49, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-426", + "nodes": [ + "issue:28282", + "issue:30990", + "issue:36010", + "issue:41628", + "issue:42175", + "issue:42371", + "issue:42673", + "issue:42831", + "issue:42994", + "issue:43232", + "issue:43475", + "issue:43482", + "issue:43504", + "issue:43525", + "issue:43526", + "issue:43643", + "issue:43673", + "issue:43704", + "issue:43824", + "issue:43873", + "issue:44038", + "issue:44188", + "issue:44222", + "issue:44351", + "issue:44496", + "issue:44623", + "issue:44661", + "issue:44749", + "issue:44792", + "issue:44933", + "issue:44945", + "issue:44964", + "issue:45081", + "issue:45092", + "issue:45103", + "issue:45137", + "issue:45230", + "issue:45310", + "issue:45375", + "issue:45381", + "issue:45397", + "issue:45478", + "issue:45561", + "issue:45712", + "issue:45753", + "issue:46077", + "issue:46123" + ], + "soft_pairs": [ + "issue:44749|issue:45310", + "issue:42994|issue:43482", + "issue:43475|issue:45381", + "issue:28282|issue:30990", + "issue:45230|issue:45397", + "issue:43475|issue:43643", + "issue:44038|issue:44623", + "issue:46077|issue:46123", + "issue:41628|issue:44933", + "issue:42371|issue:45092", + "issue:43475|issue:44661", + "issue:42994|issue:43525", + "issue:45375|issue:45478", + "issue:45081|issue:45103", + "issue:45310|issue:45375", + "issue:43873|issue:44945", + "issue:41628|issue:44351", + "issue:43232|issue:44188", + "issue:42673|issue:43704", + "issue:42831|issue:44222", + "issue:44945|issue:45753", + "issue:36010|issue:41628", + "issue:44222|issue:45310", + "issue:42994|issue:43824", + "issue:44964|issue:45137", + "issue:42175|issue:45712", + "issue:43232|issue:44792", + "issue:44038|issue:45310", + "issue:43673|issue:44188", + "issue:44496|issue:45561", + "issue:43526|issue:45561", + "issue:43504|issue:45561" + ], + "prepared_review_unit_hash": "0dacca25acb9973197aaba1b0cbaff8aa09705d29a795d57c76de68c085f3e30", + "component_budget": { + "node_count": 426, + "item_count": 426, + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 32, + "serialized_chars": 36486, + "estimated_input_tokens": 9122, + "estimated_eval_tokens": 18500 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is mostly heterogeneous, but it does contain two clear duplicate pockets: the Qwen3ForCausalLM VRAM-leak pair (42673/43704) and an exact-title duplicate pair for '[BUG] transformers>=5.4.0, Qwen3.5 Moe from_pretrained error' (45310/45478). The remaining soft pairs are largely false positives across broadly related subsystems, not the same concrete bug.", + "confidence": 0.84, + "canonical_issue_reason": "issue:42673 is the strongest canonical issue for the cluster because it anchors the clearest confirmed duplicate pair (exact same Qwen3ForCausalLM VRAM-leak report as issue:43704) and has the richer discussion thread; the packet also contains a separate exact-title duplicate pair, issue:45310 and issue:45478.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:42673 is the best single representative issue in this packet because it is part of the clearest duplicate pair and has the strongest supporting discussion, while the rest of the set is mixed and the other duplicate pocket is separate.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44749", + "right": "issue:45310", + "accept": false, + "reason": "Different failures: data filtering slowdown vs a Qwen3.5-Moe from_pretrained error." + }, + { + "left": "issue:42994", + "right": "issue:43482", + "accept": false, + "reason": "Different code paths: quantized model saving vs GGUF loading on a different model/version issue." + }, + { + "left": "issue:43475", + "right": "issue:45381", + "accept": false, + "reason": "Both are multimodal/video-related, but they describe different models and distinct bugs." + }, + { + "left": "issue:28282", + "right": "issue:30990", + "accept": false, + "reason": "Missing-PyTorch import error is unrelated to a Sentence Transformers loading hang." + }, + { + "left": "issue:45230", + "right": "issue:45397", + "accept": false, + "reason": "Generic bug report vs a specific ZeRO3 from_pretrained issue; no evidence they are the same bug." + }, + { + "left": "issue:43475", + "right": "issue:43643", + "accept": false, + "reason": "Missing model attribute in SAM 3 Video is different from trust_remote_code missing fields in AutoConfig." + }, + { + "left": "issue:44038", + "right": "issue:44623", + "accept": false, + "reason": "Qwen3-VL-Moe compatibility and processor.save_pretrained missing files are different failures." + }, + { + "left": "issue:46077", + "right": "issue:46123", + "accept": false, + "reason": "Attention-mask construction bug is unrelated to final-batch result dropping in the pipeline." + }, + { + "left": "issue:41628", + "right": "issue:44933", + "accept": false, + "reason": "Both are import-related, but they concern different missing symbols and likely different regressions." + }, + { + "left": "issue:42371", + "right": "issue:45092", + "accept": false, + "reason": "TF32 API guidance is unrelated to remote-code/meta-init incompatibility for InternVL2." + }, + { + "left": "issue:43475", + "right": "issue:44661", + "accept": false, + "reason": "Different problems: missing encoder output field vs tokenizer-mapping failure in add-new-model-like." + }, + { + "left": "issue:42994", + "right": "issue:43525", + "accept": false, + "reason": "Quantized save failure is unrelated to Llama4Config missing pad_token_id." + }, + { + "left": "issue:45375", + "right": "issue:45478", + "accept": false, + "reason": "Related Qwen family, but one is a missing config field and the other is a generic from_pretrained error." + }, + { + "left": "issue:45081", + "right": "issue:45103", + "accept": false, + "reason": "Tokenizer regex crash and auto_docstring annotation crash are unrelated." + }, + { + "left": "issue:45310", + "right": "issue:45375", + "accept": false, + "reason": "Same model family, but one is a from_pretrained failure and the other is a missing vision config field." + }, + { + "left": "issue:43873", + "right": "issue:44945", + "accept": false, + "reason": "Offloading/quantization behavior is unrelated to incorrect output under pipeline parallelism." + }, + { + "left": "issue:41628", + "right": "issue:44351", + "accept": false, + "reason": "Both are import errors, but they concern different missing names and likely different export regressions." + }, + { + "left": "issue:43232", + "right": "issue:44188", + "accept": false, + "reason": "Generation kwargs update after sync_gpus is unrelated to attention-kernel divergence under torch.compile." + }, + { + "left": "issue:42673", + "right": "issue:43704", + "accept": true, + "reason": "Exact same Qwen3ForCausalLM VRAM-leak bug in multi-threaded dataloader use; clear duplicate." + }, + { + "left": "issue:42831", + "right": "issue:44222", + "accept": false, + "reason": "Both involve FP8, but one is an accuracy issue and the other is FP8 save_pretrained for MoE." + }, + { + "left": "issue:44945", + "right": "issue:45753", + "accept": false, + "reason": "Incorrect pipeline-parallel output is unrelated to an image-triggered infinite loop in Qwen3_5." + }, + { + "left": "issue:36010", + "right": "issue:41628", + "accept": false, + "reason": "Different missing imports: GenerationMixin vs AutoImageProcessor." + }, + { + "left": "issue:44222", + "right": "issue:45310", + "accept": false, + "reason": "FP8 save_pretrained MoE is unrelated to Qwen3.5-Moe from_pretrained errors." + }, + { + "left": "issue:42994", + "right": "issue:43824", + "accept": false, + "reason": "Saving a quantized model is unrelated to a missing Qwen2.5-VL class import." + }, + { + "left": "issue:44964", + "right": "issue:45137", + "accept": false, + "reason": "Model loading incompatibility is unrelated to a DeepSpeed ZeRO3 deque error." + }, + { + "left": "issue:42175", + "right": "issue:45712", + "accept": false, + "reason": "Backend packaging/install issue is different from dummy_pt_objects leaking into dir(transformers)." + }, + { + "left": "issue:43232", + "right": "issue:44792", + "accept": false, + "reason": "Generation kwargs/sync_gpus bug is unrelated to a janus image-generation test failure." + }, + { + "left": "issue:44038", + "right": "issue:45310", + "accept": false, + "reason": "Different Qwen families and different failures: VL-Moe compatibility vs Qwen3.5-Moe from_pretrained." + }, + { + "left": "issue:43673", + "right": "issue:44188", + "accept": false, + "reason": "Chunked_prefill cache regression is unrelated to bidirectional-skip branching in attention kernels." + }, + { + "left": "issue:44496", + "right": "issue:45561", + "accept": false, + "reason": "Unrecognized model/config issue is unrelated to a pytest-xdist race on captured_info.txt." + }, + { + "left": "issue:43526", + "right": "issue:45561", + "accept": false, + "reason": "BeitImageProcessorFast reduce_labels bug is unrelated to the xdist file race." + }, + { + "left": "issue:43504", + "right": "issue:45561", + "accept": false, + "reason": "Legacy-field model loading issue is unrelated to the captured_info.txt race." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "The summary is not fully grounded in the packet. In addition to the Qwen3ForCausalLM VRAM-leak duplicate pair (42673/43704), the packet also contains an exact-title match between issue:45310 and issue:45478 ('[BUG] transformers>=5.4.0, Qwen3.5 Moe from_pretrained error'), which the analyst omitted. Because the duplicate landscape is understated, the overall cluster summary is too strong." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 38, + "review_unit_count": 49, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-426", + "nodes": [ + "issue:30333", + "issue:30990", + "issue:33290", + "issue:35141", + "issue:36331", + "issue:42175", + "issue:42222", + "issue:42831", + "issue:42898", + "issue:43278", + "issue:43377", + "issue:43479", + "issue:43644", + "issue:43723", + "issue:43784", + "issue:43824", + "issue:43874", + "issue:43976", + "issue:44060", + "issue:44079", + "issue:44188", + "issue:44190", + "issue:44222", + "issue:44242", + "issue:44295", + "issue:44297", + "issue:44373", + "issue:44448", + "issue:44568", + "issue:44623", + "issue:44734", + "issue:44779", + "issue:44869", + "issue:44933", + "issue:45092", + "issue:45103", + "issue:45125", + "issue:45137", + "issue:45200", + "issue:45245", + "issue:45310", + "issue:45356", + "issue:45405", + "issue:45656", + "issue:45712", + "issue:45812", + "issue:46082", + "issue:46132" + ], + "soft_pairs": [ + "issue:43377|issue:44242", + "issue:42222|issue:45200", + "issue:42175|issue:44933", + "issue:42898|issue:44568", + "issue:43976|issue:44933", + "issue:30990|issue:35141", + "issue:43479|issue:44869", + "issue:43479|issue:44079", + "issue:43278|issue:45137", + "issue:44623|issue:45310", + "issue:42898|issue:45812", + "issue:44188|issue:44373", + "issue:30333|issue:33290", + "issue:43784|issue:43824", + "issue:44734|issue:45103", + "issue:43723|issue:44568", + "issue:42898|issue:45356", + "issue:42898|issue:44779", + "issue:43874|issue:45092", + "issue:44060|issue:45125", + "issue:45405|issue:45712", + "issue:42898|issue:44448", + "issue:44295|issue:45656", + "issue:44448|issue:46132", + "issue:44933|issue:45712", + "issue:44222|issue:44297", + "issue:44297|issue:45656", + "issue:43479|issue:43644", + "issue:44190|issue:46082", + "issue:42831|issue:44623", + "issue:44222|issue:44623", + "issue:30333|issue:36331", + "issue:45245|issue:46132" + ], + "prepared_review_unit_hash": "922cf2ceb35bc91764dd684823bc24df138da53a6137f4bbe5dcb1b204568890", + "component_budget": { + "node_count": 426, + "item_count": 426, + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 33, + "serialized_chars": 37690, + "estimated_input_tokens": 9423, + "estimated_eval_tokens": 19102 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is mostly a mix of unrelated Transformers v5-era bug reports. The soft matches are topical similarities, but none look like true duplicates of the same underlying defect.", + "confidence": 0.62, + "canonical_issue_reason": "issue:42898 is the broadest tokenizer-regression report here and sits near several tokenizer-related soft matches, so it is the best cluster anchor.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45812 is the clearest high-impact standalone bug report: explicit v4\u2192v5 regression, broad model impact, and a concrete symptom.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43377", + "right": "issue:44242", + "accept": false, + "reason": "Different failure modes: MIMI padding-mask parity vs MoE load-balancing loss gating." + }, + { + "left": "issue:42222", + "right": "issue:45200", + "accept": false, + "reason": "Different bugs: vitpose model import breakage vs Gemma 4 mm_token_type_ids defaulting." + }, + { + "left": "issue:42175", + "right": "issue:44933", + "accept": false, + "reason": "Packaging/backend extra issue vs nonexistent import from image_utils." + }, + { + "left": "issue:42898", + "right": "issue:44568", + "accept": false, + "reason": "Both are tokenizer regressions in v5, but one is clean_up_tokenization_spaces behavior and the other is add_special_tokens/BOS-EOS handling." + }, + { + "left": "issue:43976", + "right": "issue:44933", + "accept": false, + "reason": "Python support/version mismatch vs image_utils import error." + }, + { + "left": "issue:30990", + "right": "issue:35141", + "accept": false, + "reason": "Sentence Transformers loading hang is unrelated to embedding reinitialization after resize." + }, + { + "left": "issue:43479", + "right": "issue:44869", + "accept": false, + "reason": "Config None-default initialization bug vs Whisper word-timestamp decode crash." + }, + { + "left": "issue:43479", + "right": "issue:44079", + "accept": false, + "reason": "Config initialization bug vs ModelOutput key assignment bug; different code paths." + }, + { + "left": "issue:43278", + "right": "issue:45137", + "accept": false, + "reason": "Embedding dtype drift in eval is unrelated to DeepSpeed ZeRO3 deque underflow." + }, + { + "left": "issue:44623", + "right": "issue:45310", + "accept": false, + "reason": "Processor save_pretrained missing files vs Qwen3.5 MoE from_pretrained failure." + }, + { + "left": "issue:42898", + "right": "issue:45812", + "accept": false, + "reason": "Related tokenizer regressions, but distinct defects: cleanup-space semantics vs Granite token IDs." + }, + { + "left": "issue:44188", + "right": "issue:44373", + "accept": false, + "reason": "torch.compile attention branching bug vs docstring wording issue." + }, + { + "left": "issue:30333", + "right": "issue:33290", + "accept": false, + "reason": "MLflow reporting on failed jobs is unrelated to Adafactor OOM under DeepSpeed." + }, + { + "left": "issue:43784", + "right": "issue:43824", + "accept": false, + "reason": "Sentence-transformers import NameError vs missing Qwen2.5-VL class export." + }, + { + "left": "issue:44734", + "right": "issue:45103", + "accept": false, + "reason": "KV-cache continuation indexing crash vs auto-docstring crash with future annotations." + }, + { + "left": "issue:43723", + "right": "issue:44568", + "accept": false, + "reason": "Tokenizer loading issue vs special-token insertion regression; not the same bug." + }, + { + "left": "issue:42898", + "right": "issue:45356", + "accept": false, + "reason": "Different tokenizer regressions: cleanup-space behavior vs Kimi codec handling." + }, + { + "left": "issue:42898", + "right": "issue:44779", + "accept": false, + "reason": "Different tokenizer bugs: cleanup-space behavior vs Deepseek incorrect results." + }, + { + "left": "issue:43874", + "right": "issue:45092", + "accept": false, + "reason": "Missing image patch-count API vs remote-code checkpoint/meta-init incompatibility." + }, + { + "left": "issue:44060", + "right": "issue:45125", + "accept": false, + "reason": "Incorrect tied-weights warning vs missing tensor-parallel plan." + }, + { + "left": "issue:45405", + "right": "issue:45712", + "accept": false, + "reason": "PEFT version pinning issue vs dummy-class cleanup/leakage." + }, + { + "left": "issue:42898", + "right": "issue:44448", + "accept": false, + "reason": "Tokenizer cleanup-space regression vs Pegasus output mismatch; different symptoms and likely causes." + }, + { + "left": "issue:44295", + "right": "issue:45656", + "accept": false, + "reason": "position_ids buffer access error vs optimizer step being called twice under DeepSpeed." + }, + { + "left": "issue:44297", + "right": "issue:45656", + "accept": false, + "reason": "tokenizer.save_pretrained metadata mismatch vs duplicated optimizer stepping." + }, + { + "left": "issue:43479", + "right": "issue:43644", + "accept": false, + "reason": "Config default initialization bug vs non-persistent buffer junk-fill bug." + }, + { + "left": "issue:44190", + "right": "issue:46082", + "accept": false, + "reason": "Local dataset script loading issue vs LlamaConfig head_dim validation." + }, + { + "left": "issue:42831", + "right": "issue:44623", + "accept": false, + "reason": "FineGrainedFP8 accuracy issue vs processor.save_pretrained missing files." + }, + { + "left": "issue:44222", + "right": "issue:44623", + "accept": false, + "reason": "FP8 MoE save_pretrained bug vs processor file persistence bug." + }, + { + "left": "issue:30333", + "right": "issue:36331", + "accept": false, + "reason": "Failed-job MLflow reporting vs CustomTrainer.compute_loss kwarg incompatibility." + }, + { + "left": "issue:45245", + "right": "issue:46132", + "accept": false, + "reason": "Category overflow in embeddings vs AttentionInterface.register behavior change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 39, + "review_unit_count": 49, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-426", + "nodes": [ + "issue:22355", + "issue:28282", + "issue:30333", + "issue:30990", + "issue:34689", + "issue:35141", + "issue:38175", + "issue:39401", + "issue:41628", + "issue:42757", + "issue:42831", + "issue:42907", + "issue:43232", + "issue:43479", + "issue:43723", + "issue:43761", + "issue:43824", + "issue:44038", + "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:44568", + "issue:44623", + "issue:44829", + "issue:44871", + "issue:44913", + "issue:44977", + "issue:44991", + "issue:45071", + "issue:45092", + "issue:45103", + "issue:45310", + "issue:45357", + "issue:45406", + "issue:45834", + "issue:46002", + "issue:46050", + "issue:46082" + ], + "soft_pairs": [ + "issue:30333|issue:34689", + "issue:22355|issue:28282", + "issue:41628|issue:42757", + "issue:45071|issue:46082", + "issue:44206|issue:45103", + "issue:28282|issue:30333", + "issue:30333|issue:35141", + "issue:44297|issue:45310", + "issue:43723|issue:44977", + "issue:44222|issue:44484", + "issue:43232|issue:45834", + "issue:44263|issue:45092", + "issue:43761|issue:46082", + "issue:44483|issue:45406", + "issue:44295|issue:44623", + "issue:44373|issue:44485", + "issue:44514|issue:45103", + "issue:43723|issue:44871", + "issue:44263|issue:44360", + "issue:45092|issue:45357", + "issue:30990|issue:38175", + "issue:42907|issue:46050", + "issue:43723|issue:44991", + "issue:43723|issue:44442", + "issue:44261|issue:45092", + "issue:44829|issue:46002", + "issue:41628|issue:43824", + "issue:42831|issue:45310", + "issue:43479|issue:44913", + "issue:42831|issue:44038", + "issue:44263|issue:44373", + "issue:39401|issue:44568" + ], + "prepared_review_unit_hash": "a4c5eb0b6a9270c2a115f25123527a7d08d11663a41293431ac04670770e0e70", + "component_budget": { + "node_count": 426, + "item_count": 426, + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 32, + "serialized_chars": 37080, + "estimated_input_tokens": 9270, + "estimated_eval_tokens": 18796 + }, + "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 bugs; I don\u2019t see any soft pair that looks like the same underlying issue. The best representative issue is the broad v5 tokenizer-loading regression.", + "confidence": 0.84, + "canonical_issue_reason": "issue:43723 is the broadest, most central report in the set: a version-scoped tokenizer loading failure in v5, which is the closest thing to a unifying theme here.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43723 is the clearest umbrella issue: it is specific enough to be actionable, but general enough to represent the recurring tokenizer-import/load regressions better than the model-specific reports.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:30333", + "right": "issue:34689", + "accept": false, + "reason": "Different failures: MLflow reporting on training job failure vs model loading regression for Llama 3.2 Vision." + }, + { + "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:41628", + "right": "issue:42757", + "accept": false, + "reason": "One is a missing `AutoImageProcessor` import in transformers, the other is a dependency API mismatch in `huggingface_hub`." + }, + { + "left": "issue:45071", + "right": "issue:46082", + "accept": false, + "reason": "`PretrainedConfig` type-checking regression is unrelated to LlamaConfig rejecting an explicit `head_dim` value." + }, + { + "left": "issue:44206", + "right": "issue:45103", + "accept": false, + "reason": "Different code paths: a feature extractor crashing on an unsupported arg vs an auto-docstring annotation introspection crash." + }, + { + "left": "issue:28282", + "right": "issue:30333", + "accept": false, + "reason": "PyTorch import failure and MLflow job-reporting failure are unrelated." + }, + { + "left": "issue:30333", + "right": "issue:35141", + "accept": false, + "reason": "Training-job reporting bug vs token-embedding resize/post-init reinitialization bug." + }, + { + "left": "issue:44297", + "right": "issue:45310", + "accept": false, + "reason": "Tokenizer config save metadata mismatch is not the same bug as Qwen3.5 MoE from_pretrained failing." + }, + { + "left": "issue:43723", + "right": "issue:44977", + "accept": false, + "reason": "Tokenizer loading regression and flash-attention generation failure are different symptoms in different subsystems." + }, + { + "left": "issue:44222", + "right": "issue:44484", + "accept": false, + "reason": "FP8 save_pretrained bug is unrelated to a question about the default `max_shard_size` constant." + }, + { + "left": "issue:43232", + "right": "issue:45834", + "accept": false, + "reason": "Generation kwargs update after sync_gpus is unrelated to Kosmos2.5 OCR index errors." + }, + { + "left": "issue:44263", + "right": "issue:45092", + "accept": false, + "reason": "Torch split/indexing issue in GLM MoE is unrelated to meta-initialization incompatibility for old InternVL2 checkpoints." + }, + { + "left": "issue:43761", + "right": "issue:46082", + "accept": false, + "reason": "CLIP hidden-states regression and LlamaConfig head-dim validation are different model behaviors with different fixes." + }, + { + "left": "issue:44483", + "right": "issue:45406", + "accept": false, + "reason": "Chat-completions request handling failure is not the same as Gemma4Processor missing `_tokenizer`." + }, + { + "left": "issue:44295", + "right": "issue:44623", + "accept": false, + "reason": "Reading registered `position_ids` as a buffer and missing processor save files are different save/load bugs." + }, + { + "left": "issue:44373", + "right": "issue:44485", + "accept": false, + "reason": "A wrong docstring is documentation-only; GLM-5 RoPE implementation is a code bug." + }, + { + "left": "issue:44514", + "right": "issue:45103", + "accept": false, + "reason": "Batched `apply_chat_template` padding crash is unrelated to the auto-docstring annotation AttributeError." + }, + { + "left": "issue:43723", + "right": "issue:44871", + "accept": false, + "reason": "Generic tokenizer loading failure and inconsistent `eos_token_id` config are different tokenizer issues." + }, + { + "left": "issue:44263", + "right": "issue:44360", + "accept": false, + "reason": "Both mention DSA indexers, but one is a `torch.split` return-value problem and the other is a missing ReLU." + }, + { + "left": "issue:45092", + "right": "issue:45357", + "accept": false, + "reason": "Remote-code/meta-init incompatibility is not the same as incorrect visual encoder keys being saved." + }, + { + "left": "issue:30990", + "right": "issue:38175", + "accept": false, + "reason": "Sentence-transformers loading stall and SigLIP2 zero-probability classification are unrelated model behaviors." + }, + { + "left": "issue:42907", + "right": "issue:46050", + "accept": false, + "reason": "Both are quantization/save-load related, but they involve different quantization schemes and different failure modes." + }, + { + "left": "issue:43723", + "right": "issue:44991", + "accept": false, + "reason": "Both are tokenizer load failures, but they are model-specific reports with no evidence of the same underlying root cause." + }, + { + "left": "issue:43723", + "right": "issue:44442", + "accept": false, + "reason": "FastSpeech2ConformerTokenizer load failure is a specific tokenizer issue, not clearly the same bug as the broader v5 tokenizer regression." + }, + { + "left": "issue:44261", + "right": "issue:45092", + "accept": false, + "reason": "MLA layernorm epsilon precision issue is unrelated to remote-code checkpoint meta-initialization incompatibility." + }, + { + "left": "issue:44829", + "right": "issue:46002", + "accept": false, + "reason": "Flash-attention training degeneration is not the same underlying problem as NaNs during checkpoint/init." + }, + { + "left": "issue:41628", + "right": "issue:43824", + "accept": false, + "reason": "Missing `AutoImageProcessor` import and missing `Qwen2_5_VLForConditionalGeneration` import are different symbol-resolution problems." + }, + { + "left": "issue:42831", + "right": "issue:45310", + "accept": false, + "reason": "FineGrainedFP8 accuracy issues are unrelated to Qwen3.5 MoE from_pretrained errors." + }, + { + "left": "issue:43479", + "right": "issue:44913", + "accept": false, + "reason": "Default config initialization from `None` is unrelated to `rotary_pct` silently reverting on reload." + }, + { + "left": "issue:42831", + "right": "issue:44038", + "accept": false, + "reason": "FP8 accuracy regression and Qwen3-VL-Moe v5 bug are not the same change." + }, + { + "left": "issue:44263", + "right": "issue:44373", + "accept": false, + "reason": "Code bug in DSA indexer vs documentation issue for `position_ids`; no shared underlying defect." + }, + { + "left": "issue:39401", + "right": "issue:44568", + "accept": false, + "reason": "Qwen3 tokenizer offset mapping bug and mdeberta BOS/EOS insertion bug are distinct tokenizer behaviors." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 40, + "review_unit_count": 49, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-426", + "nodes": [ + "issue:22355", + "issue:28218", + "issue:30990", + "issue:33357", + "issue:39401", + "issue:41628", + "issue:42371", + "issue:42757", + "issue:42831", + "issue:42886", + "issue:42907", + "issue:42994", + "issue:43066", + "issue:43122", + "issue:43232", + "issue:43408", + "issue:43475", + "issue:43576", + "issue:43646", + "issue:43756", + "issue:43784", + "issue:43937", + "issue:43976", + "issue:44060", + "issue:44117", + "issue:44261", + "issue:44263", + "issue:44295", + "issue:44297", + "issue:44360", + "issue:44373", + "issue:44485", + "issue:44488", + "issue:44514", + "issue:44521", + "issue:44625", + "issue:44671", + "issue:44704", + "issue:44743", + "issue:44829", + "issue:44945", + "issue:45278", + "issue:45290", + "issue:45593", + "issue:45600", + "issue:45753", + "issue:46002", + "issue:46132" + ], + "soft_pairs": [ + "issue:42757|issue:45278", + "issue:42886|issue:42994", + "issue:43232|issue:45593", + "issue:43756|issue:44060", + "issue:42831|issue:45753", + "issue:42886|issue:43066", + "issue:22355|issue:28218", + "issue:44671|issue:46132", + "issue:43122|issue:44521", + "issue:43576|issue:43784", + "issue:43646|issue:46002", + "issue:44117|issue:44704", + "issue:30990|issue:33357", + "issue:42371|issue:44263", + "issue:44295|issue:44297", + "issue:43232|issue:44743", + "issue:44704|issue:45290", + "issue:42907|issue:44945", + "issue:43122|issue:44514", + "issue:41628|issue:45278", + "issue:39401|issue:44488", + "issue:43784|issue:45600", + "issue:43976|issue:44945", + "issue:44261|issue:44373", + "issue:43408|issue:43475", + "issue:42907|issue:43937", + "issue:39401|issue:44521", + "issue:44360|issue:44485", + "issue:44625|issue:44829" + ], + "prepared_review_unit_hash": "46c44fd348fbf681721a9a3cf6292d252171980cf9b96853cb63144ce55022af", + "component_budget": { + "node_count": 426, + "item_count": 426, + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 29, + "serialized_chars": 35827, + "estimated_input_tokens": 8957, + "estimated_eval_tokens": 18170 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is heterogeneous: most items share only broad subsystem words (imports, tokenizers, multimodal, SAM3) but not the same concrete failure mode. I reject all soft edges; none look like a single fixable duplicate pair. If a hub is needed, the broadest umbrella is the import-regression report 45278.", + "confidence": 0.41, + "canonical_issue_reason": "issue:45278 is the broadest, most umbrella-like report here, covering a version-upgrade import-regression pattern that could subsume related import failures.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45278 is the most suitable triage anchor because it is broad, regression-oriented, and discussion-friendly; the other issues are too specific and mostly unrelated.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:42757", + "right": "issue:45278", + "accept": false, + "reason": "Both mention import failures, but 42757 is a specific missing symbol from huggingface_hub while 45278 is a broad multi-import regression after upgrade; not the same concrete bug." + }, + { + "left": "issue:42886", + "right": "issue:42994", + "accept": false, + "reason": "Offline tokenizer cache loading and quantized model saving are unrelated code paths." + }, + { + "left": "issue:43232", + "right": "issue:45593", + "accept": false, + "reason": "Generation-kwargs handling after sync_gpus is unrelated to D-FINE auxiliary loss behavior." + }, + { + "left": "issue:43756", + "right": "issue:44060", + "accept": false, + "reason": "Different Qwen3-related model bugs: RoPE layer dropping versus tied-weights warning/misassignment." + }, + { + "left": "issue:42831", + "right": "issue:45753", + "accept": false, + "reason": "FineGrainedFP8 accuracy drift and Qwen3_5 image-triggered infinite loop are unrelated symptoms and subsystems." + }, + { + "left": "issue:42886", + "right": "issue:43066", + "accept": false, + "reason": "Both are tokenizer-related, but one is offline cache loading and the other is a decoder-type regression in v5; not the same bug." + }, + { + "left": "issue:22355", + "right": "issue:28218", + "accept": false, + "reason": "Missing transformers.onnx module import and added-token spacing are unrelated." + }, + { + "left": "issue:44671", + "right": "issue:46132", + "accept": false, + "reason": "CamemBERT MLM output regression and AttentionInterface.register behavior change affect different components and failures." + }, + { + "left": "issue:43122", + "right": "issue:44521", + "accept": false, + "reason": "Tokenization differs across versions, but 44521 is a multimodal assistant-mask bug; no evidence they share one fix." + }, + { + "left": "issue:43576", + "right": "issue:43784", + "accept": false, + "reason": "A broken env command and a NameError importing sentence-transformers are different regressions." + }, + { + "left": "issue:43646", + "right": "issue:46002", + "accept": false, + "reason": "Custom model initialization breakage and NaNs in classification heads are not the same code-path problem." + }, + { + "left": "issue:44117", + "right": "issue:44704", + "accept": false, + "reason": "Tokenizer mapping fallback logic and AutoProcessor cached_file kwargs handling are separate issues." + }, + { + "left": "issue:30990", + "right": "issue:33357", + "accept": false, + "reason": "Sentence-Transformers loading hang and a MacOS bus error with CLIP are different failure modes." + }, + { + "left": "issue:42371", + "right": "issue:44263", + "accept": false, + "reason": "TF32 configuration guidance and GlmMoeDsaIndexer torch.split output handling are unrelated." + }, + { + "left": "issue:44295", + "right": "issue:44297", + "accept": false, + "reason": "Reading a bufferized position_ids and mismatched tokenizer_class metadata are different bugs." + }, + { + "left": "issue:43232", + "right": "issue:44743", + "accept": false, + "reason": "Both touch generation/cache logic, but one is sync_gpus kwargs propagation and the other is recurrent state reset; not mergeable as one fix." + }, + { + "left": "issue:44704", + "right": "issue:45290", + "accept": false, + "reason": "Cached-file kwargs forwarding and apply_chat_template crashing on tool-call assistant messages are separate API bugs." + }, + { + "left": "issue:42907", + "right": "issue:44945", + "accept": false, + "reason": "Quantized model saving failure and incorrect pipeline-parallel LLM output are unrelated." + }, + { + "left": "issue:43122", + "right": "issue:44514", + "accept": false, + "reason": "Version-dependent tokenization changes and batched chat-template crash with padding=False are distinct issues." + }, + { + "left": "issue:41628", + "right": "issue:45278", + "accept": false, + "reason": "Both are import-related, but 41628 is a specific missing AutoImageProcessor export while 45278 is a broad upgrade-wide import regression; too coarse to treat as one bug." + }, + { + "left": "issue:39401", + "right": "issue:44488", + "accept": false, + "reason": "Wrong offset_mapping and failure to load a specific model are unrelated tokenizer/model-loading problems." + }, + { + "left": "issue:43784", + "right": "issue:45600", + "accept": false, + "reason": "sentence-transformers import NameError and stale Sam3LiteText references in auto_mappings.py are different import regressions." + }, + { + "left": "issue:43976", + "right": "issue:44945", + "accept": false, + "reason": "Python-version compatibility breakage and pipeline-parallel output corruption are unrelated." + }, + { + "left": "issue:44261", + "right": "issue:44373", + "accept": false, + "reason": "Missing rms_norm_eps affecting precision and a wrong docstring for position_ids are not the same underlying bug." + }, + { + "left": "issue:43408", + "right": "issue:43475", + "accept": false, + "reason": "Both are SAM3-related, but one is a model-type mismatch warning and the other is a missing attribute on an output object; separate fixes." + }, + { + "left": "issue:42907", + "right": "issue:43937", + "accept": false, + "reason": "Dequantized model saving and invalid GenerationConfig are different parts of the stack." + }, + { + "left": "issue:39401", + "right": "issue:44521", + "accept": false, + "reason": "Tokenizer offset mapping for Qwen3 and multimodal assistant-mask generation are unrelated." + }, + { + "left": "issue:44360", + "right": "issue:44485", + "accept": false, + "reason": "DSA indexer activation shape/logic and GLM-5 RoPE implementation are separate model internals." + }, + { + "left": "issue:44625", + "right": "issue:44829", + "accept": false, + "reason": "num_labels propagation in Qwen3.5 config and flash_attention_3 training degeneration 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": 1, + "component_count": 17, + "review_unit_index": 41, + "review_unit_count": 49, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-426", + "nodes": [ + "issue:22355", + "issue:28218", + "issue:29127", + "issue:29159", + "issue:30990", + "issue:36296", + "issue:38617", + "issue:42222", + "issue:42371", + "issue:42673", + "issue:42757", + "issue:42907", + "issue:43066", + "issue:43116", + "issue:43122", + "issue:43278", + "issue:43441", + "issue:43452", + "issue:43493", + "issue:43611", + "issue:43784", + "issue:43824", + "issue:43874", + "issue:43937", + "issue:43976", + "issue:43992", + "issue:44038", + "issue:44075", + "issue:44222", + "issue:44295", + "issue:44297", + "issue:44351", + "issue:44485", + "issue:44704", + "issue:44779", + "issue:44936", + "issue:44945", + "issue:44964", + "issue:44977", + "issue:45072", + "issue:45200", + "issue:45216", + "issue:45278", + "issue:45446", + "issue:45478", + "issue:45753", + "issue:46002" + ], + "soft_pairs": [ + "issue:43116|issue:45200", + "issue:42907|issue:44964", + "issue:44038|issue:44297", + "issue:28218|issue:30990", + "issue:44038|issue:44295", + "issue:43122|issue:44779", + "issue:42907|issue:43452", + "issue:36296|issue:43116", + "issue:43824|issue:44351", + "issue:44038|issue:44222", + "issue:45216|issue:45478", + "issue:42757|issue:44351", + "issue:43278|issue:44945", + "issue:43784|issue:45446", + "issue:22355|issue:29159", + "issue:43122|issue:44977", + "issue:43493|issue:45753", + "issue:42673|issue:45072", + "issue:43066|issue:44779", + "issue:38617|issue:45278", + "issue:43992|issue:44704", + "issue:22355|issue:29127", + "issue:43937|issue:44485", + "issue:43611|issue:46002", + "issue:43976|issue:45446", + "issue:42222|issue:43441", + "issue:43116|issue:43493", + "issue:42371|issue:43874", + "issue:44075|issue:44936" + ], + "prepared_review_unit_hash": "3691502de7ea9dab9b6aac684994ba28b3c0e2a28ec46c1c9b85910c0020aca4", + "component_budget": { + "node_count": 426, + "item_count": 426, + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 29, + "serialized_chars": 35277, + "estimated_input_tokens": 8820, + "estimated_eval_tokens": 17896 + }, + "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; none of the suggested soft pairs are strong enough to treat as true duplicates. The most representative issue is the broad tokenizer regression report around the v4\u2192v5 behavior change.", + "confidence": 0.34, + "canonical_issue_reason": "issue:43122 is the clearest, most general regression report in the set: it has a concrete version boundary (4.57.3 \u2192 5.0), a reproducible tokenizer behavior change, and broader relevance than the model-specific reports.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43122 is the best representative issue because it describes a concrete, user-facing tokenizer regression with a clear before/after comparison and is broad enough to summarize the tokenizer-related subset of the cluster.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43116", + "right": "issue:45200", + "accept": false, + "reason": "Different problems: example script multi-label output vs Gemma 4 multimodal token-type handling. No shared code path." + }, + { + "left": "issue:42907", + "right": "issue:44964", + "accept": false, + "reason": "Saving dequantized Ministral/Devstral weights is unrelated to loading Phi-4 multimodal with latest transformers." + }, + { + "left": "issue:44038", + "right": "issue:44297", + "accept": false, + "reason": "Qwen3-VL-Moe bug is not the same as tokenizer save_pretrained metadata mismatch." + }, + { + "left": "issue:28218", + "right": "issue:30990", + "accept": false, + "reason": "Tokenizer added-space bug and Sentence Transformers loading hang are unrelated." + }, + { + "left": "issue:44038", + "right": "issue:44295", + "accept": false, + "reason": "Different symptoms and code paths: Qwen3-VL-Moe failure vs position_ids buffer read error." + }, + { + "left": "issue:43122", + "right": "issue:44779", + "accept": false, + "reason": "Both are tokenizer regressions in v5, but one is generic tokenization drift and the other is Deepseek-specific incorrect outputs; not enough evidence of the same root cause." + }, + { + "left": "issue:42907", + "right": "issue:43452", + "accept": false, + "reason": "Dequantized save failure and gguf_file loading failure are different loading/saving paths and different bugs." + }, + { + "left": "issue:36296", + "right": "issue:43116", + "accept": false, + "reason": "Tensor-parallel training bug is unrelated to the classification example script issue." + }, + { + "left": "issue:43824", + "right": "issue:44351", + "accept": false, + "reason": "Both are import errors, but for different symbols and different underlying missing exports." + }, + { + "left": "issue:44038", + "right": "issue:44222", + "accept": false, + "reason": "Different Qwen3-VL-Moe inference bug vs FP8 save_pretrained-on-MoE bug." + }, + { + "left": "issue:45216", + "right": "issue:45478", + "accept": false, + "reason": "Same model family, but save_pretrained checkpoint corruption and from_pretrained loading error are distinct code paths and not clearly the same defect." + }, + { + "left": "issue:42757", + "right": "issue:44351", + "accept": false, + "reason": "Different import failures from different libraries/modules; not a duplicate." + }, + { + "left": "issue:43278", + "right": "issue:44945", + "accept": false, + "reason": "BF16-to-FP32 dtype shift in eval is unrelated to incorrect outputs under pipeline parallelism." + }, + { + "left": "issue:43784", + "right": "issue:45446", + "accept": false, + "reason": "Sentence-transformers import NameError and flex_attention version-check bug are unrelated." + }, + { + "left": "issue:22355", + "right": "issue:29159", + "accept": false, + "reason": "Missing transformers.onnx module and slow/fast tokenizer inconsistency are separate issues." + }, + { + "left": "issue:43122", + "right": "issue:44977", + "accept": false, + "reason": "Tokenizer regression and flash-attention generation failure are different symptoms and likely different paths." + }, + { + "left": "issue:43493", + "right": "issue:45753", + "accept": false, + "reason": "SigLIP2 implementation discrepancy is unrelated to Qwen3_5 image-induced infinite loop." + }, + { + "left": "issue:42673", + "right": "issue:45072", + "accept": false, + "reason": "VRAM leak in multi-threaded Qwen3ForCausalLM use is unrelated to dtype mismatch in bfloat16 inference." + }, + { + "left": "issue:43066", + "right": "issue:44779", + "accept": false, + "reason": "Wrong tokenizer decoder type and Deepseek tokenization regression are related only at a high level; not the same concrete bug." + }, + { + "left": "issue:38617", + "right": "issue:45278", + "accept": false, + "reason": "Specific missing import versus broad import errors after upgrade; not the same issue." + }, + { + "left": "issue:43992", + "right": "issue:44704", + "accept": false, + "reason": "Missing embed_tokens.weight in UMT5Encoder and AutoProcessor kwargs forwarding are unrelated." + }, + { + "left": "issue:22355", + "right": "issue:29127", + "accept": false, + "reason": "transformers.onnx import failure and LayoutLMv3 error-message clarity are unrelated." + }, + { + "left": "issue:43937", + "right": "issue:44485", + "accept": false, + "reason": "Invalid GenerationConfig and GLM-5 RoPE discussion do not describe the same defect." + }, + { + "left": "issue:43611", + "right": "issue:46002", + "accept": false, + "reason": "base_model_prefix loading regression and NaNs in classification heads are different problems." + }, + { + "left": "issue:43976", + "right": "issue:45446", + "accept": false, + "reason": "Python version support issue and flex_attention version check bug are unrelated." + }, + { + "left": "issue:42222", + "right": "issue:43441", + "accept": false, + "reason": "VitPose model breakage and Ministral-3 FlashAttention failure are different model-specific bugs." + }, + { + "left": "issue:43116", + "right": "issue:43493", + "accept": false, + "reason": "Classification example script issue and SigLIP2 implementation discrepancy are unrelated." + }, + { + "left": "issue:42371", + "right": "issue:43874", + "accept": false, + "reason": "TF32 API guidance and missing image patch count method are unrelated." + }, + { + "left": "issue:44075", + "right": "issue:44936", + "accept": false, + "reason": "Unused SGD args and trainer.evaluate() failing after train() are different training-loop bugs." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 42, + "review_unit_count": 49, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-426", + "nodes": [ + "issue:33453", + "issue:38617", + "issue:39401", + "issue:41628", + "issue:42371", + "issue:42490", + "issue:42617", + "issue:42673", + "issue:43064", + "issue:43116", + "issue:43408", + "issue:43421", + "issue:43450", + "issue:43582", + "issue:43673", + "issue:43742", + "issue:43825", + "issue:43827", + "issue:43906", + "issue:43937", + "issue:43986", + "issue:44038", + "issue:44117", + "issue:44261", + "issue:44291", + "issue:44380", + "issue:44410", + "issue:44466", + "issue:44484", + "issue:44779", + "issue:44829", + "issue:44977", + "issue:45071", + "issue:45081", + "issue:45083", + "issue:45216", + "issue:45310", + "issue:45341", + "issue:45357", + "issue:45405", + "issue:45446", + "issue:45478", + "issue:45561", + "issue:45563", + "issue:45588", + "issue:46002", + "issue:46077" + ], + "soft_pairs": [ + "issue:43064|issue:43673", + "issue:44038|issue:45478", + "issue:45357|issue:45478", + "issue:43116|issue:43825", + "issue:45083|issue:45446", + "issue:43937|issue:45563", + "issue:44380|issue:46077", + "issue:33453|issue:39401", + "issue:39401|issue:44779", + "issue:43827|issue:45446", + "issue:44117|issue:45588", + "issue:42673|issue:43906", + "issue:43937|issue:44977", + "issue:44291|issue:45071", + "issue:43450|issue:43986", + "issue:44466|issue:46002", + "issue:45405|issue:45446", + "issue:45341|issue:45561", + "issue:44117|issue:45081", + "issue:38617|issue:41628", + "issue:44410|issue:45310", + "issue:44829|issue:45216", + "issue:39401|issue:43421", + "issue:42673|issue:43582", + "issue:42617|issue:45341", + "issue:44484|issue:45310", + "issue:42371|issue:44261", + "issue:42490|issue:42673", + "issue:43408|issue:43742" + ], + "prepared_review_unit_hash": "934fba96391f7caf15c0c8d523d2fcfb183bb63406167fd4f44c255349c1cd46", + "component_budget": { + "node_count": 426, + "item_count": 426, + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 29, + "serialized_chars": 35850, + "estimated_input_tokens": 8963, + "estimated_eval_tokens": 18182 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This is not a tight duplicate cluster; it mixes many unrelated bugs across tokenizers, Qwen/Qwen3.5, training, imports, docs, and testing. I would not merge any of the proposed soft edges.", + "confidence": 0.61, + "canonical_issue_reason": "No single true canonical exists here because the items are heterogeneous; if a placeholder is required, 43825 is the cleanest user-facing runtime bug and has some discussion/inbound references.", + "canonical_pr_reason": null, + "best_issue_reason": "43825 is the most concrete and broadly understandable issue in the set, but it still does not represent the rest of the unrelated cluster.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43064", + "right": "issue:43673", + "accept": false, + "reason": "FSDP/PEFT optimizer-state corruption is a different bug than missing GenerationMixin cache during chunked prefill." + }, + { + "left": "issue:44038", + "right": "issue:45478", + "accept": false, + "reason": "Both mention Qwen3.5-era model loading, but one is a generic Qwen3-VL-Moe v5 issue and the other is a specific from_pretrained failure; not the same bug." + }, + { + "left": "issue:45357", + "right": "issue:45478", + "accept": false, + "reason": "One is incorrect save_pretrained visual encoder keys; the other is a from_pretrained load error. Different code paths and symptoms." + }, + { + "left": "issue:43116", + "right": "issue:43825", + "accept": false, + "reason": "Multi-label classification example behavior is unrelated to the pipeline translation-task message regression." + }, + { + "left": "issue:45083", + "right": "issue:45446", + "accept": false, + "reason": "qwen3_omni_moe feature-length behavior and a flex_attention PyTorch version check are unrelated." + }, + { + "left": "issue:43937", + "right": "issue:45563", + "accept": false, + "reason": "Invalid GenerationConfig handling is not the same as a stale num_return_sequences warning in paged generate." + }, + { + "left": "issue:44380", + "right": "issue:46077", + "accept": false, + "reason": "GPT-2 attention scaling under SDPA/FlashAttention and Gemma4 bidirectional mask construction are different model/backend bugs." + }, + { + "left": "issue:33453", + "right": "issue:39401", + "accept": false, + "reason": "Tokenizer loading regression and Qwen3 offset_mapping are both tokenizer-related but are different failures with different models/symptoms." + }, + { + "left": "issue:39401", + "right": "issue:44779", + "accept": false, + "reason": "Different tokenizer bugs affecting different models: offset_mapping vs incorrect DeepSeek tokenization results." + }, + { + "left": "issue:43827", + "right": "issue:45446", + "accept": false, + "reason": "Docs still referencing pipeline() is unrelated to the flex_attention import/version check." + }, + { + "left": "issue:44117", + "right": "issue:45588", + "accept": false, + "reason": "TOKENIZER_MAPPING_NAMES returning None is unrelated to flash_attention crash on sink-less models." + }, + { + "left": "issue:42673", + "right": "issue:43906", + "accept": false, + "reason": "VRAM leak across dataloader threads is not the same as an isolated reproduction of a different issue." + }, + { + "left": "issue:43937", + "right": "issue:44977", + "accept": false, + "reason": "Generic invalid generation config handling is different from Qwen3.5 flash-attention generation failure." + }, + { + "left": "issue:44291", + "right": "issue:45071", + "accept": false, + "reason": "Unexpected _is_hf_initialized argument in init_empty_weights is unrelated to v5.4.0 PretrainedConfig type checking." + }, + { + "left": "issue:43450", + "right": "issue:43986", + "accept": false, + "reason": "Batched video processor shape bug and missing torchvision import crash are separate video-processing problems." + }, + { + "left": "issue:44466", + "right": "issue:46002", + "accept": false, + "reason": "lm_head.weight serialization inconsistency is unrelated to NaNs in classification heads after checkout/init." + }, + { + "left": "issue:45405", + "right": "issue:45446", + "accept": false, + "reason": "A PEFT version bump problem is unrelated to the flex_attention AuxRequest version check." + }, + { + "left": "issue:45341", + "right": "issue:45561", + "accept": false, + "reason": "A small testing_utils bug is not the same as the pytest-xdist captured_info.txt race." + }, + { + "left": "issue:44117", + "right": "issue:45081", + "accept": false, + "reason": "Tokenizer mapping returning None is a different failure from _patch_mistral_regex crashing on backend_tokenizer access." + }, + { + "left": "issue:38617", + "right": "issue:41628", + "accept": false, + "reason": "Both are import errors, but they concern different missing symbols from different modules and code paths." + }, + { + "left": "issue:44410", + "right": "issue:45310", + "accept": false, + "reason": "Missing qkv/gate projections in qwen3next is a model-architecture issue, not the same as Qwen3.5 Moe from_pretrained failure." + }, + { + "left": "issue:44829", + "right": "issue:45216", + "accept": false, + "reason": "FlashAttention-3 causing degenerate training is unrelated to Qwen3.5 save_pretrained checkpoint regression." + }, + { + "left": "issue:39401", + "right": "issue:43421", + "accept": false, + "reason": "Qwen3 tokenizer offset_mapping wrong results and runtime post-processor updates for special tokens are different tokenizer behaviors." + }, + { + "left": "issue:42673", + "right": "issue:43582", + "accept": false, + "reason": "A VRAM leak in multi-threaded loading is unrelated to an Apple Silicon TypeError in caching_allocator_warmup." + }, + { + "left": "issue:42617", + "right": "issue:45341", + "accept": false, + "reason": "Failure to run 3d_parallel.py is unrelated to a minor testing_utils issue." + }, + { + "left": "issue:44484", + "right": "issue:45310", + "accept": false, + "reason": "save_pretrained max_shard_size behavior is unrelated to Qwen3.5 Moe loading errors." + }, + { + "left": "issue:42371", + "right": "issue:44261", + "accept": false, + "reason": "TF32 behavior settings and MLA rms_norm_eps precision mismatch are unrelated issues." + }, + { + "left": "issue:42490", + "right": "issue:42673", + "accept": false, + "reason": "Trailing-slash import warning when loading a local model is not the same as a VRAM leak across dataloader threads." + }, + { + "left": "issue:43408", + "right": "issue:43742", + "accept": false, + "reason": "sam3_video/sam3_tracker warning is unrelated to a KeyError while loading facebook/MobileLLM-125M." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 43, + "review_unit_count": 49, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-426", + "nodes": [ + "issue:38617", + "issue:41628", + "issue:42175", + "issue:42222", + "issue:42371", + "issue:42831", + "issue:43012", + "issue:43066", + "issue:43097", + "issue:43122", + "issue:43335", + "issue:43408", + "issue:43441", + "issue:43452", + "issue:43577", + "issue:43673", + "issue:43874", + "issue:43931", + "issue:43937", + "issue:43986", + "issue:44038", + "issue:44075", + "issue:44117", + "issue:44220", + "issue:44261", + "issue:44263", + "issue:44360", + "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:45278", + "issue:45341", + "issue:45459", + "issue:45464", + "issue:45478", + "issue:45542", + "issue:45656", + "issue:46123", + "issue:46172" + ], + "soft_pairs": [ + "issue:44933|issue:45278", + "issue:38617|issue:44220", + "issue:42222|issue:45542", + "issue:43937|issue:44496", + "issue:42175|issue:42222", + "issue:44483|issue:45464", + "issue:43066|issue:43122", + "issue:42371|issue:42831", + "issue:44075|issue:45656", + "issue:43012|issue:43577", + "issue:44410|issue:45478", + "issue:43673|issue:45341", + "issue:44484|issue:45478", + "issue:43097|issue:44493", + "issue:41628|issue:45071", + "issue:43874|issue:44261", + "issue:44263|issue:44485", + "issue:45071|issue:45542", + "issue:42222|issue:45459", + "issue:44263|issue:46172", + "issue:44360|issue:46172", + "issue:43012|issue:43408", + "issue:43441|issue:44829", + "issue:43931|issue:44410", + "issue:43673|issue:46123", + "issue:43931|issue:44038", + "issue:43335|issue:45071", + "issue:43097|issue:43452", + "issue:45003|issue:45341", + "issue:43986|issue:45341", + "issue:44117|issue:44987", + "issue:43408|issue:44368", + "issue:43986|issue:44484", + "issue:43874|issue:44263" + ], + "prepared_review_unit_hash": "03800beb213eaf1f5d42c9a0cdd0e3db072ca85468a2dcceb8eb4bb0be1b4301", + "component_budget": { + "node_count": 426, + "item_count": 426, + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 34, + "serialized_chars": 37583, + "estimated_input_tokens": 9396, + "estimated_eval_tokens": 19048 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These items mostly form a noisy similarity cluster: many are Transformers regressions, but the proposed edges usually connect different models, different code paths, or only broadly related symptom classes. I found no pair that clearly looks like the same underlying bug or change.", + "confidence": 0.82, + "canonical_issue_reason": "Issue 45278 is the broadest, most umbrella-like report in the set: a post-upgrade import-regression with multiple failures, making it the best representative issue among these mostly unrelated reports.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 45278 has the clearest generalized regression framing and the widest apparent scope, so it is the strongest single issue to keep as the cluster representative.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44933", + "right": "issue:45278", + "accept": false, + "reason": "Both mention imports, but one is a specific missing import from image_utils while the other is a broad upgrade-wide import regression; not the same bug." + }, + { + "left": "issue:38617", + "right": "issue:44220", + "accept": false, + "reason": "Different failures in different areas: a missing configuration utility symbol vs. a feature-extraction helper issue." + }, + { + "left": "issue:42222", + "right": "issue:45542", + "accept": false, + "reason": "One is about broken vitpose models, the other about TensorFlow backend detection when only tensorboard is installed." + }, + { + "left": "issue:43937", + "right": "issue:44496", + "accept": false, + "reason": "Different model-loading/validation failures affecting unrelated models and error paths." + }, + { + "left": "issue:42175", + "right": "issue:42222", + "accept": false, + "reason": "TensorFlow installation/backend packaging problem vs. vitpose model breakage; too different." + }, + { + "left": "issue:44483", + "right": "issue:45464", + "accept": false, + "reason": "Both touch chat/completions, but one is a generic API rejection and the other is a specific streaming-inference failure for a particular model; not clearly the same bug." + }, + { + "left": "issue:43066", + "right": "issue:43122", + "accept": false, + "reason": "Both are tokenizer regressions in v5, but one is decoder-type mismatch and the other is changed tokenization output; not enough to treat as the same defect." + }, + { + "left": "issue:42371", + "right": "issue:42831", + "accept": false, + "reason": "TF32 API-setting guidance is unrelated to the FineGrainedFP8 accuracy regression." + }, + { + "left": "issue:44075", + "right": "issue:45656", + "accept": false, + "reason": "SGD argument handling and DeepSpeed double-stepping are different optimizer bugs." + }, + { + "left": "issue:43012", + "right": "issue:43577", + "accept": false, + "reason": "Both involve dtype behavior, but one is a PyTorch warning during compilation and the other is Blip2 load-time dtype propagation; different concrete problems." + }, + { + "left": "issue:44410", + "right": "issue:45478", + "accept": false, + "reason": "Different models and different symptoms: missing projections in qwen3next vs. Qwen3.5 MoE from_pretrained error." + }, + { + "left": "issue:43673", + "right": "issue:45341", + "accept": false, + "reason": "Generation cache/chunked-prefill bug is unrelated to a testing_utils issue." + }, + { + "left": "issue:44484", + "right": "issue:45478", + "accept": false, + "reason": "save_pretrained sharding-size question is unrelated to Qwen3.5 MoE loading failure." + }, + { + "left": "issue:43097", + "right": "issue:44493", + "accept": false, + "reason": "Removed tie_embeddings behavior and unexpected position-id keys are different regression classes." + }, + { + "left": "issue:41628", + "right": "issue:45071", + "accept": false, + "reason": "AutoImageProcessor import failure and PretrainedConfig type-checking regression are distinct issues." + }, + { + "left": "issue:43874", + "right": "issue:44261", + "accept": false, + "reason": "Missing image-patch counting helper vs. rms_norm_eps precision issue; different models and paths." + }, + { + "left": "issue:44263", + "right": "issue:44485", + "accept": false, + "reason": "DSA indexer split behavior and GLM-5 RoPE implementation are not the same concrete bug." + }, + { + "left": "issue:45071", + "right": "issue:45542", + "accept": false, + "reason": "Type-checking regression in config handling is unrelated to TensorFlow backend import/installation behavior." + }, + { + "left": "issue:42222", + "right": "issue:45459", + "accept": false, + "reason": "Vitpose model breakage and tokenizer error-hiding when protobuf is absent are unrelated." + }, + { + "left": "issue:44263", + "right": "issue:46172", + "accept": false, + "reason": "Both are tensor/indexing-related, but the concrete code paths and symptoms differ." + }, + { + "left": "issue:44360", + "right": "issue:46172", + "accept": false, + "reason": "ReLU missing in the DSA indexer and a dtype/transpose issue in deepseek-v4 are different bugs." + }, + { + "left": "issue:43012", + "right": "issue:43408", + "accept": false, + "reason": "A bfloat16 precision warning is unrelated to the sam3_video/sam3_tracker model-type mismatch." + }, + { + "left": "issue:43441", + "right": "issue:44829", + "accept": false, + "reason": "Both mention FlashAttention, but they affect different models and failure modes; not the same underlying change." + }, + { + "left": "issue:43931", + "right": "issue:44410", + "accept": false, + "reason": "Qwen3-VL weight-shape mismatch and qwen3next missing projections are different model-specific bugs." + }, + { + "left": "issue:43673", + "right": "issue:46123", + "accept": false, + "reason": "Cache handling during chunked prefill is unrelated to the final partial-batch masking pipeline bug." + }, + { + "left": "issue:43931", + "right": "issue:44038", + "accept": false, + "reason": "Different Qwen3-VL-related failures with different symptoms and likely different fixes." + }, + { + "left": "issue:43335", + "right": "issue:45071", + "accept": false, + "reason": "Sparse-layer construction bug and config type-checking regression are unrelated." + }, + { + "left": "issue:43097", + "right": "issue:43452", + "accept": false, + "reason": "Embedding-tie removal and gguf_file loading failures are different code paths." + }, + { + "left": "issue:45003", + "right": "issue:45341", + "accept": false, + "reason": "Unsafe sys.modules access in modeling_utils is unrelated to a testing_utils bug." + }, + { + "left": "issue:43986", + "right": "issue:45341", + "accept": false, + "reason": "AutoProcessor crashing without torchvision and a testing_utils issue are unrelated." + }, + { + "left": "issue:44117", + "right": "issue:44987", + "accept": false, + "reason": "Tokenizer mapping None handling and a broad model-loading failure are too different to merge." + }, + { + "left": "issue:43408", + "right": "issue:44368", + "accept": false, + "reason": "sam3 model-type warning and a Qwen3.5 LoRA tie_word_embeddings warning are unrelated." + }, + { + "left": "issue:43986", + "right": "issue:44484", + "accept": false, + "reason": "Video-model processor crash and max_shard_size behavior are unrelated." + }, + { + "left": "issue:43874", + "right": "issue:44263", + "accept": false, + "reason": "Missing image-patch helper and torch.split return-value behavior are different implementation bugs." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 44, + "review_unit_count": 49, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-426", + "nodes": [ + "issue:38617", + "issue:42757", + "issue:42907", + "issue:43012", + "issue:43064", + "issue:43066", + "issue:43097", + "issue:43208", + "issue:43352", + "issue:43502", + "issue:43531", + "issue:43550", + "issue:43792", + "issue:43824", + "issue:43825", + "issue:43873", + "issue:43906", + "issue:43976", + "issue:43986", + "issue:43992", + "issue:44038", + "issue:44062", + "issue:44117", + "issue:44162", + "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:45103", + "issue:45341", + "issue:45372", + "issue:45375", + "issue:45440", + "issue:45479", + "issue:45588", + "issue:45753", + "issue:46050", + "issue:46123", + "issue:46132", + "issue:46172" + ], + "soft_pairs": [ + "issue:45440|issue:46172", + "issue:43531|issue:45341", + "issue:43986|issue:45372", + "issue:38617|issue:42757", + "issue:44261|issue:44485", + "issue:43064|issue:44315", + "issue:43097|issue:46050", + "issue:43097|issue:45588", + "issue:44062|issue:44117", + "issue:44410|issue:45375", + "issue:38617|issue:44351", + "issue:44038|issue:44410", + "issue:43502|issue:43992", + "issue:43097|issue:43873", + "issue:44704|issue:45103", + "issue:43986|issue:44117", + "issue:43066|issue:43906", + "issue:43824|issue:43986", + "issue:43208|issue:43550", + "issue:43825|issue:44655", + "issue:43992|issue:44484", + "issue:43012|issue:43550", + "issue:43012|issue:44368", + "issue:43352|issue:44380", + "issue:42907|issue:45753", + "issue:42907|issue:44655", + "issue:44162|issue:46172", + "issue:43792|issue:46123", + "issue:43352|issue:43976", + "issue:44380|issue:46132", + "issue:44380|issue:45479", + "issue:43064|issue:44380", + "issue:44483|issue:44514" + ], + "prepared_review_unit_hash": "313f5ec691155985b7999297ba66fea9145d7134ffdacdee780b99daf5af1e91", + "component_budget": { + "node_count": 426, + "item_count": 426, + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 33, + "serialized_chars": 37467, + "estimated_input_tokens": 9367, + "estimated_eval_tokens": 18990 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These items are mostly unrelated Transformers bugs spanning imports, model configs, processor loading, training, and attention backends. The soft matches are mostly broad subsystem similarity rather than the same underlying defect, so none should be merged as duplicates.", + "confidence": 0.96, + "canonical_issue_reason": "issue:43097 is the clearest representative of a user-facing v5 regression with explicit impact and inbound references, making it a strong anchor for triage of version-breakage reports.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43873 is the best standalone issue to prioritize: it is open, active, and describes a broadly impactful, concrete failure mode around offline/remote loading behavior.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45440", + "right": "issue:46172", + "accept": false, + "reason": "Different DeepSeek model families and different bug classes; not the same code path." + }, + { + "left": "issue:43531", + "right": "issue:45341", + "accept": false, + "reason": "Sliding-window model behavior vs a testing utility bug; unrelated." + }, + { + "left": "issue:43986", + "right": "issue:45372", + "accept": false, + "reason": "Both involve processor loading, but the missing dependency/import and affected paths differ." + }, + { + "left": "issue:38617", + "right": "issue:42757", + "accept": false, + "reason": "Separate ImportErrors from different modules and symbols." + }, + { + "left": "issue:44261", + "right": "issue:44485", + "accept": false, + "reason": "Precision warning in MLA layer norm vs GLM-5 RoPE implementation; unrelated defects." + }, + { + "left": "issue:43064", + "right": "issue:44315", + "accept": false, + "reason": "Optimizer-state corruption under FSDP/PEFT vs Liger kernel not being applied; different problems." + }, + { + "left": "issue:43097", + "right": "issue:46050", + "accept": false, + "reason": "Config removal regression vs quantized save/load corruption; not the same bug." + }, + { + "left": "issue:43097", + "right": "issue:45588", + "accept": false, + "reason": "v5 config change issue vs flash-attention crash on sink-less models." + }, + { + "left": "issue:44062", + "right": "issue:44117", + "accept": false, + "reason": "AddedToken constructor argument clash vs tokenizer-mapping None handling; different failure points." + }, + { + "left": "issue:44410", + "right": "issue:45375", + "accept": false, + "reason": "Missing projections in a model layer vs a missing config field silently dropped." + }, + { + "left": "issue:38617", + "right": "issue:44351", + "accept": false, + "reason": "Different missing-import regressions involving different names and modules." + }, + { + "left": "issue:44038", + "right": "issue:44410", + "accept": false, + "reason": "Qwen3-VL-MoE bug vs Qwen3next projection bug; same broad family, different concrete issue." + }, + { + "left": "issue:43502", + "right": "issue:43992", + "accept": false, + "reason": "Offline-download behavior vs missing embedding weights on load; unrelated." + }, + { + "left": "issue:43097", + "right": "issue:43873", + "accept": false, + "reason": "Config API removal vs quantization offloading behavior; different code paths." + }, + { + "left": "issue:44704", + "right": "issue:45103", + "accept": false, + "reason": "cached_file kwargs propagation vs a docstring/annotation crash; not the same defect." + }, + { + "left": "issue:43986", + "right": "issue:44117", + "accept": false, + "reason": "Processor/torchvision loading crash vs tokenizer mapping lookup bug." + }, + { + "left": "issue:43066", + "right": "issue:43906", + "accept": false, + "reason": "Wrong tokenizer decoder type vs an unrelated reproduced issue; no concrete overlap." + }, + { + "left": "issue:43824", + "right": "issue:43986", + "accept": false, + "reason": "Missing exported model class vs AutoProcessor/torchvision crash; separate import and runtime paths." + }, + { + "left": "issue:43208", + "right": "issue:43550", + "accept": false, + "reason": "xLSTM training blockers vs Bamba compile/SDPA failure; different models and failures." + }, + { + "left": "issue:43825", + "right": "issue:44655", + "accept": false, + "reason": "Pipeline error message vs pipeline serialization bug; not the same underlying change." + }, + { + "left": "issue:43992", + "right": "issue:44484", + "accept": false, + "reason": "Missing checkpoint weight vs max_shard_size default behavior; unrelated." + }, + { + "left": "issue:43012", + "right": "issue:43550", + "accept": false, + "reason": "PyTorch warning during bfloat16 compilation vs a torch.compile SDPA model failure." + }, + { + "left": "issue:43012", + "right": "issue:44368", + "accept": false, + "reason": "Different warnings with different root causes and model setups." + }, + { + "left": "issue:43352", + "right": "issue:44380", + "accept": false, + "reason": "Unsupported Flash Attention 2.0 for one model vs GPT2 attention-scaling being ignored." + }, + { + "left": "issue:42907", + "right": "issue:45753", + "accept": false, + "reason": "Save/load dequantization bug vs an infinite-loop vision prompt bug." + }, + { + "left": "issue:42907", + "right": "issue:44655", + "accept": false, + "reason": "Model save/pretrained quantization issue vs pipeline object serialization." + }, + { + "left": "issue:44162", + "right": "issue:46172", + "accept": false, + "reason": "ESM2 breakage vs a DeepSeek V4 tensor transpose bug; unrelated." + }, + { + "left": "issue:43792", + "right": "issue:46123", + "accept": false, + "reason": "Whisper model load failure vs partial-batch mask pipeline bookkeeping bug." + }, + { + "left": "issue:43352", + "right": "issue:43976", + "accept": false, + "reason": "Attention backend support issue vs Python version incompatibility." + }, + { + "left": "issue:44380", + "right": "issue:46132", + "accept": false, + "reason": "GPT2 backend-specific scaling bug vs AttentionInterface.register changing callback behavior." + }, + { + "left": "issue:44380", + "right": "issue:45479", + "accept": false, + "reason": "Attention backend config bug vs zero-loss sequence-classification bug; different subsystems." + }, + { + "left": "issue:43064", + "right": "issue:44380", + "accept": false, + "reason": "Distributed optimizer-state corruption vs attention backend scaling behavior." + }, + { + "left": "issue:44483", + "right": "issue:44514", + "accept": false, + "reason": "Chat/completions API rejection vs batched processor apply_chat_template crash." + }, + { + "left": "issue:45440", + "right": "issue:46172", + "accept": false, + "reason": "Different DeepSeek versions and different concrete implementation bugs; not mergeable as one fix." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 45, + "review_unit_count": 49, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-426", + "nodes": [ + "issue:36010", + "issue:36683", + "issue:38617", + "issue:42490", + "issue:42757", + "issue:42898", + "issue:42994", + "issue:43012", + "issue:43066", + "issue:43208", + "issue:43299", + "issue:43352", + "issue:43475", + "issue:43502", + "issue:43525", + "issue:43526", + "issue:43582", + "issue:43611", + "issue:43618", + "issue:43756", + "issue:43761", + "issue:43824", + "issue:43867", + "issue:43901", + "issue:43937", + "issue:43976", + "issue:44038", + "issue:44077", + "issue:44222", + "issue:44230", + "issue:44380", + "issue:44483", + "issue:44485", + "issue:44625", + "issue:44655", + "issue:44683", + "issue:44938", + "issue:45216", + "issue:45276", + "issue:45290", + "issue:45381", + "issue:45478", + "issue:45593", + "issue:45698", + "issue:45753", + "issue:46077", + "issue:46082", + "issue:46172" + ], + "soft_pairs": [ + "issue:44222|issue:45478", + "issue:36683|issue:43582", + "issue:43208|issue:43526", + "issue:43012|issue:44683", + "issue:42490|issue:43012", + "issue:42757|issue:43824", + "issue:43618|issue:43761", + "issue:45698|issue:46172", + "issue:43618|issue:44380", + "issue:44038|issue:44230", + "issue:38617|issue:43502", + "issue:44483|issue:45290", + "issue:44483|issue:45381", + "issue:43299|issue:43867", + "issue:43756|issue:44485", + "issue:36683|issue:43475", + "issue:43901|issue:44655", + "issue:45276|issue:46077", + "issue:43299|issue:45753", + "issue:42994|issue:44655", + "issue:42898|issue:43066", + "issue:36010|issue:42757", + "issue:44077|issue:44380", + "issue:36683|issue:44077", + "issue:43867|issue:45216", + "issue:43352|issue:43937", + "issue:43611|issue:44655", + "issue:44625|issue:46077", + "issue:43976|issue:44938", + "issue:46077|issue:46082", + "issue:43352|issue:45593", + "issue:36683|issue:43525", + "issue:43618|issue:45216" + ], + "prepared_review_unit_hash": "bfd191d700ef37b91fea7b12a506836b3beb3b37108f7c0f917e4e55027c54dd", + "component_budget": { + "node_count": 426, + "item_count": 426, + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 33, + "serialized_chars": 37104, + "estimated_input_tokens": 9276, + "estimated_eval_tokens": 18808 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This set is mostly a grab bag of distinct Transformers regressions, import errors, and model-specific bugs. I did not find any soft pair that looks like the same underlying change, so I would not merge any of them as duplicates.", + "confidence": 0.34, + "canonical_issue_reason": "issue:45698 is the broadest representative here: a core from_pretrained/save_pretrained loading/saving bug with active discussion and inbound references, unlike the many narrow model-specific issues.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45698 is the best global representative because it targets a central library API path (model loading/saving) and has comparatively high discussion activity.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44222", + "right": "issue:45478", + "accept": false, + "reason": "Both involve Qwen3.5/MoE serialization, but one is FP8 save_pretrained and the other is a from_pretrained load error; different failure modes." + }, + { + "left": "issue:36683", + "right": "issue:43582", + "accept": false, + "reason": "Gemma3 config attribute error vs Apple Silicon allocator TypeError; unrelated bugs." + }, + { + "left": "issue:43208", + "right": "issue:43526", + "accept": false, + "reason": "xLSTM training bugs are unrelated to a BEiT image-processor label reduction bug." + }, + { + "left": "issue:43012", + "right": "issue:44683", + "accept": false, + "reason": "One is a warning during bfloat16 compilation, the other is a hard failure in compiled flex_attention on torch>=2.9; not the same issue." + }, + { + "left": "issue:42490", + "right": "issue:43012", + "accept": false, + "reason": "Local-path slash import warning and bfloat16 compile warning are different code paths." + }, + { + "left": "issue:42757", + "right": "issue:43824", + "accept": false, + "reason": "Different missing imports from different packages/symbols; not a duplicate." + }, + { + "left": "issue:43618", + "right": "issue:43761", + "accept": false, + "reason": "Both are CLIP v5 regressions, but one drops attentions and the other drops hidden_states; no clear shared concrete bug." + }, + { + "left": "issue:45698", + "right": "issue:46172", + "accept": false, + "reason": "Generic load/save module-resolution bug vs DeepSeek dtype/transpose bug; unrelated." + }, + { + "left": "issue:43618", + "right": "issue:44380", + "accept": false, + "reason": "CLIP output-field regression and GPT2 attention-scaling backend behavior are separate problems." + }, + { + "left": "issue:44038", + "right": "issue:44230", + "accept": false, + "reason": "Same family (Qwen3-VL/MoE) but different issues: generic loading breakage vs fp8 inference support." + }, + { + "left": "issue:38617", + "right": "issue:43502", + "accept": false, + "reason": "ImportError for a missing utility vs network requests despite local_files_only=True; unrelated." + }, + { + "left": "issue:44483", + "right": "issue:45290", + "accept": false, + "reason": "OpenAI chat-completions request validation and chat-template tokenization crash are different paths." + }, + { + "left": "issue:44483", + "right": "issue:45381", + "accept": false, + "reason": "API request rejection and Qwen2.5-VL video position-id bug are unrelated." + }, + { + "left": "issue:43299", + "right": "issue:43867", + "accept": false, + "reason": "Both are loading errors, but one is a Qwen3-VL/MoE regression and the other is a state_dict-order issue; not the same root cause." + }, + { + "left": "issue:43756", + "right": "issue:44485", + "accept": false, + "reason": "Smollm3 RoPE-layer mismatch and GLM-5 RoPE discussion are different model-specific bugs." + }, + { + "left": "issue:36683", + "right": "issue:43475", + "accept": false, + "reason": "Different models and different missing attributes; no shared code path." + }, + { + "left": "issue:43901", + "right": "issue:44655", + "accept": false, + "reason": "Docs/behavior mismatch for a pipeline arg is unrelated to pipeline save_pretrained failure." + }, + { + "left": "issue:45276", + "right": "issue:46077", + "accept": false, + "reason": "Gemma4 embedding resize propagation bug vs Gemma4 attention-mask bug; separate issues." + }, + { + "left": "issue:42994", + "right": "issue:44655", + "accept": false, + "reason": "Quantized model saving failure is not the same as saving Pipeline objects." + }, + { + "left": "issue:42898", + "right": "issue:43066", + "accept": false, + "reason": "Tokenization cleanup behavior change and wrong tokenizer decoder type are distinct tokenizer regressions." + }, + { + "left": "issue:36010", + "right": "issue:42757", + "accept": false, + "reason": "Both are import errors, but for different symbols/modules and likely unrelated fixes." + }, + { + "left": "issue:44077", + "right": "issue:44380", + "accept": false, + "reason": "patchtsmixer post_init validation is unrelated to GPT2 attention scaling under SDPA/FlashAttention." + }, + { + "left": "issue:36683", + "right": "issue:44077", + "accept": false, + "reason": "Gemma3 config attribute bug and patchtsmixer post_init policy change are unrelated." + }, + { + "left": "issue:43867", + "right": "issue:45216", + "accept": false, + "reason": "Sorted-state_dict load failure and incorrect Qwen3.5 checkpoint save are different serialization bugs." + }, + { + "left": "issue:43352", + "right": "issue:43937", + "accept": false, + "reason": "Flash Attention support warning and invalid GenerationConfig error are unrelated." + }, + { + "left": "issue:43611", + "right": "issue:44655", + "accept": false, + "reason": "Base-model-prefix loading regression and pipeline save_pretrained failure are different code paths." + }, + { + "left": "issue:44625", + "right": "issue:46077", + "accept": false, + "reason": "Qwen3.5 num_labels propagation and Gemma4 attention-mask construction are unrelated." + }, + { + "left": "issue:43976", + "right": "issue:44938", + "accept": false, + "reason": "Both are Python-version compatibility reports, but for different versions and likely different causes." + }, + { + "left": "issue:46077", + "right": "issue:46082", + "accept": false, + "reason": "Gemma4 attention-mask bug vs LlamaConfig head_dim validation are unrelated." + }, + { + "left": "issue:43352", + "right": "issue:45593", + "accept": false, + "reason": "Unsupported Flash Attention 2.0 and D-FINE auxiliary-loss behavior are unrelated." + }, + { + "left": "issue:36683", + "right": "issue:43525", + "accept": false, + "reason": "Different model configs and different missing attributes." + }, + { + "left": "issue:43618", + "right": "issue:45216", + "accept": false, + "reason": "CLIP output regression and Qwen3.5 save_pretrained regression 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": 1, + "component_count": 17, + "review_unit_index": 46, + "review_unit_count": 49, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-426", + "nodes": [ + "issue:28282", + "issue:33453", + "issue:36296", + "issue:36683", + "issue:41669", + "issue:42490", + "issue:42890", + "issue:43010", + "issue:43064", + "issue:43352", + "issue:43502", + "issue:43575", + "issue:43576", + "issue:43618", + "issue:43723", + "issue:43867", + "issue:43931", + "issue:43976", + "issue:43992", + "issue:44062", + "issue:44075", + "issue:44077", + "issue:44112", + "issue:44242", + "issue:44246", + "issue:44355", + "issue:44360", + "issue:44387", + "issue:44484", + "issue:44492", + "issue:44556", + "issue:44855", + "issue:44857", + "issue:44938", + "issue:44964", + "issue:44977", + "issue:45071", + "issue:45278", + "issue:45310", + "issue:45357", + "issue:45375", + "issue:45381", + "issue:45440", + "issue:45478", + "issue:45561", + "issue:45600", + "issue:45850", + "issue:46132" + ], + "soft_pairs": [ + "issue:36683|issue:44242", + "issue:43352|issue:43931", + "issue:43064|issue:44492", + "issue:43010|issue:45278", + "issue:43992|issue:45310", + "issue:44355|issue:44855", + "issue:43867|issue:45375", + "issue:41669|issue:44246", + "issue:42890|issue:45561", + "issue:36683|issue:43064", + "issue:44938|issue:45071", + "issue:43867|issue:45357", + "issue:43010|issue:45850", + "issue:44938|issue:45381", + "issue:43010|issue:43976", + "issue:42490|issue:44075", + "issue:43992|issue:45478", + "issue:43618|issue:44077", + "issue:36296|issue:43575", + "issue:28282|issue:33453", + "issue:43867|issue:44484", + "issue:43618|issue:44857", + "issue:44360|issue:45440", + "issue:43618|issue:44977", + "issue:42890|issue:44964", + "issue:42890|issue:44112", + "issue:42890|issue:45600", + "issue:43502|issue:44062", + "issue:43723|issue:44556", + "issue:43576|issue:46132", + "issue:44387|issue:44556" + ], + "prepared_review_unit_hash": "75aafdbfe773c959501ea31618bbac392a786b912b8de88ba12744b0ed89c37d", + "component_budget": { + "node_count": 426, + "item_count": 426, + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 31, + "serialized_chars": 36499, + "estimated_input_tokens": 9125, + "estimated_eval_tokens": 18506 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "The cluster is still best treated as a loose collection of Transformers v5 regressions, model-loading bugs, and CI/test issues. The broad v5 import/loading report is the best anchor, but the soft edges are mostly false positives; in particular the UMT5Encoder loading bug should not be merged with the Qwen3.5 MoE from_pretrained reports.", + "confidence": 0.71, + "canonical_issue_reason": "issue:45278 is the best broad representative of the recurring v5 upgrade/import-regression theme, and it is concrete enough to anchor the cluster without collapsing distinct model-specific failures.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45278 is the strongest single issue to represent the cluster because it covers the dominant upgrade/load regression pattern seen across several items, while remaining more specific than the generic complaints.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:36683", + "right": "issue:44242", + "accept": false, + "reason": "Different bugs: Gemma3Config vocab_size access vs missing load-balancing loss when router logits are disabled." + }, + { + "left": "issue:43352", + "right": "issue:43931", + "accept": false, + "reason": "Unrelated failure modes: Flash Attention unsupported error vs Qwen3-VL weight-shape mismatch." + }, + { + "left": "issue:43064", + "right": "issue:44492", + "accept": false, + "reason": "Different subsystems and symptoms: distributed optimizer-state corruption vs a cache-strategy typo." + }, + { + "left": "issue:43010", + "right": "issue:45278", + "accept": false, + "reason": "A no_grad decoration request for cache/layer updates is not the same concrete bug as a broad v5 import/load regression." + }, + { + "left": "issue:43992", + "right": "issue:45310", + "accept": false, + "reason": "Not the same bug: UMT5Encoder missing embed_tokens.weight is a different model/path from Qwen3.5 MoE from_pretrained errors." + }, + { + "left": "issue:44355", + "right": "issue:44855", + "accept": false, + "reason": "Compiled-Python import failures and a Python 3.13 torch.jit parsing issue are different code paths." + }, + { + "left": "issue:43867", + "right": "issue:45375", + "accept": false, + "reason": "State_dict sorting load error and missing deepstack_visual_indexes in config are different model-loading problems." + }, + { + "left": "issue:41669", + "right": "issue:44246", + "accept": false, + "reason": "Both mention import performance, but one is a specific import-star cleanup proposal and the other is an intermittent slowdown report." + }, + { + "left": "issue:42890", + "right": "issue:45561", + "accept": false, + "reason": "Test flakiness from missing set_seed is unrelated to the xdist captured_info.txt race." + }, + { + "left": "issue:36683", + "right": "issue:43064", + "accept": false, + "reason": "Config attribute error and distributed optimizer-state corruption are unrelated failures." + }, + { + "left": "issue:44938", + "right": "issue:45071", + "accept": false, + "reason": "Python-version load failure and PretrainedConfig type-checking breakage are different regressions." + }, + { + "left": "issue:43867", + "right": "issue:45357", + "accept": false, + "reason": "Different save/load bugs: sorted state_dict handling vs incorrect visual encoder keys on save_pretrained." + }, + { + "left": "issue:43010", + "right": "issue:45850", + "accept": false, + "reason": "A cache update decoration issue is not the same bug as a broad 'v5 breaks too many models' complaint." + }, + { + "left": "issue:44938", + "right": "issue:45381", + "accept": false, + "reason": "Generic Python 3.14 import failure is unrelated to Qwen2.5-VL video position-id behavior." + }, + { + "left": "issue:43010", + "right": "issue:43976", + "accept": false, + "reason": "Cache update decoration and Python version compatibility are unrelated issues." + }, + { + "left": "issue:42490", + "right": "issue:44075", + "accept": false, + "reason": "Local-files-only path handling is unrelated to SGD argument handling." + }, + { + "left": "issue:43992", + "right": "issue:45478", + "accept": false, + "reason": "Not the same bug: UMT5Encoder embed_tokens.weight omission is a different failure than Qwen3.5 MoE from_pretrained errors." + }, + { + "left": "issue:43618", + "right": "issue:44077", + "accept": false, + "reason": "CLIP attentions regression and patchtsmixer post_init eligibility are different model-specific bugs." + }, + { + "left": "issue:36296", + "right": "issue:43575", + "accept": false, + "reason": "Tensor-parallel training bug and tp-related OOM on model load are different failures." + }, + { + "left": "issue:28282", + "right": "issue:33453", + "accept": false, + "reason": "AutoModel missing PyTorch and tokenizer loading regression are unrelated issues." + }, + { + "left": "issue:43867", + "right": "issue:44484", + "accept": false, + "reason": "Sorted state_dict load error is unrelated to max_shard_size default behavior." + }, + { + "left": "issue:43618", + "right": "issue:44857", + "accept": false, + "reason": "Attention-output regression and float16 AMP crash are different bugs." + }, + { + "left": "issue:44360", + "right": "issue:45440", + "accept": false, + "reason": "DSA indexer ReLU discussion is unrelated to DeepSeekV3MoE implementation divergence." + }, + { + "left": "issue:43618", + "right": "issue:44977", + "accept": false, + "reason": "CLIP attentions regression does not match Qwen3.5 flash-attention generation failure." + }, + { + "left": "issue:42890", + "right": "issue:44964", + "accept": false, + "reason": "SamHQ test seeding flakiness is unrelated to Phi-4 multimodal model-loading failure." + }, + { + "left": "issue:42890", + "right": "issue:44112", + "accept": false, + "reason": "CI flakiness in a model test and a stale device-override test are different problems." + }, + { + "left": "issue:42890", + "right": "issue:45600", + "accept": false, + "reason": "Captured-info race and removed-config references are distinct CI issues." + }, + { + "left": "issue:43502", + "right": "issue:44062", + "accept": false, + "reason": "Local-files-only API requests and AddedToken argument duplication are unrelated." + }, + { + "left": "issue:43723", + "right": "issue:44556", + "accept": false, + "reason": "Tokenizer loading and checkpoint reload after upgrade are both load-related, but the packet does not show the same concrete bug." + }, + { + "left": "issue:43576", + "right": "issue:46132", + "accept": false, + "reason": "Broken env CLI behavior and AttentionInterface.register behavior change are different features." + }, + { + "left": "issue:44387", + "right": "issue:44556", + "accept": false, + "reason": "Int4 CUDA memory regression is unrelated to checkpoint reload failures after upgrade." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "Reject: the accepted soft-edge merges are not grounded in the packet. issue:43992 is \"UMT5Encoder.from_pretrained misses embed_tokens.weight,\" while issue:45310 and issue:45478 are Qwen3.5 MoE from_pretrained errors; the packet does not support treating 43992 as the same bug as either of those. The rest of the verdicts are generally conservative, and the cluster summary is plausible, but these accepted pairings overstate evidence." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 47, + "review_unit_count": 49, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-426", + "nodes": [ + "issue:30333", + "issue:30990", + "issue:31795", + "issue:33453", + "issue:34689", + "issue:36296", + "issue:38617", + "issue:41628", + "issue:42548", + "issue:42757", + "issue:42831", + "issue:42890", + "issue:43010", + "issue:43065", + "issue:43232", + "issue:43299", + "issue:43502", + "issue:43526", + "issue:43576", + "issue:43673", + "issue:43824", + "issue:43827", + "issue:43874", + "issue:43901", + "issue:44222", + "issue:44230", + "issue:44315", + "issue:44351", + "issue:44355", + "issue:44360", + "issue:44485", + "issue:44492", + "issue:44556", + "issue:44655", + "issue:44749", + "issue:44908", + "issue:44961", + "issue:45071", + "issue:45125", + "issue:45341", + "issue:45362", + "issue:45375", + "issue:45468", + "issue:45478", + "issue:45850", + "issue:46143", + "issue:46144" + ], + "soft_pairs": [ + "issue:42890|issue:45341", + "issue:42548|issue:44351", + "issue:44222|issue:45125", + "issue:43010|issue:43065", + "issue:44355|issue:44749", + "issue:43502|issue:44485", + "issue:45125|issue:45478", + "issue:30333|issue:33453", + "issue:43010|issue:43232", + "issue:45125|issue:45375", + "issue:30990|issue:31795", + "issue:43502|issue:44222", + "issue:43299|issue:44556", + "issue:42831|issue:43874", + "issue:44315|issue:44360", + "issue:43502|issue:44961", + "issue:30333|issue:36296", + "issue:44355|issue:45071", + "issue:43827|issue:44961", + "issue:44355|issue:44655", + "issue:33453|issue:34689", + "issue:42757|issue:43576", + "issue:44222|issue:44230", + "issue:43576|issue:45071", + "issue:33453|issue:36296", + "issue:43576|issue:43824", + "issue:41628|issue:42548", + "issue:31795|issue:38617", + "issue:44961|issue:45468", + "issue:38617|issue:43576", + "issue:46143|issue:46144", + "issue:44908|issue:44961", + "issue:44485|issue:44961", + "issue:43673|issue:44492", + "issue:36296|issue:43901", + "issue:41628|issue:44355", + "issue:36296|issue:43526", + "issue:45362|issue:45850" + ], + "prepared_review_unit_hash": "d77ad5e89078983e727462da05c7e855d10ef4e411d1999ca4abc0412ea6571b", + "component_budget": { + "node_count": 426, + "item_count": 426, + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 + }, + "budget": { + "node_count": 47, + "item_count": 47, + "soft_pair_count": 38, + "serialized_chars": 37818, + "estimated_input_tokens": 9455, + "estimated_eval_tokens": 19166 + }, + "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 regressions, import errors, model-loading issues, and docs/test bugs. I do not see any soft pair that is convincingly the same underlying defect, so I would not merge any of these issues as duplicates.", + "confidence": 0.93, + "canonical_issue_reason": "If a single anchor is needed, #45850 is the broadest umbrella issue for the v5 breakage theme, but the cluster is too heterogeneous for a true duplicate canonical.", + "canonical_pr_reason": null, + "best_issue_reason": "#45850 is the most general and central issue in the set, so it is the best umbrella candidate even though it is not a strong duplicate match for the rest.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:42890", + "right": "issue:45341", + "accept": false, + "reason": "Different bugs: seed instability in a SAM HQ test vs a utility bug in testing_utils.py." + }, + { + "left": "issue:42548", + "right": "issue:44351", + "accept": false, + "reason": "Different missing symbols/import paths: PreTrainedModel vs HybridCache." + }, + { + "left": "issue:44222", + "right": "issue:45125", + "accept": false, + "reason": "Different FP8/serialization bug vs tensor-parallel `_tp_plan` loading bug." + }, + { + "left": "issue:43010", + "right": "issue:43065", + "accept": false, + "reason": "Cache `update()` no_grad concern is unrelated to the dummy Conv2d in Sam3PixelDecoder." + }, + { + "left": "issue:44355", + "right": "issue:44749", + "accept": false, + "reason": "Compiled Python file errors and a data-filtering slowdown are unrelated." + }, + { + "left": "issue:43502", + "right": "issue:44485", + "accept": false, + "reason": "Offline/local-files-only request leakage is unrelated to GLM-5 RoPE implementation." + }, + { + "left": "issue:45125", + "right": "issue:45478", + "accept": false, + "reason": "Both mention Qwen3.5 MoE, but one is tensor-parallel `_tp_plan` and the other is a broader from_pretrained failure; not clearly the same defect." + }, + { + "left": "issue:30333", + "right": "issue:33453", + "accept": false, + "reason": "MLFlow failure reporting and tokenizer loading regression are different code paths." + }, + { + "left": "issue:43010", + "right": "issue:43232", + "accept": false, + "reason": "`no_grad` decoration for cache update is not the same as the generation/sync_gpus kwargs issue." + }, + { + "left": "issue:45125", + "right": "issue:45375", + "accept": false, + "reason": "Tensor-parallel plan bug and missing vision config field are different model/config issues." + }, + { + "left": "issue:30990", + "right": "issue:31795", + "accept": false, + "reason": "Loading hang in Sentence Transformers is unrelated to a documentation wording issue." + }, + { + "left": "issue:43502", + "right": "issue:44222", + "accept": false, + "reason": "Local-files-only network leakage is unrelated to FP8 save_pretrained on MoE models." + }, + { + "left": "issue:43299", + "right": "issue:44556", + "accept": false, + "reason": "Both are loading-related, but they concern different versions/models and different failure modes." + }, + { + "left": "issue:42831", + "right": "issue:43874", + "accept": false, + "reason": "FP8 accuracy drift and a missing image-patch helper are separate problems." + }, + { + "left": "issue:44315", + "right": "issue:44360", + "accept": false, + "reason": "Liger Kernel/model_init behavior is unrelated to the DSA ReLU discussion." + }, + { + "left": "issue:43502", + "right": "issue:44961", + "accept": false, + "reason": "Local-files-only API leakage and a one-word issue are unrelated." + }, + { + "left": "issue:30333", + "right": "issue:36296", + "accept": false, + "reason": "MLFlow reporting on failed jobs is unrelated to tensor-parallel training." + }, + { + "left": "issue:44355", + "right": "issue:45071", + "accept": false, + "reason": "Compiled Python file execution errors and PretrainedConfig type checking are unrelated." + }, + { + "left": "issue:43827", + "right": "issue:44961", + "accept": false, + "reason": "Docs still referencing pipeline() and a placeholder issue are unrelated." + }, + { + "left": "issue:44355", + "right": "issue:44655", + "accept": false, + "reason": "Compiled-file runtime errors and inability to save Pipeline objects are different issues." + }, + { + "left": "issue:33453", + "right": "issue:34689", + "accept": false, + "reason": "Tokenizer-loading regression and Llama 3.2 vision model loading breakage are distinct." + }, + { + "left": "issue:42757", + "right": "issue:43576", + "accept": false, + "reason": "Importing `is_offline_mode` and the `transformers env` CLI failure are related at most by version churn, not the same concrete bug." + }, + { + "left": "issue:44222", + "right": "issue:44230", + "accept": false, + "reason": "FP8 MoE save_pretrained and FP8 inference support are different requests/failures." + }, + { + "left": "issue:43576", + "right": "issue:45071", + "accept": false, + "reason": "Broken CLI env command and PretrainedConfig type checking are unrelated." + }, + { + "left": "issue:43824", + "right": "issue:44961", + "accept": false, + "reason": "Qwen2.5-VL import failure and the placeholder issue are unrelated." + }, + { + "left": "issue:41628", + "right": "issue:42548", + "accept": false, + "reason": "Different missing imports (`AutoImageProcessor` vs `PreTrainedModel`)." + }, + { + "left": "issue:31795", + "right": "issue:38617", + "accept": false, + "reason": "Docs confusion and an import error for `layer_type_validation` are not the same bug." + }, + { + "left": "issue:44961", + "right": "issue:45468", + "accept": false, + "reason": "Placeholder issue and Gemma-4 audio positional encoding bug are unrelated." + }, + { + "left": "issue:38617", + "right": "issue:43576", + "accept": false, + "reason": "Import error for `layer_type_validation` is not the same as the broken env command." + }, + { + "left": "issue:46143", + "right": "issue:46144", + "accept": false, + "reason": "Both are RoFormer-related, but one is kwargs forwarding and the other is attention-interface usage; not the same concrete defect." + }, + { + "left": "issue:44908", + "right": "issue:44961", + "accept": false, + "reason": "Scheduler kwargs handling and a placeholder issue are unrelated." + }, + { + "left": "issue:44485", + "right": "issue:44961", + "accept": false, + "reason": "GLM-5 RoPE implementation and a placeholder issue are unrelated." + }, + { + "left": "issue:43673", + "right": "issue:44492", + "accept": false, + "reason": "GenerationMixin cache-missing behavior and a cache-strategy typo are different issues." + }, + { + "left": "issue:36296", + "right": "issue:43901", + "accept": false, + "reason": "Tensor-parallel training bug and TextClassificationPipeline docs mismatch are unrelated." + }, + { + "left": "issue:41628", + "right": "issue:44355", + "accept": false, + "reason": "Importing `AutoImageProcessor` and compiled-file execution errors are unrelated." + }, + { + "left": "issue:36296", + "right": "issue:43526", + "accept": false, + "reason": "Tensor-parallel training and BeitImageProcessorFast label reduction are different code paths." + }, + { + "left": "issue:45362", + "right": "issue:45850", + "accept": false, + "reason": "Qwen3.5 chat crash is a specific model bug; #45850 is a broad meta issue about many v5 breakages, 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": 1, + "component_count": 17, + "review_unit_index": 48, + "review_unit_count": 49, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-426", + "nodes": [ + "issue:31795", + "issue:33453", + "issue:41669", + "issue:42371", + "issue:42890", + "issue:43352", + "issue:43408", + "issue:43525", + "issue:43572", + "issue:43576", + "issue:43673", + "issue:43827", + "issue:43874", + "issue:43976", + "issue:44038", + "issue:44077", + "issue:44230", + "issue:44246", + "issue:44273", + "issue:44295", + "issue:44297", + "issue:44315", + "issue:44351", + "issue:44360", + "issue:44393", + "issue:44492", + "issue:44556", + "issue:44623", + "issue:44655", + "issue:44704", + "issue:44829", + "issue:44861", + "issue:44908", + "issue:44945", + "issue:45003", + "issue:45071", + "issue:45092", + "issue:45125", + "issue:45230", + "issue:45278", + "issue:45310", + "issue:45341", + "issue:45468", + "issue:45478", + "issue:45542", + "issue:45753", + "issue:45850", + "issue:46095" + ], + "soft_pairs": [ + "issue:44623|issue:45230", + "issue:44273|issue:45542", + "issue:44273|issue:45003", + "issue:44038|issue:45125", + "issue:44393|issue:45071", + "issue:44230|issue:46095", + "issue:44393|issue:45753", + "issue:43576|issue:44556", + "issue:43525|issue:45850", + "issue:41669|issue:43673", + "issue:44556|issue:45092", + "issue:43572|issue:45850", + "issue:44492|issue:45850", + "issue:42371|issue:44492", + "issue:45278|issue:45850", + "issue:44246|issue:44273", + "issue:42890|issue:44908", + "issue:44704|issue:44908", + "issue:43408|issue:44315", + "issue:43976|issue:44393", + "issue:43352|issue:44393", + "issue:44297|issue:45230", + "issue:43576|issue:44908", + "issue:44295|issue:45230", + "issue:44908|issue:45071", + "issue:41669|issue:44351", + "issue:41669|issue:45341", + "issue:44492|issue:44945", + "issue:43874|issue:45850", + "issue:44393|issue:45478", + "issue:44077|issue:44861", + "issue:43827|issue:44908", + "issue:41669|issue:44908", + "issue:31795|issue:33453", + "issue:44315|issue:44829", + "issue:43576|issue:44273", + "issue:44360|issue:45468", + "issue:44655|issue:44861", + "issue:45230|issue:45310" + ], + "prepared_review_unit_hash": "a5464ac8b45fa8ec5eacbec5038d45724344def9fcabc9fba2ca76df419c82ce", + "component_budget": { + "node_count": 426, + "item_count": 426, + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 39, + "serialized_chars": 38845, + "estimated_input_tokens": 9712, + "estimated_eval_tokens": 19680 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is mostly a grab bag of unrelated closed issues, with only a loose common theme of Transformers v5 regressions and loading/import breakages. I found no pair that clearly describes the exact same underlying bug/change.", + "confidence": 0.77, + "canonical_issue_reason": "issue:45278 is the best umbrella representative: it is a broad, concrete regression report about upgrading from 4.57.0 to 5.5.0 and getting many import errors, which matches the cluster\u2019s overall v5 breakage theme better than the more generic or highly specific issues.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:45278 has the clearest high-level scope and is the most representative of the repeated upgrade/regression complaints in this set.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44623", + "right": "issue:45230", + "accept": false, + "reason": "44623 is a specific processor.save_pretrained missing-files bug; 45230 is just a generic bug report with no evidence it targets the same failure." + }, + { + "left": "issue:44273", + "right": "issue:45542", + "accept": false, + "reason": "One is a lazy-loading correctness problem; the other is a TensorFlow/tensorboard backend error. Different code paths and symptoms." + }, + { + "left": "issue:44273", + "right": "issue:45003", + "accept": false, + "reason": "Both touch import/loading behavior, but 44273 is about lazy loading not working while 45003 is about unsafe sys.modules access; not the same concrete bug." + }, + { + "left": "issue:44038", + "right": "issue:45125", + "accept": false, + "reason": "Same general model family, but one is a broad Qwen3-VL-Moe bug and the other is a missing tensor-parallel plan attribute. Different fixes." + }, + { + "left": "issue:44393", + "right": "issue:45071", + "accept": false, + "reason": "Qwen3-VL bbox/output issues are unrelated to a PretrainedConfig type-checking regression." + }, + { + "left": "issue:44230", + "right": "issue:46095", + "accept": false, + "reason": "FP8 inference support for Qwen models is unrelated to a DeepSeekV4 weight-conversion question." + }, + { + "left": "issue:44393", + "right": "issue:45753", + "accept": false, + "reason": "Both are Qwen3-related, but one is bbox output corruption and the other is an infinite loop on a specific image; not the same underlying bug." + }, + { + "left": "issue:43576", + "right": "issue:44556", + "accept": false, + "reason": "Both are v5 regressions, but one breaks the transformers env command and the other breaks checkpoint reload after upgrade. Different failures." + }, + { + "left": "issue:43525", + "right": "issue:45850", + "accept": false, + "reason": "A missing pad_token_id on Llama4Config is a specific config bug, not the same as the broad 'too many models break in v5' complaint." + }, + { + "left": "issue:41669", + "right": "issue:43673", + "accept": false, + "reason": "Import-star slowdowns and a missing GenerationMixin cache in chunked_prefill are unrelated issues." + }, + { + "left": "issue:44556", + "right": "issue:45092", + "accept": false, + "reason": "Both involve v5 compatibility, but one is generic checkpoint reload failure and the other is remote-code/meta-init incompatibility for InternVL2. Not the same bug." + }, + { + "left": "issue:43572", + "right": "issue:45850", + "accept": false, + "reason": "StableLmConfig pad_token_idx loss is a model-specific config regression, not the same as a broad v5 breakage report." + }, + { + "left": "issue:44492", + "right": "issue:45850", + "accept": false, + "reason": "A typo in cache strategies is too specific and unrelated to the broad model-breakage complaint." + }, + { + "left": "issue:45278", + "right": "issue:45850", + "accept": false, + "reason": "Both are broad upgrade complaints, but one is specifically about import errors and the other about too many models breaking; similar theme, not the same underlying issue." + }, + { + "left": "issue:44246", + "right": "issue:44273", + "accept": false, + "reason": "Import slowness and lazy loading malfunction are related only at a high level; they are not the same concrete defect." + }, + { + "left": "issue:42890", + "right": "issue:44908", + "accept": false, + "reason": "Missing set_seed in tests and ignored lr_scheduler_kwargs are unrelated." + }, + { + "left": "issue:44704", + "right": "issue:44908", + "accept": false, + "reason": "AutoProcessor kwargs propagation and scheduler kwargs handling are different API surfaces and bugs." + }, + { + "left": "issue:43408", + "right": "issue:44315", + "accept": false, + "reason": "A model-type warning for sam3_* and Liger Kernel not applied with model_init are unrelated." + }, + { + "left": "issue:43976", + "right": "issue:44393", + "accept": false, + "reason": "Python version compatibility and Qwen3-VL bbox output errors are unrelated." + }, + { + "left": "issue:43352", + "right": "issue:44393", + "accept": false, + "reason": "Flash Attention 2 support in NemotronH vs Qwen3-VL output issues are different code paths." + }, + { + "left": "issue:44297", + "right": "issue:45230", + "accept": false, + "reason": "Tokenizer save_pretrained mismatch is a concrete serialization bug; 45230 is too generic to match." + }, + { + "left": "issue:43576", + "right": "issue:44908", + "accept": false, + "reason": "transformers env CLI breakage and inverse_sqrt scheduler kwargs handling are unrelated." + }, + { + "left": "issue:44295", + "right": "issue:45230", + "accept": false, + "reason": "Position_ids buffer access failure does not match the generic bug report." + }, + { + "left": "issue:44908", + "right": "issue:45071", + "accept": false, + "reason": "Scheduler kwargs handling and PretrainedConfig type checking are unrelated." + }, + { + "left": "issue:41669", + "right": "issue:44351", + "accept": false, + "reason": "Import-star performance regression and missing HybridCache export are different import-related bugs." + }, + { + "left": "issue:41669", + "right": "issue:45341", + "accept": false, + "reason": "A slow import regression and a small testing_utils bug are not the same issue." + }, + { + "left": "issue:44492", + "right": "issue:44945", + "accept": false, + "reason": "Cache strategy typo and pipeline-parallelism output errors are unrelated." + }, + { + "left": "issue:43874", + "right": "issue:45850", + "accept": false, + "reason": "A missing image-patch helper in one processor is too specific to be the same as a broad v5 breakage report." + }, + { + "left": "issue:44393", + "right": "issue:45478", + "accept": false, + "reason": "Qwen3-VL bounding-box hallucination/error and Qwen3.5-Moe from_pretrained failure are different bugs." + }, + { + "left": "issue:44077", + "right": "issue:44861", + "accept": false, + "reason": "Optional post_init policy and tied-weight-key AttributeError are separate model-loading issues." + }, + { + "left": "issue:43827", + "right": "issue:44908", + "accept": false, + "reason": "Docs still referencing pipeline() and scheduler kwargs handling are unrelated." + }, + { + "left": "issue:41669", + "right": "issue:44908", + "accept": false, + "reason": "Import-star slowdown and scheduler kwargs handling do not describe the same bug." + }, + { + "left": "issue:31795", + "right": "issue:33453", + "accept": false, + "reason": "Documentation confusion around forward args and tokenizer loading regression are unrelated." + }, + { + "left": "issue:44315", + "right": "issue:44829", + "accept": false, + "reason": "Liger Kernel not applied with model_init and degenerate training under flash_attention_3 are different concrete problems." + }, + { + "left": "issue:43576", + "right": "issue:44273", + "accept": false, + "reason": "A broken env command and lazy-loading malfunction are different symptoms and likely different fixes." + }, + { + "left": "issue:44360", + "right": "issue:45468", + "accept": false, + "reason": "DSA indexer discussion and Gemma-4 audio positional encoding are unrelated." + }, + { + "left": "issue:44655", + "right": "issue:44861", + "accept": false, + "reason": "Pipeline save_pretrained failure and tied-weight-key AttributeError are different serialization/model-loading bugs." + }, + { + "left": "issue:45230", + "right": "issue:45310", + "accept": false, + "reason": "Generic bug report versus a specific Qwen3.5-Moe from_pretrained regression; no evidence they are the same issue." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 1, + "component_count": 17, + "review_unit_index": 49, + "review_unit_count": 49, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-10000-426", + "nodes": [ + "issue:30333", + "issue:31515", + "issue:31795", + "issue:33453", + "issue:34689", + "issue:36296", + "issue:37428", + "issue:37934", + "issue:38617", + "issue:41669", + "issue:42371", + "issue:42913", + "issue:42936", + "issue:43066", + "issue:43240", + "issue:43298", + "issue:43502", + "issue:43519", + "issue:43525", + "issue:43572", + "issue:43606", + "issue:43618", + "issue:43668", + "issue:43704", + "issue:43717", + "issue:43813", + "issue:44031", + "issue:44230", + "issue:44261", + "issue:44263", + "issue:44273", + "issue:44315", + "issue:44485", + "issue:44492", + "issue:44623", + "issue:44861", + "issue:44908", + "issue:44998", + "issue:45092", + "issue:45290", + "issue:45310", + "issue:45341", + "issue:45468", + "issue:45850", + "issue:46018", + "issue:46095", + "issue:46144", + "issue:46172" + ], + "soft_pairs": [ + "issue:46018|issue:46172", + "issue:44273|issue:44908", + "issue:43519|issue:44485", + "issue:43519|issue:44908", + "issue:31795|issue:34689", + "issue:41669|issue:43704", + "issue:31795|issue:37428", + "issue:44315|issue:45092", + "issue:44485|issue:45468", + "issue:43704|issue:44485", + "issue:43298|issue:46095", + "issue:43502|issue:43519", + "issue:42371|issue:43704", + "issue:43519|issue:44623", + "issue:43519|issue:43704", + "issue:43606|issue:45468", + "issue:44263|issue:45468", + "issue:44261|issue:45468", + "issue:42913|issue:43066", + "issue:44908|issue:44998", + "issue:44492|issue:44998", + "issue:44861|issue:45290", + "issue:43618|issue:44998", + "issue:44998|issue:45850", + "issue:44998|issue:45341", + "issue:43525|issue:44861", + "issue:43572|issue:44861", + "issue:31515|issue:38617", + "issue:44230|issue:44623", + "issue:30333|issue:31515", + "issue:31515|issue:36296", + "issue:31515|issue:31795", + "issue:44230|issue:45310", + "issue:31515|issue:33453", + "issue:43298|issue:44230", + "issue:43717|issue:46144", + "issue:42936|issue:46018", + "issue:43668|issue:44031", + "issue:42936|issue:43240", + "issue:43240|issue:46144", + "issue:42936|issue:44031", + "issue:42936|issue:46144", + "issue:42936|issue:43813", + "issue:43240|issue:43668", + "issue:43668|issue:43813", + "issue:37934|issue:43668", + "issue:42936|issue:43668" + ], + "prepared_review_unit_hash": "e851ffd5df1e8179b16c83ac8695695fc89847d6112ada97b8b9e8f892c2d876", + "component_budget": { + "node_count": 426, + "item_count": 426, + "soft_pair_count": 1571, + "serialized_chars": 813804, + "estimated_input_tokens": 203451, + "estimated_eval_tokens": 407158 + }, + "budget": { + "node_count": 48, + "item_count": 48, + "soft_pair_count": 47, + "serialized_chars": 40969, + "estimated_input_tokens": 10243, + "estimated_eval_tokens": 20742 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This is a noisy similarity cluster, mostly mixing assorted Transformers v5/model-loading regressions with unrelated tokenizer, scheduler, vision, and docs issues. The broadest shared theme is version-5 compatibility breakage, but none of the soft pairs are true duplicates or mergeable PR-equivalents.", + "confidence": 0.66, + "canonical_issue_reason": "Issue 45850 is the best umbrella issue because it is the only broadly scoped report capturing the shared theme of Transformers 5.0 breaking many models, while the others are narrower, model-specific, or unrelated subsystem bugs.", + "canonical_pr_reason": null, + "best_issue_reason": "45850 is the most representative cluster anchor: it generalizes the common v5 regression pattern better than the specific loading/config/tokenizer bugs elsewhere in the set.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:46018", + "right": "issue:46172", + "accept": false, + "reason": "Different DeepSeek-v4 symptoms: shared expert gating vs a tensor transpose bug." + }, + { + "left": "issue:44273", + "right": "issue:44908", + "accept": false, + "reason": "Lazy-loading behavior and inverse_sqrt scheduler kwargs are unrelated." + }, + { + "left": "issue:43519", + "right": "issue:44485", + "accept": false, + "reason": "Qwen3VL timestamp parsing and GLM-5 RoPE implementation are different bugs." + }, + { + "left": "issue:43519", + "right": "issue:44908", + "accept": false, + "reason": "Processor timestamp math is unrelated to scheduler kwargs handling." + }, + { + "left": "issue:31795", + "right": "issue:34689", + "accept": false, + "reason": "Documentation confusion and Llama 3.2 model-loading regression are not the same issue." + }, + { + "left": "issue:41669", + "right": "issue:43704", + "accept": false, + "reason": "Import-star performance cleanup is unrelated to a VRAM leak in multi-threaded inference." + }, + { + "left": "issue:31795", + "right": "issue:37428", + "accept": false, + "reason": "Docs wording for forward args and a missing internal flash-attn import are different problems." + }, + { + "left": "issue:44315", + "right": "issue:45092", + "accept": false, + "reason": "Liger Kernel application with model_init and InternVL2 meta-initialization incompatibility are separate integration bugs." + }, + { + "left": "issue:44485", + "right": "issue:45468", + "accept": false, + "reason": "GLM-5 RoPE and Gemma-4 audio positional encoding affect different models and code paths." + }, + { + "left": "issue:43704", + "right": "issue:44485", + "accept": false, + "reason": "VRAM leak in Qwen3ForCausalLM is unrelated to GLM-5 RoPE math." + }, + { + "left": "issue:43298", + "right": "issue:46095", + "accept": false, + "reason": "Continuous batching for audio models and a DeepSeek-V4 weight-conversion question are not duplicates." + }, + { + "left": "issue:43502", + "right": "issue:43519", + "accept": false, + "reason": "local_files_only network leakage and timestamp calculation in a processor are different failures." + }, + { + "left": "issue:42371", + "right": "issue:43704", + "accept": false, + "reason": "TF32 behavior settings and a Qwen3 memory leak do not match." + }, + { + "left": "issue:43519", + "right": "issue:44623", + "accept": false, + "reason": "Timestamp calculation and processor.save_pretrained missing files are different processor bugs." + }, + { + "left": "issue:43519", + "right": "issue:43704", + "accept": false, + "reason": "Processor timestamp math and VRAM leakage are unrelated." + }, + { + "left": "issue:43606", + "right": "issue:45468", + "accept": false, + "reason": "CPU-offload device mismatch and Gemma-4 audio positional encoding are different model paths." + }, + { + "left": "issue:44263", + "right": "issue:45468", + "accept": false, + "reason": "torch.split return handling and Gemma-4 audio positional encoding are unrelated." + }, + { + "left": "issue:44261", + "right": "issue:45468", + "accept": false, + "reason": "MLA RMS-eps precision discussion is not the same as Gemma-4 audio encoding." + }, + { + "left": "issue:42913", + "right": "issue:43066", + "accept": false, + "reason": "Tokenizer v4/v5 behavior differences and wrong decoder type are both tokenizer-related but distinct bugs with different fixes." + }, + { + "left": "issue:44908", + "right": "issue:44998", + "accept": false, + "reason": "Scheduler kwargs handling and an unrelated placeholder issue are not duplicates." + }, + { + "left": "issue:44492", + "right": "issue:44998", + "accept": false, + "reason": "Cache-strategy typo and an unrelated placeholder issue are not the same bug." + }, + { + "left": "issue:44861", + "right": "issue:45290", + "accept": false, + "reason": "Tied-weights helper crash and chat template tool-call crash affect different code paths." + }, + { + "left": "issue:43618", + "right": "issue:44998", + "accept": false, + "reason": "CLIPOutput attentions regression and an unrelated placeholder issue are not duplicates." + }, + { + "left": "issue:44998", + "right": "issue:45850", + "accept": false, + "reason": "Placeholder issue text is not a real report about Transformers 5.0 breakages." + }, + { + "left": "issue:44998", + "right": "issue:45341", + "accept": false, + "reason": "Placeholder issue and a testing_utils bug are unrelated." + }, + { + "left": "issue:43525", + "right": "issue:44861", + "accept": false, + "reason": "Missing pad_token_id in Llama4Config and tied-weight-key crashes are different configuration/model bugs." + }, + { + "left": "issue:43572", + "right": "issue:44861", + "accept": false, + "reason": "StableLmConfig pad_token_idx regression and tied-weight-key crashes are unrelated." + }, + { + "left": "issue:31515", + "right": "issue:38617", + "accept": false, + "reason": "Slow checkpoint loading and a missing configuration_utils import are not the same bug." + }, + { + "left": "issue:44230", + "right": "issue:44623", + "accept": false, + "reason": "FP8 support request and missing processor save files are unrelated." + }, + { + "left": "issue:30333", + "right": "issue:31515", + "accept": false, + "reason": "MLflow job status reporting and slow from_pretrained checkpoint loading are different issues." + }, + { + "left": "issue:31515", + "right": "issue:36296", + "accept": false, + "reason": "Checkpoint load slowness and tensor-parallel training bugs are separate topics." + }, + { + "left": "issue:31515", + "right": "issue:31795", + "accept": false, + "reason": "Model load performance and forward-signature documentation do not describe the same defect." + }, + { + "left": "issue:44230", + "right": "issue:45310", + "accept": false, + "reason": "FP8 inference support and Qwen3.5 MoE from_pretrained error are not the same change." + }, + { + "left": "issue:31515", + "right": "issue:33453", + "accept": false, + "reason": "Slow checkpoint loading and tokenizer-loading regression are different load-path problems." + }, + { + "left": "issue:43298", + "right": "issue:44230", + "accept": false, + "reason": "Audio continuous batching support and FP8 inference support are unrelated feature areas." + }, + { + "left": "issue:43717", + "right": "issue:46144", + "accept": false, + "reason": "Mamba init_weights behavior and RoFormer attention interface usage are unrelated model implementations." + }, + { + "left": "issue:42936", + "right": "issue:46018", + "accept": false, + "reason": "Mask2Former ignore_value handling and DeepSeek-V4 expert gating are different model bugs." + }, + { + "left": "issue:43668", + "right": "issue:44031", + "accept": false, + "reason": "Config type-hint correctness and tokenizer regex warning behavior are unrelated." + }, + { + "left": "issue:42936", + "right": "issue:43240", + "accept": false, + "reason": "Mask2Former ignore_value usage and loss kwargs propagation are different training bugs." + }, + { + "left": "issue:43240", + "right": "issue:46144", + "accept": false, + "reason": "Loss-function kwargs forwarding and RoFormer attention interface usage are unrelated." + }, + { + "left": "issue:42936", + "right": "issue:44031", + "accept": false, + "reason": "Mask2Former ignore_value handling is unrelated to tokenizer regex warnings." + }, + { + "left": "issue:42936", + "right": "issue:46144", + "accept": false, + "reason": "Mask2Former ignore_value and RoFormer attention interface bugs do not share a code path." + }, + { + "left": "issue:42936", + "right": "issue:43813", + "accept": false, + "reason": "Mask2Former ignore_value handling is unrelated to the quantization typo." + }, + { + "left": "issue:43240", + "right": "issue:43668", + "accept": false, + "reason": "Loss kwargs forwarding and config type-hint correction are not the same issue." + }, + { + "left": "issue:43668", + "right": "issue:43813", + "accept": false, + "reason": "Incorrect type hint and quantization typo are different, unrelated fixes." + }, + { + "left": "issue:37934", + "right": "issue:43668", + "accept": false, + "reason": "Llama4 norm question and ModernBERT type-hint issue are unrelated model-specific concerns." + }, + { + "left": "issue:42936", + "right": "issue:43668", + "accept": false, + "reason": "Mask2Former ignore_value bug and ModernBERT type-hint issue 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": 17, + "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-10519-3", + "nodes": [ + "issue:10519", + "issue:11005", + "issue:11307" + ], + "soft_pairs": [ + "issue:10519|issue:11307", + "issue:10519|issue:11005", + "issue:11005|issue:11307" + ], + "prepared_review_unit_hash": "ecc5c441706e5a9bac00c2a9e2e6ed5acb96f40a24c1e5a3ed8adcc419812bc4", + "component_budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 3, + "serialized_chars": 2877, + "estimated_input_tokens": 720, + "estimated_eval_tokens": 1696 + }, + "budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 3, + "serialized_chars": 2877, + "estimated_input_tokens": 720, + "estimated_eval_tokens": 1696 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "All three issues are about different features/topics: token truncation behavior, ReduceLROnPlateau-like training logic, and Wav2Vec2 word time offsets. They are not duplicates of the same underlying bug or change.", + "confidence": 0.99, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": "Issue #10519 is the clearest self-contained feature request, but it is not a duplicate/canonical target for the others.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:10519", + "right": "issue:11307", + "accept": false, + "reason": "Different domains and requests: truncation strategy vs. Wav2Vec2 word boundary offsets. No shared code path or concrete overlap." + }, + { + "left": "issue:10519", + "right": "issue:11005", + "accept": false, + "reason": "Unrelated functionality: sequence truncation behavior vs. scheduler-like training control. Similar wording only, not the same bug/change." + }, + { + "left": "issue:11005", + "right": "issue:11307", + "accept": false, + "reason": "Completely different features: learning-rate scheduler behavior vs. speech model time offsets. No plausible duplicate relationship." + } + ] + }, + "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": 17, + "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": "c844e731f4dd665315b8f75e9aa9c463f0707f96b01778ce85d5826e4f407878", + "component_budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 2479, + "estimated_input_tokens": 620, + "estimated_eval_tokens": 1496 + }, + "budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 2479, + "estimated_input_tokens": 620, + "estimated_eval_tokens": 1496 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "One issue is unrelated tokenization behavior, while the other two are both model-loading failures but appear to target different failure modes rather than the same concrete bug.", + "confidence": 0.79, + "canonical_issue_reason": "Issue #16618 is the broadest and most central model-loading report in the cluster, so it is the best representative of the loading-failure theme.", + "canonical_pr_reason": null, + "best_issue_reason": "#16618 best captures the shared model-loading symptom; #15980 is narrower and focused on a bad auth/error-message path, and #13244 is unrelated.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:13244", + "right": "issue:16618", + "accept": false, + "reason": "Tokenization differences in TAPAS vs. a model-load failure are different bugs in different code paths." + }, + { + "left": "issue:15980", + "right": "issue:16618", + "accept": false, + "reason": "Both involve model loading, but #15980 is specifically about private-model login/error handling, while #16618 is a generic inability to load 'bert-base-uncased'; not clearly the same 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": 4, + "component_count": 17, + "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": "d7f96afad994a04dedda35e173755ada1dd5f2e09a0fedd7a875873cd12cfffe", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1588, + "estimated_input_tokens": 397, + "estimated_eval_tokens": 1050 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1588, + "estimated_input_tokens": 397, + "estimated_eval_tokens": 1050 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are two separate community contribution tracking issues for different features: Flash Attention 2 support vs GGUF support. They are both broad architecture-support requests, but not the same underlying bug or change.", + "confidence": 0.98, + "canonical_issue_reason": "Issue #26350 is the better canonical issue because it is older and appears to be the earlier open community-tracking thread, but it is only canonical for its own feature area, not for the pair.", + "canonical_pr_reason": null, + "best_issue_reason": "#26350 is the stronger issue candidate by age and discussion activity, but #33260 is equally valid as a separate tracker for a different feature request.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:26350", + "right": "issue:33260", + "accept": false, + "reason": "Different requested features and code paths: Flash Attention 2 support for more architectures versus GGUF support for more architectures. Similar framing as community contributions is not enough 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": "issue", + "component_index": 5, + "component_count": 17, + "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-27088-6", + "nodes": [ + "issue:27088", + "issue:28936", + "issue:38435", + "issue:43834", + "issue:44365", + "issue:44370" + ], + "soft_pairs": [ + "issue:27088|issue:44370", + "issue:27088|issue:44365", + "issue:27088|issue:43834", + "issue:27088|issue:38435", + "issue:27088|issue:28936", + "issue:28936|issue:43834", + "issue:28936|issue:44365", + "issue:28936|issue:38435", + "issue:28936|issue:44370", + "issue:38435|issue:44370", + "issue:38435|issue:44365", + "issue:38435|issue:43834", + "issue:43834|issue:44370", + "issue:43834|issue:44365", + "issue:44365|issue:44370" + ], + "prepared_review_unit_hash": "f16fbed2cd573279a698a7c35051a0b971b06d85153c38f871de40c5aeadf51f", + "component_budget": { + "node_count": 6, + "item_count": 6, + "soft_pair_count": 15, + "serialized_chars": 8791, + "estimated_input_tokens": 2198, + "estimated_eval_tokens": 4652 + }, + "budget": { + "node_count": 6, + "item_count": 6, + "soft_pair_count": 15, + "serialized_chars": 8791, + "estimated_input_tokens": 2198, + "estimated_eval_tokens": 4652 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are translation-request issues for different languages plus near-identical template-placeholder issues. They share a common documentation-translation pattern, but they are not duplicates of the same underlying bug/change.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": "issue 43834 is the best representative of the shared template because it is the generic placeholder form of the docs-translation request, but it is still not a duplicate of the language-specific issues.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:27088", + "right": "issue:44370", + "accept": false, + "reason": "Different translation request instances; one is Turkish and the other is a generic placeholder template, so they are not the same underlying issue." + }, + { + "left": "issue:27088", + "right": "issue:44365", + "accept": false, + "reason": "Different language-specific request vs. generic placeholder; only the template is shared." + }, + { + "left": "issue:27088", + "right": "issue:43834", + "accept": false, + "reason": "Turkish docs translation request and a generic placeholder issue are not the same concrete issue." + }, + { + "left": "issue:27088", + "right": "issue:38435", + "accept": false, + "reason": "Different target languages (Turkish vs Romanian), so these are separate translation requests, not duplicates." + }, + { + "left": "issue:27088", + "right": "issue:28936", + "accept": false, + "reason": "Different target languages (Turkish vs Spanish); same topic, but distinct issues." + }, + { + "left": "issue:28936", + "right": "issue:43834", + "accept": false, + "reason": "Spanish translation request and a placeholder template are not the same underlying issue." + }, + { + "left": "issue:28936", + "right": "issue:44365", + "accept": false, + "reason": "Spanish-specific request vs. generic placeholder; shared template only." + }, + { + "left": "issue:28936", + "right": "issue:38435", + "accept": false, + "reason": "Different target languages (Spanish vs Romanian), so not duplicates." + }, + { + "left": "issue:28936", + "right": "issue:44370", + "accept": false, + "reason": "Spanish translation request and a generic placeholder issue are distinct items." + }, + { + "left": "issue:38435", + "right": "issue:44370", + "accept": false, + "reason": "Romanian-specific request vs. placeholder template; not the same issue." + }, + { + "left": "issue:38435", + "right": "issue:44365", + "accept": false, + "reason": "Romanian translation request and a generic placeholder template are not duplicates." + }, + { + "left": "issue:38435", + "right": "issue:43834", + "accept": false, + "reason": "Romanian-specific request vs. generic placeholder; only the translation-request template matches." + }, + { + "left": "issue:43834", + "right": "issue:44370", + "accept": false, + "reason": "Both are placeholder-style translation issues, but they are separate issue instances rather than one concrete bug/change." + }, + { + "left": "issue:43834", + "right": "issue:44365", + "accept": false, + "reason": "Near-identical placeholder template, but not the same underlying request." + }, + { + "left": "issue:44365", + "right": "issue:44370", + "accept": false, + "reason": "Two placeholder template issues; same form, but not a single duplicate report." + } + ] + }, + "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": 17, + "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-36817-16", + "nodes": [ + "issue:36817", + "issue:36879", + "issue:37015", + "issue:41115", + "issue:41211", + "issue:42503", + "issue:42738", + "issue:43671", + "issue:43837", + "issue:43909", + "issue:44121", + "issue:44205", + "issue:45027", + "issue:45177", + "issue:45322", + "issue:46133" + ], + "soft_pairs": [ + "issue:36817|issue:41211", + "issue:36817|issue:43671", + "issue:36817|issue:36879", + "issue:36879|issue:41211", + "issue:36879|issue:43909", + "issue:36879|issue:44121", + "issue:36879|issue:43837", + "issue:36879|issue:43671", + "issue:37015|issue:43837", + "issue:37015|issue:44121", + "issue:37015|issue:43909", + "issue:36879|issue:37015", + "issue:37015|issue:42503", + "issue:41115|issue:43837", + "issue:41115|issue:44121", + "issue:41115|issue:42503", + "issue:41115|issue:45027", + "issue:41211|issue:43671", + "issue:42503|issue:43837", + "issue:42503|issue:43671", + "issue:42503|issue:42738", + "issue:42738|issue:45027", + "issue:43671|issue:43837", + "issue:43671|issue:45027", + "issue:43671|issue:45177", + "issue:43671|issue:44121", + "issue: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:37015|issue:45027", + "issue:43837|issue:45177", + "issue:45322|issue:46133" + ], + "prepared_review_unit_hash": "121b167e370e3a0fe54c4c2373a23a6cf72bc855aa51d94d41c2eb239b63991a", + "component_budget": { + "node_count": 16, + "item_count": 16, + "soft_pair_count": 36, + "serialized_chars": 21655, + "estimated_input_tokens": 5414, + "estimated_eval_tokens": 11084 + }, + "budget": { + "node_count": 16, + "item_count": 16, + "soft_pair_count": 36, + "serialized_chars": 21655, + "estimated_input_tokens": 5414, + "estimated_eval_tokens": 11084 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These items are all broad model-request issues, but they do not describe the same underlying bug or change. The similarities are mostly topical (\"add model support\"), not duplicates.", + "confidence": 0.94, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": "issue:43671 is the strongest representative because it is a concrete, active request with notable discussion activity, but it should still be treated as distinct from the other model requests.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:36817", + "right": "issue:41211", + "accept": false, + "reason": "Different model requests: EuroBert vs DEIMv2. Same broad theme, but not the same model or code change." + }, + { + "left": "issue:36817", + "right": "issue:43671", + "accept": false, + "reason": "EuroBert config support and Qwen3-TTS support are unrelated model additions with different architectures and tasks." + }, + { + "left": "issue:36817", + "right": "issue:36879", + "accept": false, + "reason": "EuroBert and RF-DETR are completely different models and use cases." + }, + { + "left": "issue:36879", + "right": "issue:41211", + "accept": false, + "reason": "RF-DETR and DEIMv2 are different vision models; not the same underlying request." + }, + { + "left": "issue:36879", + "right": "issue:43909", + "accept": false, + "reason": "RF-DETR is a detection model, while LFM2.5 Audio 1.5B is an audio model request." + }, + { + "left": "issue:36879", + "right": "issue:44121", + "accept": false, + "reason": "RF-DETR and OpenAI Weight-Sparse Transformer target different model families and code paths." + }, + { + "left": "issue:36879", + "right": "issue:43837", + "accept": false, + "reason": "RF-DETR and Qwen3-ASR are unrelated; one is object detection, the other speech recognition." + }, + { + "left": "issue:36879", + "right": "issue:43671", + "accept": false, + "reason": "RF-DETR and Qwen3-TTS are different tasks and model implementations." + }, + { + "left": "issue:37015", + "right": "issue:43837", + "accept": false, + "reason": "NeoBERT and Qwen3-ASR are unrelated model requests; only the same generic \"add support\" pattern is shared." + }, + { + "left": "issue:37015", + "right": "issue:44121", + "accept": false, + "reason": "NeoBERT and the weight-sparse transformer request are different architectures with different integration work." + }, + { + "left": "issue:37015", + "right": "issue:43909", + "accept": false, + "reason": "NeoBERT is a text model request, while LFM2.5 Audio 1.5B is an audio model request." + }, + { + "left": "issue:36879", + "right": "issue:37015", + "accept": false, + "reason": "RF-DETR and NeoBERT are unrelated model additions with no shared concrete bug or patch." + }, + { + "left": "issue:37015", + "right": "issue:42503", + "accept": false, + "reason": "NeoBERT and ModernVBERT are distinct BERT-style model requests, not the same model or implementation." + }, + { + "left": "issue:41115", + "right": "issue:43837", + "accept": false, + "reason": "MiniCPM3 architecture support and Qwen3-ASR support are different model integrations." + }, + { + "left": "issue:41115", + "right": "issue:44121", + "accept": false, + "reason": "MiniCPM3 and the weight-sparse transformer request are different architectures and not the same code-path change." + }, + { + "left": "issue:41115", + "right": "issue:42503", + "accept": false, + "reason": "MiniCPM3 and ModernVBERT are separate model requests, not duplicates." + }, + { + "left": "issue:41115", + "right": "issue:45027", + "accept": false, + "reason": "MiniCPM3 and Voxtral TTS target different model families and modalities." + }, + { + "left": "issue:41211", + "right": "issue:43671", + "accept": false, + "reason": "DEIMv2 and Qwen3-TTS are unrelated model additions." + }, + { + "left": "issue:42503", + "right": "issue:43837", + "accept": false, + "reason": "ModernVBERT and Qwen3-ASR are different model families and tasks." + }, + { + "left": "issue:42503", + "right": "issue:43671", + "accept": false, + "reason": "ModernVBERT and Qwen3-TTS are not the same model or change." + }, + { + "left": "issue:42503", + "right": "issue:42738", + "accept": false, + "reason": "ModernVBERT and the generic BERT-like RoPE request are related only by theme, not by a specific duplicate fix." + }, + { + "left": "issue:42738", + "right": "issue:45027", + "accept": false, + "reason": "A generic BERT-like RoPE request and Voxtral TTS support are unrelated." + }, + { + "left": "issue:43671", + "right": "issue:43837", + "accept": false, + "reason": "Qwen3-TTS vs Qwen3-ASR are different modalities with separate implementations; same family, not same change." + }, + { + "left": "issue:43671", + "right": "issue:45027", + "accept": false, + "reason": "Qwen3-TTS support and Voxtral-4B-TTS support are different model requests, even though both are speech-related." + }, + { + "left": "issue:43671", + "right": "issue:45177", + "accept": false, + "reason": "Qwen3-TTS and DeepSeek-OCR2 are unrelated capabilities and model code paths." + }, + { + "left": "issue:43671", + "right": "issue:44121", + "accept": false, + "reason": "Qwen3-TTS and the weight-sparse transformer request are unrelated model additions." + }, + { + "left": "issue:43671", + "right": "issue:43909", + "accept": false, + "reason": "Qwen3-TTS and LFM2.5 Audio are both audio-related, but they are different model requests and not the same fix." + }, + { + "left": "issue:43837", + "right": "issue:44205", + "accept": false, + "reason": "Qwen3-ASR and SAM3-LiteText are different tasks and model integrations." + }, + { + "left": "issue:44121", + "right": "issue:45322", + "accept": false, + "reason": "Weight-sparse transformer support and EUPE are distinct model requests with no shared concrete patch." + }, + { + "left": "issue:43837", + "right": "issue:45322", + "accept": false, + "reason": "Qwen3-ASR and EUPE are unrelated model families and use cases." + }, + { + "left": "issue:45322", + "right": "issue:46133", + "accept": false, + "reason": "EUPE and TIPSv2 are both vision-language related, but they are different architectures/papers and not the same change." + }, + { + "left": "issue:43837", + "right": "issue:43909", + "accept": false, + "reason": "Qwen3-ASR and LFM2.5 Audio are distinct audio-model support requests, not duplicates." + }, + { + "left": "issue:43837", + "right": "issue:44121", + "accept": false, + "reason": "Speech recognition support and a sparse-transformer model request are unrelated." + }, + { + "left": "issue:43837", + "right": "issue:45177", + "accept": false, + "reason": "Qwen3-ASR and DeepSeek-OCR2 are separate modalities and different model integrations." + }, + { + "left": "issue:37015", + "right": "issue:45027", + "accept": false, + "reason": "NeoBERT and Voxtral TTS are different model families and tasks." + }, + { + "left": "issue:43909", + "right": "issue:44205", + "accept": false, + "reason": "LFM2.5 Audio is an audio model request, while SAM3-LiteText is a segmentation/text-related request." + } + ] + }, + "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": 17, + "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": "1e07463b0c45a1ac2a897eb2282b801eb6ea33f06e698fb4a0431d57e876b4fe", + "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": "The two issues are unrelated: one is about block decoding behavior, the other requests support for DeepSeek v3.2. No duplicate cluster here.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:41084", + "right": "issue:43037", + "accept": false, + "reason": "The titles and implied scopes do not match: block decoding is a low-level decoding behavior issue, while DeepSeek v3.2 support is a model-compatibility request. They do not appear to describe the same underlying bug or change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 8, + "component_count": 17, + "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": "6121320bf6993239ff9918d84eadfc7e9162ee12c9e1425158cc5031a10677cd", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1584, + "estimated_input_tokens": 396, + "estimated_eval_tokens": 1048 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1584, + "estimated_input_tokens": 396, + "estimated_eval_tokens": 1048 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Two closed issues about Trainer extensibility/custom trainers, but they target different behaviors: column removal customizability vs FLOPS estimation. They are related in theme, not duplicates.", + "confidence": 0.89, + "canonical_issue_reason": "Issue #43598 is the broader, more central Trainer customizability request and is the better representative of this cluster.", + "canonical_pr_reason": null, + "best_issue_reason": "#43598 is the most representative issue because it focuses on the core Trainer customization concern rather than a narrower FLOPS-estimation symptom.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43598", + "right": "issue:43600", + "accept": false, + "reason": "Both mention custom trainers, but they concern different code paths and outcomes: `remove_unused_column` behavior versus FLOPS estimation. They are related, not the same underlying bug or change." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 9, + "component_count": 17, + "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": "429e69995ba19e34da2f7c2541faf825dab3ae8a62b0e797d6013b2b216a9427", + "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 look like two separate feature requests, not duplicates: one asks for Qwen3-Omni registration in AutoModel/AutoModelForConditionalGeneration, the other asks for AutoModelForSequenceClassification support for Qwen3.5. Different model families and different auto-class targets.", + "confidence": 0.96, + "canonical_issue_reason": "Issue #43684 is the broader request and names explicit auto-class registration changes, so it is the better representative issue for the cluster even though the two issues are not duplicates.", + "canonical_pr_reason": null, + "best_issue_reason": "#43684 is the most general/central item in the set; #44405 is a distinct, narrower support request for a different model family and task head.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43684", + "right": "issue:44405", + "accept": false, + "reason": "Reject: they target different functionality. #43684 is about Qwen3-Omni registration for AutoModel/AutoModelForConditionalGeneration, while #44405 is about AutoModelForSequenceClassification support for Qwen3.5 (Qwen3_5Config)." + } + ] + }, + "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": 17, + "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": "ad48316e08176e03b2e29562203b8dc389155c714c49f10a077890c035e789d1", + "component_budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 6, + "serialized_chars": 4267, + "estimated_input_tokens": 1067, + "estimated_eval_tokens": 2390 + }, + "budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 6, + "serialized_chars": 4267, + "estimated_input_tokens": 1067, + "estimated_eval_tokens": 2390 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "43835 is unrelated, 43845 and 43846 look like the strongest duplicate pair, and 43859 is only a weaker title match because its body is noticeably shorter. I would anchor the cluster on 43845, but only accept the 43845\u219443846 soft edge.", + "confidence": 0.66, + "canonical_issue_reason": "issue:43845 is the earliest strong match in the cluster and pairs cleanly with issue:43846, while issue:43859 is less convincing on body evidence.", + "canonical_pr_reason": null, + "best_issue_reason": "issue:43845 is the best issue anchor because it is earliest among the near-duplicate huggingface reports and has the strongest duplicate signal with issue:43846.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:43835", + "right": "issue:43845", + "accept": false, + "reason": "'Fett' appears unrelated to the huggingface reports and does not share enough evidence of the same underlying issue." + }, + { + "left": "issue:43835", + "right": "issue:43846", + "accept": false, + "reason": "'Fett' is textually unrelated to the huggingface issues and does not look like the same report." + }, + { + "left": "issue:43835", + "right": "issue:43859", + "accept": false, + "reason": "'Fett' does not appear to describe the same underlying issue as the huggingface reports." + }, + { + "left": "issue:43845", + "right": "issue:43846", + "accept": true, + "reason": "Same title and matching body length make these look like the same report." + }, + { + "left": "issue:43845", + "right": "issue:43859", + "accept": false, + "reason": "The title matches, but the body is much shorter here, so the evidence is not strong enough to treat this as the same issue." + }, + { + "left": "issue:43846", + "right": "issue:43859", + "accept": false, + "reason": "They share a title, but the shorter body on 43859 makes this too weak to confidently merge as a duplicate." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "The cluster judgment is mostly plausible, but the writeup overstates the evidence. In the packet, 43845 and 43846 share the same title and body length, but 43859 has a much shorter body length (46 vs 84), so the repeated claims about \u201csame body characteristics\u201d / \u201csimilar body length\u201d are not grounded. The canonical-earliest choice for 43845 is fine, but the soft-edge rationales for 43845\u219443859 and 43846\u219443859 should be phrased more conservatively. The 43835 \u2018Fett\u2019 rejection is reasonable, but overall this is not conservative enough to accept as written." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 11, + "component_count": 17, + "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": "f17b9dcb542b461d01da14fe5df908e93606c85f96dd0bca0ba4ba155d8588e8", + "component_budget": { + "node_count": 5, + "item_count": 5, + "soft_pair_count": 9, + "serialized_chars": 5788, + "estimated_input_tokens": 1447, + "estimated_eval_tokens": 3150 + }, + "budget": { + "node_count": 5, + "item_count": 5, + "soft_pair_count": 9, + "serialized_chars": 5788, + "estimated_input_tokens": 1447, + "estimated_eval_tokens": 3150 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "All items are low-signal spam/generic placeholders with no substantive problem statement, and they do not appear to describe the same underlying issue. I would not merge any of the soft pairs.", + "confidence": 0.96, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:44371", + "right": "issue:44572", + "accept": false, + "reason": "Same placeholder title, but there is no substantive description to confirm a shared underlying bug/change; treat as separate spam artifacts." + }, + { + "left": "issue:44371", + "right": "issue:44573", + "accept": false, + "reason": "Both are generic spam placeholders, but there is no concrete evidence they refer to the same issue." + }, + { + "left": "issue:44371", + "right": "issue:44418", + "accept": false, + "reason": "One is spam and the other is a vague documentation suggestion; too little concrete overlap to call them duplicates." + }, + { + "left": "issue:44371", + "right": "issue:44937", + "accept": false, + "reason": "Different titles and no meaningful body detail; not enough evidence of the same underlying issue." + }, + { + "left": "issue:44418", + "right": "issue:44572", + "accept": false, + "reason": "A generic docs suggestion and a spam placeholder are not the same concrete issue." + }, + { + "left": "issue:44418", + "right": "issue:44573", + "accept": false, + "reason": "No shared bug/change is described; these look like unrelated low-information issues." + }, + { + "left": "issue:44572", + "right": "issue:44573", + "accept": false, + "reason": "Matching spam-style titles alone is not enough; there is no evidence they describe one concrete issue." + }, + { + "left": "issue:44572", + "right": "issue:44937", + "accept": false, + "reason": "Generic spam placeholder versus unrelated named-title issue; no clear duplicate relationship." + }, + { + "left": "issue:44573", + "right": "issue:44937", + "accept": false, + "reason": "Different titles and no substantive content to support merging as duplicates." + } + ] + }, + "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": 17, + "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": "22c27b177c03f29cdfffe2026c6bb2e657f7fcd695f13379907153b50ab3b495", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1645, + "estimated_input_tokens": 412, + "estimated_eval_tokens": 1080 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1645, + "estimated_input_tokens": 412, + "estimated_eval_tokens": 1080 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "#45145 is the best canonical issue anchor because it was opened first and already has an inbound reference. The pair looks related, but the evidence here is only identical titles plus a soft similarity signal, so I would treat the duplicate relationship as tentative rather than confirmed.", + "confidence": 0.79, + "canonical_issue_reason": "#45145 was created earlier and has one inbound reference, so it is the strongest anchor for the cluster.", + "canonical_pr_reason": null, + "best_issue_reason": "#45145 is the best representative issue because it is the earliest instance and already has slightly more observed attention.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45145", + "right": "issue:45151", + "accept": false, + "reason": "The pair shares an identical title and a moderate soft-similarity signal, but the packet does not provide enough direct evidence to conclude they are definitely the same underlying report." + } + ] + }, + "evaluator_result": { + "accept": false, + "feedback": "The canonical-choice rationale is grounded (earlier creation time and one inbound reference), but the duplicate verdict is a bit overstated. The packet only supports same title plus a soft-similarity signal (Jaccard 0.63); it does not directly show the bodies are \"very similar\" or that the issues are definitively the same underlying report. Please keep the soft-edge language more conservative." + }, + "evaluator_used": true, + "retried": true, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 13, + "component_count": 17, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45295-2", + "nodes": [ + "issue:45295", + "issue:45373" + ], + "soft_pairs": [], + "prepared_review_unit_hash": "f5fd95fada2f530ab0999e343c17c16266cde0c7182bb22509972ca44827cd49", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 0, + "serialized_chars": 1286, + "estimated_input_tokens": 322, + "estimated_eval_tokens": 900 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 0, + "serialized_chars": 1286, + "estimated_input_tokens": 322, + "estimated_eval_tokens": 900 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These two issues describe the same missing Gemma 4 sequence-classification support; the later, more specific open issue is the better cluster representative.", + "confidence": 0.94, + "canonical_issue_reason": "issue #45373 is the clearest canonical tracker: it is open, explicitly names the missing class `Gemma4ForSequenceClassification`, and states the module-level gap relative to Gemma 2/3.", + "canonical_pr_reason": null, + "best_issue_reason": "issue #45373 best captures the concrete request and likely remaining work, while #45295 is a shorter earlier duplicate with less detail and is already closed.", + "best_pr_reason": null, + "soft_edge_verdicts": [] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 14, + "component_count": 17, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45672-2", + "nodes": [ + "issue:45672", + "issue:45674" + ], + "soft_pairs": [ + "issue:45672|issue:45674" + ], + "prepared_review_unit_hash": "fb1b74ec895b58e48e3988f7988f784318fadee4c5f485eb4329ec1ed3c01c5f", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1669, + "estimated_input_tokens": 418, + "estimated_eval_tokens": 1092 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1669, + "estimated_input_tokens": 418, + "estimated_eval_tokens": 1092 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are two distinct open issues in the 4-bit inference/BitsAndBytes area, but they describe different failures: one is a Gemma4 audio-module `torch.finfo()` TypeError on `uint8` weights, the other is a `BitsAndBytesConfig.llm_int8_skip_modules` regression that drops the default `lm_head` exclusion and triggers an AssertionError. They should not be merged as duplicates.", + "confidence": 0.95, + "canonical_issue_reason": "Issue #45674 is the better canonical representative because it states a more concrete configuration-level regression and has a cleaner, actionable failure mode.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue #45674 appears to be the best issue in the cluster: it has a precise title, a specific root-cause hint, and is likely easier to triage and fix than the narrower Gemma4 audio-module report.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45672", + "right": "issue:45674", + "accept": false, + "reason": "Same broad 4-bit inference area, but different code paths and symptoms: Gemma4 audio modules hit `torch.finfo()` on `uint8`, while the other is a BitsAndBytesConfig `lm_head` exclusion regression causing an AssertionError. Not the same 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": 15, + "component_count": 17, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45762-2", + "nodes": [ + "issue:45762", + "issue:45763" + ], + "soft_pairs": [ + "issue:45762|issue:45763" + ], + "prepared_review_unit_hash": "2c726ebeeb7bb79fdf99689aabddaf190cc53b9ebd6679ebeeabfb6e8a1e6277", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1477, + "estimated_input_tokens": 370, + "estimated_eval_tokens": 996 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1477, + "estimated_input_tokens": 370, + "estimated_eval_tokens": 996 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Two sparse closed issues with different personal-name titles and dates; there is no evidence they describe the same underlying bug or request.", + "confidence": 0.96, + "canonical_issue_reason": "No canonical issue: the issues have different titles, no body, no shared targets, and no shared bug context.", + "canonical_pr_reason": null, + "best_issue_reason": "Neither issue is a good canonical choice; if forced, 45762 is only slightly earlier, but the pair does not look like duplicates.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45762", + "right": "issue:45763", + "accept": false, + "reason": "Reject: the titles refer to different people and dates, with no supporting discussion or linkage indicating the same underlying issue." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 16, + "component_count": 17, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45764-2", + "nodes": [ + "issue:45764", + "issue:45765" + ], + "soft_pairs": [ + "issue:45764|issue:45765" + ], + "prepared_review_unit_hash": "5d0004029bf5f12fd9ff7021273cce9580bb120c45c3e9a3fc928795a69e395f", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1481, + "estimated_input_tokens": 371, + "estimated_eval_tokens": 998 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 1481, + "estimated_input_tokens": 371, + "estimated_eval_tokens": 998 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are two unrelated, content-free issues with only personal-name/date titles; there is no evidence they describe the same underlying bug or request.", + "confidence": 0.97, + "canonical_issue_reason": "If a representative must be chosen, issue 45764 is the earliest, but neither issue has substantive content to justify a true canonical selection.", + "canonical_pr_reason": null, + "best_issue_reason": "Issue 45764 is only marginally better as a cluster representative because it appears first; otherwise it is no more informative than 45765.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45764", + "right": "issue:45765", + "accept": false, + "reason": "The titles reference different names and dates and contain no shared problem statement, so they do not look like duplicates." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "issue", + "component_index": 17, + "component_count": 17, + "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-45869-5", + "nodes": [ + "issue:45869", + "issue:45870", + "issue:45871", + "issue:45872", + "issue:45873" + ], + "soft_pairs": [ + "issue:45869|issue:45870", + "issue:45869|issue:45871", + "issue:45870|issue:45871", + "issue:45870|issue:45872", + "issue:45870|issue:45873", + "issue:45871|issue:45872", + "issue:45871|issue:45873", + "issue:45872|issue:45873" + ], + "prepared_review_unit_hash": "4afbd60b1863eeaaf60e38b49f402f472c08832d4b036ef02601a26fb8e2762f", + "component_budget": { + "node_count": 5, + "item_count": 5, + "soft_pair_count": 8, + "serialized_chars": 5726, + "estimated_input_tokens": 1432, + "estimated_eval_tokens": 3120 + }, + "budget": { + "node_count": 5, + "item_count": 5, + "soft_pair_count": 8, + "serialized_chars": 5726, + "estimated_input_tokens": 1432, + "estimated_eval_tokens": 3120 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "This cluster actually contains two separate duplicate pairs: 45870/45871 are exact title/body duplicates, and 45872/45873 are exact title/body duplicates. Issue 45869 is a different financial dataset item and should not be merged with either pair.", + "confidence": 0.98, + "canonical_issue_reason": "No single canonical issue cleanly represents the whole cluster because it mixes two distinct duplicate pairs plus one unrelated issue; within the \"Example: Financial time series prediction with DataForge\" pair, 45870 is the better canonical representative because it is slightly earlier and has more discussion activity.", + "canonical_pr_reason": null, + "best_issue_reason": "45870 is the best overall issue representative only among the exact duplicate pair it belongs to, but the cluster is not a single coherent duplicate set.", + "best_pr_reason": null, + "soft_edge_verdicts": [ + { + "left": "issue:45869", + "right": "issue:45870", + "accept": false, + "reason": "Different subjects: a financial data bundle listing vs an example/tutorial issue." + }, + { + "left": "issue:45869", + "right": "issue:45871", + "accept": false, + "reason": "Different subjects: a financial data bundle listing vs an example/tutorial issue." + }, + { + "left": "issue:45870", + "right": "issue:45871", + "accept": true, + "reason": "Same title and matching content pattern; these look like exact duplicate issues." + }, + { + "left": "issue:45870", + "right": "issue:45872", + "accept": false, + "reason": "Related finance/DataForge theme, but different requested change: example/tutorial vs scraping-service integration." + }, + { + "left": "issue:45870", + "right": "issue:45873", + "accept": false, + "reason": "Related finance/DataForge theme, but different requested change: example/tutorial vs scraping-service integration." + }, + { + "left": "issue:45871", + "right": "issue:45872", + "accept": false, + "reason": "Related finance/DataForge theme, but different requested change: example/tutorial vs scraping-service integration." + }, + { + "left": "issue:45871", + "right": "issue:45873", + "accept": false, + "reason": "Related finance/DataForge theme, but different requested change: example/tutorial vs scraping-service integration." + }, + { + "left": "issue:45872", + "right": "issue:45873", + "accept": true, + "reason": "Same title and matching content pattern; these look like exact duplicate issues." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is conservative and grounded in the packet: 45870/45871 and 45872/45873 have identical titles and max similarity, while 45869 is clearly a different dataset issue. The soft-edge verdicts do not overstate evidence, and the canonical choice for 45870 over 45871 is supported by earlier creation time and higher discussion activity." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 1, + "component_count": 27, + "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-39895-14", + "nodes": [ + "pull_request:39895", + "pull_request:41855", + "pull_request:42668", + "pull_request:43251", + "pull_request:43254", + "pull_request:43424", + "pull_request:43823", + "pull_request:43995", + "pull_request:44827", + "pull_request:45170", + "pull_request:45213", + "pull_request:45655", + "pull_request:46048", + "pull_request:46067" + ], + "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:46067", + "pull_request:43251|pull_request:43823", + "pull_request:43251|pull_request:44827", + "pull_request:43424|pull_request:46067", + "pull_request:43424|pull_request:44827", + "pull_request:43424|pull_request:43823", + "pull_request:43823|pull_request:45655", + "pull_request:43823|pull_request:43995", + "pull_request:43823|pull_request:46067", + "pull_request:43995|pull_request:44827", + "pull_request:43995|pull_request:45213", + "pull_request:44827|pull_request:46067", + "pull_request:44827|pull_request:45170", + "pull_request:44827|pull_request:46048", + "pull_request:44827|pull_request:45655", + "pull_request:44827|pull_request:45213", + "pull_request:43823|pull_request:45213", + "pull_request:43251|pull_request:43254" + ], + "prepared_review_unit_hash": "34711673bc32d80f93c4a3866901d04d2a4b9abe06a94df4ff932e8fa8deba11", + "component_budget": { + "node_count": 14, + "item_count": 14, + "soft_pair_count": 22, + "serialized_chars": 38369, + "estimated_input_tokens": 9593, + "estimated_eval_tokens": 19442 + }, + "budget": { + "node_count": 14, + "item_count": 14, + "soft_pair_count": 22, + "serialized_chars": 38369, + "estimated_input_tokens": 9593, + "estimated_eval_tokens": 19442 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "This cluster is mostly heterogeneous: model additions, tokenizer/API refactors, doc/test-only updates, and unrelated loading/conversion fixes. The only true near-duplicate pair is the fixed_cross_entropy kwargs change, where one PR is effectively a narrower variant of the other. ", + "confidence": 0.96, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 43251 is the better canonical representative: it has the explicit issue target (43240) and implements the concrete cross-entropy kwargs fix in one place. PR 43254 is a close variant of the same change.", + "best_issue_reason": null, + "best_pr_reason": "PR 43251 is the strongest cluster anchor because it directly addresses the underlying bug, has the explicit issue link, and is the only pair here that cleanly collapses into one duplicate fix.", + "soft_edge_verdicts": [ + { + "left": "pull_request:39895", + "right": "pull_request:43424", + "accept": false, + "reason": "Both are PRs, but one adds VideoPrism and the other adds a dynamic-shape export test; unrelated code paths and no plausible single merged fix." + }, + { + "left": "pull_request:41855", + "right": "pull_request:43823", + "accept": false, + "reason": "Tokenizer missing methods for MistralCommon is unrelated to adding MobileLLM; different models and different API surfaces." + }, + { + "left": "pull_request:41855", + "right": "pull_request:43424", + "accept": false, + "reason": "Mistral tokenizer method additions and ExecuTorch dynamic-shape tests are unrelated changes." + }, + { + "left": "pull_request:42668", + "right": "pull_request:43823", + "accept": false, + "reason": "Processor auto-loading robustness and adding a new model are not the same underlying bug/change." + }, + { + "left": "pull_request:43251", + "right": "pull_request:46067", + "accept": false, + "reason": "Cross-entropy kwargs support is unrelated to conversion/base_model_prefix loading fixes." + }, + { + "left": "pull_request:43251", + "right": "pull_request:43823", + "accept": false, + "reason": "A loss-function fix and a new model addition do not address the same code-path problem." + }, + { + "left": "pull_request:43251", + "right": "pull_request:44827", + "accept": false, + "reason": "Cross-entropy kwargs support is unrelated to Mistral4 test fixes and MOE contiguity changes." + }, + { + "left": "pull_request:43424", + "right": "pull_request:46067", + "accept": false, + "reason": "ExecuTorch export tests and loading/conversion prefix handling are unrelated." + }, + { + "left": "pull_request:43424", + "right": "pull_request:44827", + "accept": false, + "reason": "Dynamic-shape export tests and Mistral4 test fixes are different issues." + }, + { + "left": "pull_request:43424", + "right": "pull_request:43823", + "accept": false, + "reason": "A test for exportability does not match adding a new model implementation." + }, + { + "left": "pull_request:43823", + "right": "pull_request:45655", + "accept": false, + "reason": "New MobileLLM model support and config-resolution ordering in modeling_utils are unrelated." + }, + { + "left": "pull_request:43823", + "right": "pull_request:43995", + "accept": false, + "reason": "MobileLLM addition and Falcon output-interface refactoring are different model-specific changes." + }, + { + "left": "pull_request:43823", + "right": "pull_request:46067", + "accept": false, + "reason": "New model support and conversion prefix handling are not the same concrete fix." + }, + { + "left": "pull_request:43995", + "right": "pull_request:44827", + "accept": false, + "reason": "Falcon output collection refactor and Mistral4 test fixes do not target the same bug or change." + }, + { + "left": "pull_request:43995", + "right": "pull_request:45213", + "accept": false, + "reason": "Falcon refactoring and the model-creation skill are unrelated." + }, + { + "left": "pull_request:44827", + "right": "pull_request:46067", + "accept": false, + "reason": "Mistral4 test fixes and loading-prefix conversion fixes are unrelated." + }, + { + "left": "pull_request:44827", + "right": "pull_request:45170", + "accept": false, + "reason": "Mistral4 tests and layrnorm->layernorm renaming are different problems." + }, + { + "left": "pull_request:44827", + "right": "pull_request:46048", + "accept": false, + "reason": "Mistral4 tests and TDT loss kernel integration are unrelated." + }, + { + "left": "pull_request:44827", + "right": "pull_request:45655", + "accept": false, + "reason": "A model-specific test fix and a class-annotation resolution fix are not the same code-path problem." + }, + { + "left": "pull_request:44827", + "right": "pull_request:45213", + "accept": false, + "reason": "Mistral4 test fixes and the model creation skill are not mergeable as one duplicate change." + }, + { + "left": "pull_request:43823", + "right": "pull_request:45213", + "accept": false, + "reason": "Adding MobileLLM and adding an internal model-creation skill are unrelated." + }, + { + "left": "pull_request:43251", + "right": "pull_request:43254", + "accept": true, + "reason": "These are two near-identical fixes for the same fixed_cross_entropy kwargs issue in the same file and same issue target; they could plausibly be merged into one PR." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The cluster summary is grounded in the packet and stays conservative. The only accepted soft-edge pair, PR 43251 vs 43254, is clearly the same fixed_cross_entropy kwargs issue in the same file with the same issue target, so mergeability is plausible. The other rejections are appropriately cautious and supported by the diffs/titles." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 2, + "component_count": 27, + "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": "fddbcc199c0d370c1e3a472884f26270f09a2fcc808b96c123c6e8c7418b33d9", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 7223, + "estimated_input_tokens": 1806, + "estimated_eval_tokens": 3868 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 7223, + "estimated_input_tokens": 1806, + "estimated_eval_tokens": 3868 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Both PRs are related VibeVoice work, but they are not the same change: #40546 adds the core VibeVoice model/docs, while #42781 extends the stack with realtime support and a VibeVoice acoustic tokenizer. They share some scaffolding files, but the concrete scope differs enough that they should not be merged as duplicates.", + "confidence": 0.91, + "canonical_issue_reason": null, + "canonical_pr_reason": "#40546 is the cleaner baseline implementation of VibeVoice and is less specialized than #42781, which adds a separate realtime/tokenizer extension on top of that base work.", + "best_issue_reason": null, + "best_pr_reason": "#40546 is the best representative PR for the main VibeVoice integration because it focuses on the core model import path and documentation rather than the additional realtime-specific pieces.", + "soft_edge_verdicts": [ + { + "left": "pull_request:40546", + "right": "pull_request:42781", + "accept": false, + "reason": "Shared VibeVoice scaffolding is not enough here: #40546 implements the initial VibeVoice model, while #42781 adds realtime behavior plus an acoustic tokenizer and related generation changes. They are related, but not the same underlying bug/change and 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": 3, + "component_count": 27, + "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": "7c7c580b7abf5d0ac6dbf6cbe4b4ee641004624064b217a1095a17cacd224ec5", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4676, + "estimated_input_tokens": 1169, + "estimated_eval_tokens": 2594 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4676, + "estimated_input_tokens": 1169, + "estimated_eval_tokens": 2594 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are two unrelated feature PRs: one adds TimesFM covariate forecasting, the other adds DINOv3 image-classification support. They share only broad similarity as model feature work, not the same underlying change.", + "confidence": 0.96, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 41224 is the cleaner canonical representative: it is a focused, concrete feature addition with explicit model, auto-class, docs, and tests changes.", + "best_issue_reason": null, + "best_pr_reason": "PR 41224 is the strongest standalone PR in the cluster because it targets one specific code path (DINOv3 image classification support) and appears well-scoped for mergeability.", + "soft_edge_verdicts": [ + { + "left": "pull_request:40755", + "right": "pull_request:41224", + "accept": false, + "reason": "Reject: these PRs implement different features in different model families. They do not appear to fix the same bug or deliver the same change, and would not plausibly be merged 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": 4, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-41356-2", + "nodes": [ + "pull_request:41356", + "pull_request:44339" + ], + "soft_pairs": [ + "pull_request:41356|pull_request:44339" + ], + "prepared_review_unit_hash": "aead1308c167b0bae6ff11bf9358340baa5e76f816618383418b505a10a9cd01", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 6050, + "estimated_input_tokens": 1513, + "estimated_eval_tokens": 3282 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 6050, + "estimated_input_tokens": 1513, + "estimated_eval_tokens": 3282 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Two DEIMv2 pull requests target the same model family, but they are not the same change: PR 41356 is an early/basic model + image processor add, while PR 44339 is a much broader follow-up that adds loss code, auto-mappings, conversion tooling, modular files, and more integration work.", + "confidence": 0.91, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 44339 is the more complete and representative DEIMv2 implementation; PR 41356 looks like an earlier partial version of the same model bring-up.", + "best_issue_reason": null, + "best_pr_reason": "PR 44339 is the strongest single artifact for the cluster because it subsumes the basic model work and includes the broader integration needed for a full Transformers addition.", + "soft_edge_verdicts": [ + { + "left": "pull_request:41356", + "right": "pull_request:44339", + "accept": false, + "reason": "Same model area, but not the same concrete change: 44339 expands the implementation substantially and adds new code paths beyond the basic PR 41356. They look like related iterations, not mergeable duplicates." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 5, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-41797-3", + "nodes": [ + "pull_request:41797", + "pull_request:45186", + "pull_request:45630" + ], + "soft_pairs": [ + "pull_request:41797|pull_request:45186", + "pull_request:45186|pull_request:45630" + ], + "prepared_review_unit_hash": "6890fd3ec6a5a4ecde2b95032cdc9bbf0d5c780675cae266ca0a974ed9062bbf", + "component_budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 10435, + "estimated_input_tokens": 2609, + "estimated_eval_tokens": 5474 + }, + "budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 10435, + "estimated_input_tokens": 2609, + "estimated_eval_tokens": 5474 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are three separate model-addition pull requests (DeepSeek OCR, Isaac, Kimi2-6). They share repository scaffolding for adding new models, but they target different architectures and code paths, so they should not be deduplicated.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 45630 is the broadest and most complete of the three, with the largest integration surface and highest review activity, so it is the best representative PR for this cluster if one must be chosen.", + "best_issue_reason": null, + "best_pr_reason": "PR 45630 is the strongest standalone example of a full model integration PR: it adds docs, auto mappings, model code, processors, and tests for a distinct model family.", + "soft_edge_verdicts": [ + { + "left": "pull_request:41797", + "right": "pull_request:45186", + "accept": false, + "reason": "Different models and different implementations; overlap is only generic new-model scaffolding in docs/auto-registration, not the same bug or change." + }, + { + "left": "pull_request:45186", + "right": "pull_request:45630", + "accept": false, + "reason": "Separate model families with distinct files, configs, and tests; they do not fix the same code-path problem and would not plausibly be merged 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": 6, + "component_count": 27, + "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": "0cfbf19903bf1e47cfe00559a5c0f1fa58a0197681698021756d675fa9f5d8bc", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5935, + "estimated_input_tokens": 1484, + "estimated_eval_tokens": 3224 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5935, + "estimated_input_tokens": 1484, + "estimated_eval_tokens": 3224 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Both PRs are GitHub Actions maintenance updates, but they target different underlying changes: PR 42975 updates core JS-based workflow actions for Node 24 compatibility, while PR 42976 mainly bumps docker/build-push-action to v6. They overlap in a few workflow files, but they are not the same fix and should not be merged as duplicates.", + "confidence": 0.94, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 42975 is the broader, clearer representative of the cluster theme: it updates multiple GitHub Actions used across workflows specifically for Node 24 compatibility.", + "best_issue_reason": null, + "best_pr_reason": "PR 42975 is the best cluster representative because it has the widest scope and the most directly stated purpose, even though it is not a duplicate of PR 42976.", + "soft_edge_verdicts": [ + { + "left": "pull_request:42975", + "right": "pull_request:42976", + "accept": false, + "reason": "Different concrete changes: 42975 upgrades several JS-based actions for Node 24 compatibility, while 42976 mainly upgrades docker/build-push-action to v6. Shared workflow filenames alone are not enough to treat them as the same 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": 7, + "component_count": 27, + "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-42978-5", + "nodes": [ + "pull_request:42978", + "pull_request:43448", + "pull_request:43451", + "pull_request:45490", + "pull_request:46162" + ], + "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", + "pull_request:45490|pull_request:46162" + ], + "prepared_review_unit_hash": "ed7a3768c62003d32cab6c8a255bee3074b662598d489724ba9cb94e21c4b669", + "component_budget": { + "node_count": 5, + "item_count": 5, + "soft_pair_count": 7, + "serialized_chars": 16470, + "estimated_input_tokens": 4118, + "estimated_eval_tokens": 8492 + }, + "budget": { + "node_count": 5, + "item_count": 5, + "soft_pair_count": 7, + "serialized_chars": 16470, + "estimated_input_tokens": 4118, + "estimated_eval_tokens": 8492 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These PRs are only loosely similar as model-addition or test-change scaffolding, but they target different models/code paths and are not duplicates.", + "confidence": 0.95, + "canonical_issue_reason": null, + "canonical_pr_reason": "No canonical PR: 42978 adds ViT NEPA, 43448 adds Molmo, 43451 adds Molmo2, 45490 adds CTSM, and 46162 is a small test-only XPU tweak; they share template files but not the same underlying change.", + "best_issue_reason": null, + "best_pr_reason": "If one PR must represent the cluster, 43451 is the strongest standalone candidate because it is the largest and most complete model-addition effort, with the highest review activity; however, it is 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 generic scaffolding, not the same bug or change." + }, + { + "left": "pull_request:42978", + "right": "pull_request:43451", + "accept": false, + "reason": "Different model additions (ViT NEPA vs Molmo2); overlap is only in shared registration/docs files." + }, + { + "left": "pull_request:42978", + "right": "pull_request:45490", + "accept": false, + "reason": "Different model additions (ViT NEPA vs CTSM); same broad transformer repository plumbing, but unrelated concrete changes." + }, + { + "left": "pull_request:43448", + "right": "pull_request:43451", + "accept": false, + "reason": "Molmo and Molmo2 are distinct model ports with different files and functionality; too closely named, but not the same underlying PR." + }, + { + "left": "pull_request:43448", + "right": "pull_request:45490", + "accept": false, + "reason": "Molmo vs CTSM are unrelated model additions; shared auto/model registry edits are not enough to merge them as duplicates." + }, + { + "left": "pull_request:43451", + "right": "pull_request:45490", + "accept": false, + "reason": "Molmo2 and CTSM are separate model implementations with different conversion/modeling/processor code paths." + }, + { + "left": "pull_request:45490", + "right": "pull_request:46162", + "accept": false, + "reason": "CTSM model integration and a small Florence2/LFM2_VL deterministic test tweak are unrelated changes and could not plausibly be merged 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": 8, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43291-4", + "nodes": [ + "pull_request:43291", + "pull_request:43665", + "pull_request:44815", + "pull_request:46102" + ], + "soft_pairs": [ + "pull_request:43291|pull_request:43665", + "pull_request:43291|pull_request:44815", + "pull_request:43665|pull_request:46102", + "pull_request:43665|pull_request:44815" + ], + "prepared_review_unit_hash": "b80fc94beb90bd761466ca832ecb5f9ea59c818fbba9bcf703be777a77a71cb7", + "component_budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 4, + "serialized_chars": 9382, + "estimated_input_tokens": 2346, + "estimated_eval_tokens": 4948 + }, + "budget": { + "node_count": 4, + "item_count": 4, + "soft_pair_count": 4, + "serialized_chars": 9382, + "estimated_input_tokens": 2346, + "estimated_eval_tokens": 4948 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are four unrelated pull requests that only share shallow signals like test edits or generic fix titles. None appear to be duplicate fixes for the same underlying bug, so no soft edges should be merged.", + "confidence": 0.91, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 44815 is the most substantive and clearly scoped fix: it updates core loading and FP8 dequant logic across multiple affected files, with a concrete code-path change rather than test-only edits.", + "best_issue_reason": null, + "best_pr_reason": "PR 44815 is the best representative fix because it contains the real implementation change and associated test coverage, whereas the others are either test-only, unrelated, or narrower in scope.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43291", + "right": "pull_request:43665", + "accept": false, + "reason": "Both touch tests, but they target different components and different failure modes; no shared concrete bug or mergeable single fix is evident." + }, + { + "left": "pull_request:43291", + "right": "pull_request:44815", + "accept": false, + "reason": "Whisper tokenizer/test changes are unrelated to the dequant/core-loading FP8 fix; they do not address the same code-path problem." + }, + { + "left": "pull_request:43665", + "right": "pull_request:46102", + "accept": false, + "reason": "These are separate test/code changes in different subsystems (vision model tests vs FSDP tied-weights handling); they are not the same bug." + }, + { + "left": "pull_request:43665", + "right": "pull_request:44815", + "accept": false, + "reason": "Despite both mentioning tests, the underlying fixes are unrelated: CLIP/ViT test edits versus FP8 dequant and model-loading behavior." + } + ] + }, + "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": 27, + "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-43488-5", + "nodes": [ + "pull_request:43488", + "pull_request:43532", + "pull_request:45932", + "pull_request:46100", + "pull_request:46114" + ], + "soft_pairs": [ + "pull_request:43488|pull_request:46114", + "pull_request:43488|pull_request:43532", + "pull_request:43532|pull_request:46114", + "pull_request:45932|pull_request:46100", + "pull_request:46100|pull_request:46114" + ], + "prepared_review_unit_hash": "2c6668ccf409a08ab4bb022c52fbc3b3841cbb81e37ec1bbb42117662256e736", + "component_budget": { + "node_count": 5, + "item_count": 5, + "soft_pair_count": 5, + "serialized_chars": 11341, + "estimated_input_tokens": 2836, + "estimated_eval_tokens": 5928 + }, + "budget": { + "node_count": 5, + "item_count": 5, + "soft_pair_count": 5, + "serialized_chars": 11341, + "estimated_input_tokens": 2836, + "estimated_eval_tokens": 5928 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Only one clear duplicate pair appears here: the two ViT test-migration PRs with the same exact failing asserts. The other PRs are all different bot/CI or broader instrumentation experiments and should not be merged as duplicates.", + "confidence": 0.84, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #46100 is the cleanest representative of the duplicate ViT test-migration change: it is a minimal, self-contained diff in one test file, while #46114 appears to be the forked copy of the same patch.", + "best_issue_reason": null, + "best_pr_reason": "PR #46100 is the best standalone PR for the cluster because it is the smallest and clearest instance of the duplicated change, with no extra unrelated workflow or instrumentation edits.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43488", + "right": "pull_request:46114", + "accept": false, + "reason": "Both mention the ViT test file, but #43488 is mainly a setup/version/dependency bot-check change while #46114 is a test-migration fork; they are not the same concrete fix." + }, + { + "left": "pull_request:43488", + "right": "pull_request:43532", + "accept": false, + "reason": "These are different repository-bot/CI experiments with different edited files and goals; they do not share the same underlying change." + }, + { + "left": "pull_request:43532", + "right": "pull_request:46114", + "accept": false, + "reason": "#43532 is a workflow/self-comment CI change, whereas #46114 is a ViT test migration patch; same broad area, but not the same code-path change." + }, + { + "left": "pull_request:45932", + "right": "pull_request:46100", + "accept": false, + "reason": "#45932 is a much broader grafana/telemetry and model-file change set; despite both touching ViT tests, it does not look like the same concrete change as #46100." + }, + { + "left": "pull_request:46100", + "right": "pull_request:46114", + "accept": true, + "reason": "These are effectively the same PR: same single-file ViT test diff, same added assertions, and the titles indicate duplicate forked test-migration efforts." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The review is grounded in the packet and stays conservative. The only clear duplicate pair is 46100/46114: they have the same single-file ViT test diff with identical added assertions, and the titles support the forked-copy interpretation. The other pairings are correctly rejected because they share only loose topical similarity, not the same concrete fix." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 10, + "component_count": 27, + "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": "e2f796eda585d6c510b729fd1c38b7213100a2344f318cabf238a3677f0c6b69", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 6082, + "estimated_input_tokens": 1521, + "estimated_eval_tokens": 3298 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 6082, + "estimated_input_tokens": 1521, + "estimated_eval_tokens": 3298 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Two open PRs add similarly named but distinct pipelines: one for Promptable Concept Segmentation and one for Promptable Visual Segmentation. They share repository plumbing and docs, but the underlying feature scope, model integrations, and tests are different, so they should not be merged as duplicates.", + "confidence": 0.96, + "canonical_issue_reason": null, + "canonical_pr_reason": "No single canonical PR for a duplicate cluster: these are separate feature additions with different tasks, model stacks, and test coverage.", + "best_issue_reason": null, + "best_pr_reason": "PR 43613 is the better representative of the cluster only because it is the broader, more comprehensive pipeline addition; however, it is not a duplicate of 43612.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43612", + "right": "pull_request:43613", + "accept": false, + "reason": "Different concrete features: concept segmentation vs visual segmentation, with different docs, model files, and tests. Shared infrastructure alone is not enough to treat them 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": 27, + "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": "a7e226d0ba4b8d3c03e016adddcd7163523c471ee7dd8593843bfc7524e048be", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4293, + "estimated_input_tokens": 1074, + "estimated_eval_tokens": 2404 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4293, + "estimated_input_tokens": 1074, + "estimated_eval_tokens": 2404 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Two open PRs in `trainer.py` touch trainer customization, but they address different code paths: gradient-accumulation loss scaling vs. signature-column filtering. They are not duplicates.", + "confidence": 0.96, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 43651 is the more concrete behavior change: it updates `training_step` loss scaling semantics and adds a dedicated hook for custom trainers. If one PR must represent the cluster, this is the clearer canonical candidate.", + "best_issue_reason": null, + "best_pr_reason": "PR 43651 is the strongest standalone PR here because it fixes a specific training-path behavior with a direct code change, whereas PR 43663 is a separate refactor/hook for column filtering.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43651", + "right": "pull_request:43663", + "accept": false, + "reason": "Different trainer extensibility features in the same file, but not the same underlying bug or change: one concerns GA loss scaling, the other signature-column inspection. They are not plausibly 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": 12, + "component_count": 27, + "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-43656-2", + "nodes": [ + "pull_request:43656", + "pull_request:43842" + ], + "soft_pairs": [ + "pull_request:43656|pull_request:43842" + ], + "prepared_review_unit_hash": "eb024b05c1c4f838ea3d6e833be36b0219d8574a00350eab1b603a7b554c137a", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 3664, + "estimated_input_tokens": 916, + "estimated_eval_tokens": 2088 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 3664, + "estimated_input_tokens": 916, + "estimated_eval_tokens": 2088 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Two open PRs touch the same CLI file and both address annotation-related NameErrors, but they fix different runtime causes: one is a broader typing/future-annotations cleanup, the other specifically guards pydantic/TypeAdapter for environments without pydantic installed.", + "confidence": 0.76, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 43842 is the tighter, more concrete fix: it directly addresses the TypeAdapter NameError when pydantic is absent and has a clear, narrowly scoped change set.", + "best_issue_reason": null, + "best_pr_reason": "PR 43842 is the best standalone candidate because it fixes a specific, reproducible runtime failure with minimal collateral changes and an explicit issue target.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43656", + "right": "pull_request:43842", + "accept": false, + "reason": "Reject: despite similar titles and shared filename, these are not the same underlying bug. 43656 is a broader typing/future-annotations refactor around serve.py, while 43842 specifically fixes TypeAdapter import/name resolution when pydantic is not installed." + } + ] + }, + "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": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-43779-2", + "nodes": [ + "pull_request:43779", + "pull_request:43816" + ], + "soft_pairs": [ + "pull_request:43779|pull_request:43816" + ], + "prepared_review_unit_hash": "e7c674adcc031aa8883109a5dfc678b4407fb20b10764302092e384d031ba83a", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4362, + "estimated_input_tokens": 1091, + "estimated_eval_tokens": 2438 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4362, + "estimated_input_tokens": 1091, + "estimated_eval_tokens": 2438 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Both pull requests appear to implement the same SwanLab integration change: adding run id/resume support in `integration_utils.py` for issue #43698, with one variant also expanding the docs. They are strong duplicate PR candidates.", + "confidence": 0.94, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #43816 is the cleaner representative: it carries the same SwanLab id/resume support plus the accompanying documentation updates, making it the more complete version of the shared change.", + "best_issue_reason": null, + "best_pr_reason": "PR #43816 is the best global representative because it appears to include the full feature change and docs for SwanLab run id/resume support, while #43779 is the narrower implementation-focused variant.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43779", + "right": "pull_request:43816", + "accept": true, + "reason": "Same underlying SwanLab callback enhancement for id/resume support in the same file and same issue target; the doc differences do not change the core code-path change, so these are duplicate PRs." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet: both PRs target issue #43698, touch the same file, and the previews indicate the same SwanLab id/resume support with #43816 adding docs. The duplicate/mergeable verdict is conservative enough given the shared target and overlapping code change." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 14, + "component_count": 27, + "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: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:44018|pull_request:44116", + "pull_request:44071|pull_request:44722", + "pull_request:44018|pull_request:44086", + "pull_request:44018|pull_request:44085", + "pull_request:44068|pull_request:44086", + "pull_request:44071|pull_request:44085", + "pull_request:44059|pull_request:44086", + "pull_request:44085|pull_request:44086", + "pull_request:44098|pull_request:44101", + "pull_request:44059|pull_request:44722", + "pull_request:44018|pull_request:44071", + "pull_request:44017|pull_request:44722", + "pull_request:44017|pull_request:44161", + "pull_request:44017|pull_request:44073", + "pull_request:44068|pull_request:44071", + "pull_request:43997|pull_request:44073", + "pull_request:44071|pull_request:44086", + "pull_request:44073|pull_request:44101", + "pull_request:44017|pull_request:44071", + "pull_request:44059|pull_request:44071", + "pull_request:44056|pull_request:44101", + "pull_request:44073|pull_request:44161", + "pull_request:44059|pull_request:44068", + "pull_request:44018|pull_request:44059", + "pull_request:44101|pull_request:44161", + "pull_request:43997|pull_request:44017", + "pull_request:43997|pull_request:44056", + "pull_request:44071|pull_request:44076", + "pull_request:44074|pull_request:44101", + "pull_request:44056|pull_request:44161", + "pull_request:43997|pull_request:44018", + "pull_request:43997|pull_request:44161", + "pull_request:44074|pull_request:44161", + "pull_request:44018|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:44071|pull_request:44129", + "pull_request:44010|pull_request:44018", + "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" + ], + "prepared_review_unit_hash": "b410f3365977c6c09ed1b21c6b872882f6e649192146e6ac416ecf40cfda6f1c", + "component_budget": { + "node_count": 28, + "item_count": 28, + "soft_pair_count": 78, + "serialized_chars": 84902, + "estimated_input_tokens": 21226, + "estimated_eval_tokens": 42708 + }, + "budget": { + "node_count": 27, + "item_count": 27, + "soft_pair_count": 72, + "serialized_chars": 75098, + "estimated_input_tokens": 18775, + "estimated_eval_tokens": 37806 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "This cluster is centered on the umbrella standardized output-tracing refactor tracked by issue #43979, with multiple model-specific PRs. The clearest duplicate set is the GPT-J/GPT-Neo line; most other pairs are just unrelated models sharing the same refactor theme and should stay separate.", + "confidence": 0.82, + "canonical_issue_reason": "#43979 is the shared tracking issue for the standardized output-tracing work and is referenced by many of the PRs in this cluster.", + "canonical_pr_reason": "#44722 is the most complete representative of the GPT-J/CodeGen standardized output-tracing refactor and best subsumes the narrower GPT-J-only variants.", + "best_issue_reason": "#43979 is the only clear cluster-level issue anchor and matches the common refactor theme across the accepted duplicates.", + "best_pr_reason": "#44722 best represents the actual concrete change set among the duplicates because it covers the shared GPT-J path and also includes the related CodeGen update.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43996", + "right": "pull_request:44085", + "accept": false, + "reason": "Both are output-tracing refactors, but they target different model families and code paths (CVT/FNet vs GPT-J)." + }, + { + "left": "pull_request:43996", + "right": "pull_request:44044", + "accept": false, + "reason": "Different models and implementations; only the umbrella refactor theme is shared, not the concrete bug/change." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44085", + "accept": true, + "reason": "Both change the GPT-J output-tracing path in modeling_gptj.py; 44066 is a broader variant that also touches CodeGen, but the shared GPT-J fix is the same underlying change." + }, + { + "left": "pull_request:44007", + "right": "pull_request:44072", + "accept": false, + "reason": "Same tracking issue, but the actual code paths are unrelated model families (ResNet/RegNet vs EfficientNet/RT-DETR ResNet)." + }, + { + "left": "pull_request:44072", + "right": "pull_request:44722", + "accept": false, + "reason": "Both are output-tracing refactors, but they apply to different models and are not the same concrete fix." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44072", + "accept": false, + "reason": "GPT-J/CodeGen versus EfficientNet; too different to merge as the same underlying change." + }, + { + "left": "pull_request:44013", + "right": "pull_request:44044", + "accept": false, + "reason": "Different model families and different forward/output plumbing; shared issue target is not enough." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44086", + "accept": false, + "reason": "GPT-J/CodeGen refactor vs MGP-STR; same theme, but not the same code-path problem." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44071", + "accept": false, + "reason": "GPT-J/CodeGen and MPT are separate implementations; not the same concrete bug/change." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44068", + "accept": true, + "reason": "Both are GPT-Neo output-tracing refactors on the same model/file, so they look like the same underlying change." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44068", + "accept": false, + "reason": "GPT-J/CodeGen versus GPT-Neo; these are separate model-specific refactors." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44066", + "accept": false, + "reason": "Both are output-tracing work, but for different model code paths (GPT-Neo vs GPT-J/CodeGen)." + }, + { + "left": "pull_request:44019", + "right": "pull_request:44722", + "accept": false, + "reason": "ResNet output tracing and GPT-J/CodeGen output tracing are unrelated model changes." + }, + { + "left": "pull_request:44086", + "right": "pull_request:44722", + "accept": false, + "reason": "MGP-STR and GPT-J/CodeGen are different model implementations; not mergeable as one fix." + }, + { + "left": "pull_request:44019", + "right": "pull_request:44085", + "accept": false, + "reason": "Different models and different refactor scopes; only the generic output-tracing idea overlaps." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44019", + "accept": false, + "reason": "GPT-Neo and ResNet are unrelated code paths." + }, + { + "left": "pull_request:44019", + "right": "pull_request:44071", + "accept": false, + "reason": "ResNet versus MPT; same broad pattern, different concrete changes." + }, + { + "left": "pull_request:44068", + "right": "pull_request:44116", + "accept": false, + "reason": "GPT-Neo and Flaubert are unrelated model families." + }, + { + "left": "pull_request:44068", + "right": "pull_request:44722", + "accept": false, + "reason": "GPT-Neo refactor vs GPT-J/CodeGen refactor; not the same underlying model change." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44116", + "accept": false, + "reason": "GPT-Neo and Flaubert touch different model code paths." + }, + { + "left": "pull_request:44071", + "right": "pull_request:44722", + "accept": false, + "reason": "MPT and GPT-J/CodeGen are separate implementations, so these are not duplicates." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44086", + "accept": false, + "reason": "GPT-Neo versus MGP-STR; shared topic is too broad." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44085", + "accept": false, + "reason": "GPT-Neo and GPT-J are different model-specific output-tracing updates." + }, + { + "left": "pull_request:44068", + "right": "pull_request:44086", + "accept": false, + "reason": "Different model families and code paths." + }, + { + "left": "pull_request:44071", + "right": "pull_request:44085", + "accept": false, + "reason": "MPT and GPT-J are not the same fix." + }, + { + "left": "pull_request:44059", + "right": "pull_request:44086", + "accept": false, + "reason": "GPT-2 output tracing and MGP-STR output tracing are unrelated implementations." + }, + { + "left": "pull_request:44085", + "right": "pull_request:44086", + "accept": false, + "reason": "GPT-J/CodeGen versus MGP-STR; only the refactor style overlaps." + }, + { + "left": "pull_request:44098", + "right": "pull_request:44101", + "accept": false, + "reason": "ViLT and Flaubert are different model families and do not share the same code-path change." + }, + { + "left": "pull_request:44059", + "right": "pull_request:44722", + "accept": false, + "reason": "GPT-2 and GPT-J/CodeGen are separate model refactors." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44071", + "accept": false, + "reason": "GPT-Neo and MPT are distinct model implementations." + }, + { + "left": "pull_request:44017", + "right": "pull_request:44722", + "accept": false, + "reason": "SegFormer and GPT-J/CodeGen are unrelated concrete changes." + }, + { + "left": "pull_request:44017", + "right": "pull_request:44161", + "accept": false, + "reason": "SegFormer and LongT5 are different model code paths." + }, + { + "left": "pull_request:44017", + "right": "pull_request:44073", + "accept": false, + "reason": "SegFormer and VisualBert are unrelated model-specific refactors." + }, + { + "left": "pull_request:44068", + "right": "pull_request:44071", + "accept": false, + "reason": "GPT-Neo and MPT are separate implementations." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44073", + "accept": false, + "reason": "RegNet and VisualBert are different models; same refactor theme only." + }, + { + "left": "pull_request:44071", + "right": "pull_request:44076", + "accept": false, + "reason": "MPT and ImageGPT are unrelated model code paths." + }, + { + "left": "pull_request:44074", + "right": "pull_request:44101", + "accept": false, + "reason": "TextNet and Flaubert are different underlying changes." + }, + { + "left": "pull_request:44056", + "right": "pull_request:44161", + "accept": false, + "reason": "MPNet and LongT5 are not the same concrete bug/change." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44017", + "accept": false, + "reason": "RegNet and SegFormer are unrelated model refactors." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44056", + "accept": false, + "reason": "RegNet and MPNet are separate model-specific updates." + }, + { + "left": "pull_request:44071", + "right": "pull_request:44129", + "accept": false, + "reason": "MPT and SpeechT5 are different code paths." + }, + { + "left": "pull_request:44010", + "right": "pull_request:44071", + "accept": false, + "reason": "SqueezeBert and MPT are unrelated model implementations." + }, + { + "left": "pull_request:44010", + "right": "pull_request:44073", + "accept": false, + "reason": "SqueezeBert and VisualBert do not fix the same underlying issue." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44129", + "accept": false, + "reason": "GPT-Neo and SpeechT5 are different model families." + }, + { + "left": "pull_request:44074", + "right": "pull_request:44161", + "accept": false, + "reason": "TextNet and LongT5 are unrelated concrete changes." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44018", + "accept": false, + "reason": "RegNet and GPT-Neo are different model-specific refactors." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44161", + "accept": false, + "reason": "RegNet and LongT5 are unrelated code paths." + }, + { + "left": "pull_request:44074", + "right": "pull_request:44161", + "accept": false, + "reason": "TextNet and LongT5 are different model implementations." + }, + { + "left": "pull_request:44018", + "right": "pull_request:44059", + "accept": false, + "reason": "GPT-Neo and GPT-2 are separate model refactors." + }, + { + "left": "pull_request:44101", + "right": "pull_request:44161", + "accept": false, + "reason": "Flaubert and LongT5 are unrelated underlying changes." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44074", + "accept": false, + "reason": "RegNet and TextNet are different model code paths." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44129", + "accept": false, + "reason": "RegNet and SpeechT5 are unrelated model implementations." + }, + { + "left": "pull_request:44071", + "right": "pull_request:44129", + "accept": false, + "reason": "MPT and SpeechT5 are different concrete changes." + }, + { + "left": "pull_request:44010", + "right": "pull_request:44018", + "accept": false, + "reason": "SqueezeBert and GPT-Neo are separate model-specific refactors." + }, + { + "left": "pull_request:43996", + "right": "pull_request:44161", + "accept": false, + "reason": "CVT/FNet and LongT5 are different model families." + }, + { + "left": "pull_request:43996", + "right": "pull_request:44017", + "accept": false, + "reason": "CVT/FNet and SegFormer do not share the same concrete code path." + }, + { + "left": "pull_request:44010", + "right": "pull_request:44044", + "accept": false, + "reason": "SqueezeBert and DeBERTa v2 are unrelated model-specific changes." + }, + { + "left": "pull_request:44044", + "right": "pull_request:44074", + "accept": false, + "reason": "DeBERTa v2 and TextNet are different underlying fixes." + }, + { + "left": "pull_request:44066", + "right": "pull_request:44722", + "accept": true, + "reason": "They address the same GPT-J output-tracing refactor; 44722 is a later, broader variant that still includes the shared GPT-J fix." + }, + { + "left": "pull_request:44085", + "right": "pull_request:44722", + "accept": true, + "reason": "Both touch the GPT-J output-tracing path in modeling_gptj.py, so they look like the same underlying bug/change despite 44722 also including CodeGen." + }, + { + "left": "pull_request:44085", + "right": "pull_request:44129", + "accept": false, + "reason": "GPT-J and SpeechT5 are unrelated model code paths." + }, + { + "left": "pull_request:43996", + "right": "pull_request:44007", + "accept": false, + "reason": "Both are output-tracing refactors, but for different model families and concrete implementations." + }, + { + "left": "pull_request:43996", + "right": "pull_request:44129", + "accept": false, + "reason": "CVT/FNet and SpeechT5 are not the same fix." + }, + { + "left": "pull_request:44044", + "right": "pull_request:44129", + "accept": false, + "reason": "DeBERTa v2 and SpeechT5 are separate model-specific changes." + }, + { + "left": "pull_request:44044", + "right": "pull_request:44154", + "accept": false, + "reason": "DeBERTa v2 and VITS are unrelated model implementations." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The summary is grounded in the packet: #43979 is the clear shared issue anchor, and the chosen canonical PR #44722 is a reasonable broader representative of the GPT-J/CodeGen subset. The soft-edge verdicts are conservative overall, with accepts only where the diffs share the same model/file and likely the same bugfix, and rejects where only the high-level refactor theme overlaps." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 14, + "component_count": 27, + "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:44024", + "pull_request:44056", + "pull_request:44073", + "pull_request:44074", + "pull_request:44076", + "pull_request:44129" + ], + "soft_pairs": [ + "pull_request:44024|pull_request:44076", + "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": "91f4fcae2c6b189afeb35d562bbc343bcdc5a5db23bcf096c76932889c8500bc", + "component_budget": { + "node_count": 28, + "item_count": 28, + "soft_pair_count": 78, + "serialized_chars": 84902, + "estimated_input_tokens": 21226, + "estimated_eval_tokens": 42708 + }, + "budget": { + "node_count": 8, + "item_count": 8, + "soft_pair_count": 6, + "serialized_chars": 17125, + "estimated_input_tokens": 4282, + "estimated_eval_tokens": 8820 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": true, + "analyst_result": { + "summary": "These PRs are all model-specific refactors toward standardized output tracing/capture, but they target different architectures and code paths. They are related by pattern, not duplicates.", + "confidence": 0.96, + "canonical_issue_reason": null, + "canonical_pr_reason": "pull_request:44073 is the clearest representative of the shared refactor pattern: it explicitly adds _can_record_outputs plus can_return_tuple/capture_outputs for a single model, with a clean, self-contained change.", + "best_issue_reason": null, + "best_pr_reason": "pull_request:44073 is the best single representative PR for this cluster because it most clearly expresses the standardized output-tracing migration pattern, while still being narrowly scoped to one model.", + "soft_edge_verdicts": [ + { + "left": "pull_request:44024", + "right": "pull_request:44076", + "accept": false, + "reason": "Same broad output-tracing migration theme, but different models (FocalNet vs ImageGPT) and different code paths; not plausibly one merged fix." + }, + { + "left": "pull_request:44073", + "right": "pull_request:44074", + "accept": false, + "reason": "VisualBert and TextNet are separate model implementations; these are parallel refactors, not the same underlying bug/change." + }, + { + "left": "pull_request:44056", + "right": "pull_request:44073", + "accept": false, + "reason": "Both refactor output recording, but they touch different model stacks (MPNet vs VisualBert) and are not the same concrete change." + }, + { + "left": "pull_request:44056", + "right": "pull_request:44074", + "accept": false, + "reason": "Different models and different forward/output plumbing; too broad to treat as the same PR." + }, + { + "left": "pull_request:43997", + "right": "pull_request:44010", + "accept": false, + "reason": "RegNet and SqueezeBert are unrelated model code paths; shared decorator migration alone is not enough for duplication." + }, + { + "left": "pull_request:44010", + "right": "pull_request:44129", + "accept": false, + "reason": "SqueezeBert and SpeechT5 address different model internals and output semantics; not the same concrete fix." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 15, + "component_count": 27, + "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-43998-14", + "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", + "pull_request:46106" + ], + "soft_pairs": [ + "pull_request:43998|pull_request:44002", + "pull_request:43998|pull_request:44000", + "pull_request:43998|pull_request:44001", + "pull_request:43998|pull_request:44003", + "pull_request:43998|pull_request:44029", + "pull_request:43999|pull_request:44001", + "pull_request:43999|pull_request:44002", + "pull_request:43999|pull_request:44025", + "pull_request:43999|pull_request:44026", + "pull_request:43999|pull_request:44027", + "pull_request:44000|pull_request:44001", + "pull_request:44000|pull_request:44002", + "pull_request:44000|pull_request:44003", + "pull_request:44000|pull_request:44025", + "pull_request:44001|pull_request:44002", + "pull_request:44001|pull_request:44003", + "pull_request:44001|pull_request:44025", + "pull_request:44002|pull_request:44029", + "pull_request:44002|pull_request:44003", + "pull_request:44003|pull_request:44028", + "pull_request:44003|pull_request:44004", + "pull_request:43998|pull_request:44004", + "pull_request:44001|pull_request:44004", + "pull_request:44002|pull_request:44004", + "pull_request:44000|pull_request:44004", + "pull_request:44025|pull_request:44026", + "pull_request:44025|pull_request:44027", + "pull_request:44025|pull_request:44028", + "pull_request:44025|pull_request:44029", + "pull_request:44025|pull_request:44030", + "pull_request:44026|pull_request:44029", + "pull_request:44026|pull_request:44027", + "pull_request:44026|pull_request:44028", + "pull_request:44026|pull_request:44030", + "pull_request:44027|pull_request:44028", + "pull_request:44027|pull_request:44029", + "pull_request:44027|pull_request:44030", + "pull_request:44028|pull_request:44029", + "pull_request:44028|pull_request:44030", + "pull_request:44029|pull_request:44030", + "pull_request:44027|pull_request:46106", + "pull_request:44026|pull_request:46106", + "pull_request:44001|pull_request:46106", + "pull_request:44002|pull_request:46106", + "pull_request:43998|pull_request:46106" + ], + "prepared_review_unit_hash": "907983b1b73ad8ec5a92e896544ed32c60d398c871404c7f03b71207c9ac5307", + "component_budget": { + "node_count": 14, + "item_count": 14, + "soft_pair_count": 45, + "serialized_chars": 43785, + "estimated_input_tokens": 10947, + "estimated_eval_tokens": 22150 + }, + "budget": { + "node_count": 14, + "item_count": 14, + "soft_pair_count": 45, + "serialized_chars": 43785, + "estimated_input_tokens": 10947, + "estimated_eval_tokens": 22150 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are mostly separate model-specific output-tracing refactors (plus one docs toctree tweak), not duplicates of the same underlying bug/change.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": "PR 44003 is the broadest representative of the pattern, touching both `mamba` and `falcon_mamba`; however, it is still not a duplicate target for the other PRs.", + "soft_edge_verdicts": [ + { + "left": "pull_request:43998", + "right": "pull_request:44002", + "accept": false, + "reason": "Different model families/files; same refactor pattern, not the same concrete change." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44000", + "accept": false, + "reason": "Different model families/files; same refactor pattern, not the same concrete change." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44001", + "accept": false, + "reason": "Different model families/files; same refactor pattern, not the same concrete change." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44003", + "accept": false, + "reason": "Timm backbone vs Mamba/Falcon-Mamba code paths; only the tracing theme is shared." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44029", + "accept": false, + "reason": "Unrelated architectures (`timm_backbone` vs `rwkv`); not the same bug or change." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44001", + "accept": false, + "reason": "MobileNetV1 vs UnivNet are unrelated model paths; similar refactor style only." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44002", + "accept": false, + "reason": "MobileNetV1 vs UperNet are different code paths; not a duplicate change." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44025", + "accept": false, + "reason": "MobileNetV1 vs Depth Anything; same tracing refactor theme, but different models." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44026", + "accept": false, + "reason": "MobileNetV1 vs VisionEncoderDecoder are unrelated implementations." + }, + { + "left": "pull_request:43999", + "right": "pull_request:44027", + "accept": false, + "reason": "MobileNetV1 vs SpeechEncoderDecoder are different model wrappers; not the same fix." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44001", + "accept": false, + "reason": "Vision-Text dual encoder vs UnivNet are unrelated code paths." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44002", + "accept": false, + "reason": "Vision-Text dual encoder vs UperNet are different model families." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44003", + "accept": false, + "reason": "Vision-Text dual encoder vs Mamba/Falcon-Mamba are unrelated internals." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44025", + "accept": false, + "reason": "Vision-Text dual encoder vs Depth Anything are separate model implementations." + }, + { + "left": "pull_request:44001", + "right": "pull_request:44002", + "accept": false, + "reason": "UnivNet vs UperNet are different code paths; not a duplicate." + }, + { + "left": "pull_request:44001", + "right": "pull_request:44003", + "accept": false, + "reason": "UnivNet vs Mamba/Falcon-Mamba are unrelated model internals." + }, + { + "left": "pull_request:44001", + "right": "pull_request:44025", + "accept": false, + "reason": "UnivNet vs Depth Anything share only the tracing-refactor theme." + }, + { + "left": "pull_request:44002", + "right": "pull_request:44029", + "accept": false, + "reason": "UperNet vs RWKV are unrelated model families." + }, + { + "left": "pull_request:44002", + "right": "pull_request:44003", + "accept": false, + "reason": "UperNet vs Mamba/Falcon-Mamba are different code paths." + }, + { + "left": "pull_request:44003", + "right": "pull_request:44028", + "accept": false, + "reason": "Mamba/Falcon-Mamba vs SuperPoint are unrelated model implementations." + }, + { + "left": "pull_request:44003", + "right": "pull_request:44004", + "accept": false, + "reason": "Mamba/Falcon-Mamba refactor vs CodeGen refactor; same theme, different fix." + }, + { + "left": "pull_request:43998", + "right": "pull_request:44004", + "accept": false, + "reason": "Timm backbone vs CodeGen are unrelated architectures." + }, + { + "left": "pull_request:44001", + "right": "pull_request:44004", + "accept": false, + "reason": "UnivNet vs CodeGen are different model paths." + }, + { + "left": "pull_request:44002", + "right": "pull_request:44004", + "accept": false, + "reason": "UperNet vs CodeGen are unrelated code paths." + }, + { + "left": "pull_request:44000", + "right": "pull_request:44004", + "accept": false, + "reason": "Vision-Text dual encoder vs CodeGen are different implementations." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44026", + "accept": false, + "reason": "Depth Anything vs VisionEncoderDecoder are separate model families." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44027", + "accept": false, + "reason": "Depth Anything vs SpeechEncoderDecoder are unrelated code paths." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44028", + "accept": false, + "reason": "Depth Anything vs SuperPoint are unrelated model implementations." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44029", + "accept": false, + "reason": "Depth Anything vs RWKV are different architectures." + }, + { + "left": "pull_request:44025", + "right": "pull_request:44030", + "accept": false, + "reason": "Depth Anything vs DPR are unrelated model paths." + }, + { + "left": "pull_request:44026", + "right": "pull_request:44027", + "accept": false, + "reason": "VisionEncoderDecoder vs SpeechEncoderDecoder are distinct wrappers; not the same change." + }, + { + "left": "pull_request:44026", + "right": "pull_request:44028", + "accept": false, + "reason": "VisionEncoderDecoder vs SuperPoint are unrelated code paths." + }, + { + "left": "pull_request:44026", + "right": "pull_request:44029", + "accept": false, + "reason": "VisionEncoderDecoder vs RWKV are different model families." + }, + { + "left": "pull_request:44026", + "right": "pull_request:44030", + "accept": false, + "reason": "VisionEncoderDecoder vs DPR are unrelated implementations." + }, + { + "left": "pull_request:44027", + "right": "pull_request:44028", + "accept": false, + "reason": "SpeechEncoderDecoder vs SuperPoint are different model paths." + }, + { + "left": "pull_request:44027", + "right": "pull_request:44029", + "accept": false, + "reason": "SpeechEncoderDecoder vs RWKV are unrelated architectures." + }, + { + "left": "pull_request:44027", + "right": "pull_request:44030", + "accept": false, + "reason": "SpeechEncoderDecoder vs DPR are different code paths." + }, + { + "left": "pull_request:44028", + "right": "pull_request:44029", + "accept": false, + "reason": "SuperPoint vs RWKV are unrelated model implementations." + }, + { + "left": "pull_request:44028", + "right": "pull_request:44030", + "accept": false, + "reason": "SuperPoint vs DPR are different code paths." + }, + { + "left": "pull_request:44029", + "right": "pull_request:44030", + "accept": false, + "reason": "RWKV vs DPR are unrelated model families." + }, + { + "left": "pull_request:44027", + "right": "pull_request:46106", + "accept": false, + "reason": "Model code refactor vs docs-only toctree edit; unrelated changes." + }, + { + "left": "pull_request:44026", + "right": "pull_request:46106", + "accept": false, + "reason": "Model code refactor vs docs-only toctree edit; unrelated changes." + }, + { + "left": "pull_request:44001", + "right": "pull_request:46106", + "accept": false, + "reason": "Model code refactor vs docs-only toctree edit; unrelated changes." + }, + { + "left": "pull_request:44002", + "right": "pull_request:46106", + "accept": false, + "reason": "Model code refactor vs docs-only toctree edit; unrelated changes." + }, + { + "left": "pull_request:43998", + "right": "pull_request:46106", + "accept": false, + "reason": "Model code refactor vs docs-only toctree edit; unrelated changes." + } + ] + }, + "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": 27, + "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": "5abf70594dffdac0306fb70326d5db90eca828859196b9e7e2076574713ca9b8", + "component_budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 3, + "serialized_chars": 10912, + "estimated_input_tokens": 2728, + "estimated_eval_tokens": 5712 + }, + "budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 3, + "serialized_chars": 10912, + "estimated_input_tokens": 2728, + "estimated_eval_tokens": 5712 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are three separate model-introduction PRs (PenguinVL, DeepSeek-OCR-2, Voxtral TTS). They share generic HF wiring files, but they target different models, different code paths, and different upstream tasks, so they are not duplicates.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": null, + "best_issue_reason": null, + "best_pr_reason": "PR 45075 is the strongest standalone representative: it is a full model addition with conversion code, tests, and broad integration, while the others are either a draft or an unrelated model implementation.", + "soft_edge_verdicts": [ + { + "left": "pull_request:44662", + "right": "pull_request:45401", + "accept": false, + "reason": "Different model families and different implementations (PenguinVL vs Voxtral TTS). The overlap is only shared registration/docs plumbing, not the same bug or change." + }, + { + "left": "pull_request:44662", + "right": "pull_request:45075", + "accept": false, + "reason": "Unrelated model additions for different architectures and tasks. Shared auto-mapping and docs files are expected for any new model, not evidence of duplication." + }, + { + "left": "pull_request:45075", + "right": "pull_request:45401", + "accept": false, + "reason": "Different concrete changes: DeepSeek-OCR-2 vision/OCR model vs Voxtral TTS speech model. They touch similar integration files but fix/add entirely separate code paths." + } + ] + }, + "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": 27, + "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": "0128bde623a28b227d67692204923896334f6472aa9796418a3bb986c043b3fd", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4408, + "estimated_input_tokens": 1102, + "estimated_eval_tokens": 2460 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4408, + "estimated_input_tokens": 1102, + "estimated_eval_tokens": 2460 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are related kernel-fusion feature PRs, but they implement different layers of the stack: one introduces a generic module fusion API, while the other adds KernelConfig-driven n-to-1 fusion integration and an example. They are not duplicates.", + "confidence": 0.91, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #44979 is the more foundational and self-contained feature: it adds a new `Module Fusion API` with its own implementation and tests, making it the best representative of the underlying cluster theme.", + "best_issue_reason": null, + "best_pr_reason": "PR #45363 is the broader end-user-facing integration of kernel fusion via `KernelConfig`, touching runtime plumbing and an example, so it is the strongest single PR for the practical feature direction.", + "soft_edge_verdicts": [ + { + "left": "pull_request:44979", + "right": "pull_request:45363", + "accept": false, + "reason": "Related feature area, but not the same concrete change. #44979 adds a generic module-fusion API, while #45363 adds KernelConfig-based n-to-1 fusion integration and example usage; they would not plausibly merge 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": 18, + "component_count": 27, + "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": "2d800f4c74ef4dd36824e0f767f7478ee7697429eb17359f73984b8c3d0aa3ec", + "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 PRs address the same Mistral regex patching AttributeError in the same function and file, with 45317 appearing to be a follow-up/expanded version of 45086 that adds tests.", + "confidence": 0.98, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 45317 is the better canonical PR because it covers the same fix as 45086 but is more complete, including test changes and more discussion activity.", + "best_issue_reason": null, + "best_pr_reason": "45317 is the strongest representative of the cluster: same concrete bug fix, same code path, and it adds validation via tests, making it the best stand-alone PR for this issue.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45086", + "right": "pull_request:45317", + "accept": true, + "reason": "Same underlying AttributeError in _patch_mistral_regex, same file and issue target, and the second PR is essentially the same fix plus tests rather than a different 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": 19, + "component_count": 27, + "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": "8fb0f926f81cb1f4135ca2b2bb29ce30f99835e864588285ea85bac1deb19ad0", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 6065, + "estimated_input_tokens": 1517, + "estimated_eval_tokens": 3290 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 6065, + "estimated_input_tokens": 1517, + "estimated_eval_tokens": 3290 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "The two PRs are related to NemotronH, but they are not the same change: #45115 is a conversion/refactor toward GraniteMoeHybrid-compatible inheritance, while #45569 is a broader model split introducing NemotronH dense/sparse variants plus docs and auto-mapping updates. They should not be deduplicated.", + "confidence": 0.9, + "canonical_issue_reason": null, + "canonical_pr_reason": "No clear canonical duplicate PR: #45115 and #45569 address different code paths and scopes within NemotronH rather than the same concrete fix.", + "best_issue_reason": null, + "best_pr_reason": "#45569 is the better representative PR for the cluster because it is broader and more complete, covering the dedicated NemotronH dense/sparse model split and associated integration changes.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45115", + "right": "pull_request:45569", + "accept": false, + "reason": "Both touch NemotronH internals, but #45115 is a targeted conversion-mapping/inheritance refactor while #45569 restructures the model into dense/sparse variants with docs and auto-class updates. Different underlying changes, not mergeable 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": 20, + "component_count": 27, + "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": "6289851ccc737cd6b8fc57e885c7770ad7fa9ef1213d05014240fde1315a0ece", + "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: one adds heterogeneous configuration plumbing, while the other builds broader heterogeneous model/runtime support on top of it. They touch overlapping files because the model-support PR depends on the config support, but they are distinct changes and not the same fix/feature.", + "confidence": 0.91, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #45332 is the better canonical representative because it is the broader end-to-end feature PR: it includes the runtime/model changes, cache/masking updates, and tests, while #45333 is narrower foundational config plumbing.", + "best_issue_reason": null, + "best_pr_reason": "PR #45332 is the strongest cluster anchor since it covers the most user-visible functionality and the wider code path impacted by heterogeneous support.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45332", + "right": "pull_request:45333", + "accept": false, + "reason": "Shared heterogeneity/config infrastructure, but different scope: #45333 adds per-layer config support, while #45332 adds heterogeneous modeling/runtime behavior. They 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": 21, + "component_count": 27, + "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": "788b84e897a4369a96032119ae1332cdc2b3cecf44cb44f923d2c76879435e1f", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5514, + "estimated_input_tokens": 1379, + "estimated_eval_tokens": 3014 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 5514, + "estimated_input_tokens": 1379, + "estimated_eval_tokens": 3014 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are two separate typing-focused PRs. PR 45415 is a broad mechanical type-checking pass across many transformer modules, while PR 45425 adds specific typing protocols/helpers and related adjustments. They overlap in a few files but not in the same concrete change or bug fix.", + "confidence": 0.92, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 45425 is the better canonical representative because it targets a specific typing gap with concrete helper protocols and narrow follow-up changes, whereas PR 45415 is a large broad-sweep type-checking pass.", + "best_issue_reason": null, + "best_pr_reason": "PR 45425 is the stronger standalone PR: it is more focused, more cohesive, and easier to merge as a single change set than the wide-ranging type-annotation sweep in PR 45415.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45415", + "right": "pull_request:45425", + "accept": false, + "reason": "Both are about typing, but they do not fix the same underlying problem. PR 45415 is a broad annotation/type-ignore cleanup across many files; PR 45425 adds specific typing protocols and utility fixes in a different code path. Shared files alone are not enough 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": 22, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45487-5", + "nodes": [ + "pull_request:45487", + "pull_request:46101", + "pull_request:46109", + "pull_request:46135", + "pull_request:46151" + ], + "soft_pairs": [ + "pull_request:45487|pull_request:46101", + "pull_request:45487|pull_request:46151", + "pull_request:46101|pull_request:46151", + "pull_request:46109|pull_request:46151", + "pull_request:46135|pull_request:46151" + ], + "prepared_review_unit_hash": "be8cbaedf4f262355bfc732659cb71f8c3abe5037793bf873db5f1e1a2fb230d", + "component_budget": { + "node_count": 5, + "item_count": 5, + "soft_pair_count": 5, + "serialized_chars": 12530, + "estimated_input_tokens": 3133, + "estimated_eval_tokens": 6522 + }, + "budget": { + "node_count": 5, + "item_count": 5, + "soft_pair_count": 5, + "serialized_chars": 12530, + "estimated_input_tokens": 3133, + "estimated_eval_tokens": 6522 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These PRs are not duplicates of one another overall: they split into three different themes\u2014model device-placement fixes, distributed-test wording updates, and test expectation updates for unrelated suites. The only close resemblance is between the model-parallel PRs, but they still target different model code paths and should not be merged as one change.", + "confidence": 0.91, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 45487 is the broadest and most substantive fix: it addresses an actual model-parallel device placement bug across several model implementations, not just a test adjustment.", + "best_issue_reason": null, + "best_pr_reason": "PR 45487 is the best representative of the cluster because it fixes a real runtime bug in core model code and has the clearest production impact.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45487", + "right": "pull_request:46101", + "accept": false, + "reason": "Both are device-placement fixes for model parallelism, but they touch different concrete model code paths (AltCLIP/ChineseCLIP vs Qwen2-Audio) and do not look like the same underlying bug." + }, + { + "left": "pull_request:45487", + "right": "pull_request:46151", + "accept": false, + "reason": "45487 fixes model code; 46151 only updates test expectations for pipeline outputs. Same broad theme of test/runtime compatibility, but not the same bug or change." + }, + { + "left": "pull_request:46101", + "right": "pull_request:46151", + "accept": false, + "reason": "Qwen2-Audio runtime masking logic and token-classification test expectations are unrelated changes, despite both mentioning accelerator compatibility." + }, + { + "left": "pull_request:46109", + "right": "pull_request:46151", + "accept": false, + "reason": "These are both test-only changes, but they affect different distributed-trainer vs pipeline test suites and different issues; not one underlying change." + }, + { + "left": "pull_request:46135", + "right": "pull_request:46151", + "accept": false, + "reason": "Both are test expectation updates, but for different model/pipeline areas with no shared code path or same concrete bug." + } + ] + }, + "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": 27, + "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-45513-3", + "nodes": [ + "pull_request:45513", + "pull_request:46084", + "pull_request:46098" + ], + "soft_pairs": [ + "pull_request:45513|pull_request:46084", + "pull_request:46084|pull_request:46098" + ], + "prepared_review_unit_hash": "e5890af0825d84e02e814b57a6f4e042b526fed41454c837e9ee58357ab32d86", + "component_budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 7426, + "estimated_input_tokens": 1857, + "estimated_eval_tokens": 3970 + }, + "budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 2, + "serialized_chars": 7426, + "estimated_input_tokens": 1857, + "estimated_eval_tokens": 3970 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Two PRs are about the same Mamba2 cache-length bug, while the Qwen3.5 PR is a separate model-specific fix. The Mamba2 docs+code PR is the best representative of the shared bug fix.", + "confidence": 0.91, + "canonical_issue_reason": null, + "canonical_pr_reason": "pull_request:46098 is the broadest, most complete Mamba2 fix: it targets the same issue, covers the same code path, and includes tests/docs in addition to the implementation change.", + "best_issue_reason": null, + "best_pr_reason": "pull_request:46098 best represents the cluster because it addresses the shared Mamba2 use_cache seq_len>1 bug with the most complete scope and supporting updates.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45513", + "right": "pull_request:46084", + "accept": false, + "reason": "Different model families and different code paths: Qwen3.5 GDN linear attention vs Mamba2 cache handling. Similar symptom, but not the same underlying change." + }, + { + "left": "pull_request:46084", + "right": "pull_request:46098", + "accept": true, + "reason": "Same underlying Mamba2 use_cache seq_len>1 bug in the same module and same issue target; 46098 appears to be the broader, more complete version of the same fix." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The clustering is grounded in the packet: 46084 and 46098 share the same Mamba2 `use_cache`/`seq_len > 1` issue target and code path, while 45513 is a different model-family/code-path fix. The mergeability judgment for 46084 + 46098 is conservative and reasonable, and the best-PR choice for 46098 is supported by its broader scope (docs + tests + implementation)." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 24, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45548-2", + "nodes": [ + "pull_request:45548", + "pull_request:45662" + ], + "soft_pairs": [ + "pull_request:45548|pull_request:45662" + ], + "prepared_review_unit_hash": "dee3acf86fc055c7478981185eb4a3b15fbdcba1b4ad5c608aeb5b4f28ff672b", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4508, + "estimated_input_tokens": 1127, + "estimated_eval_tokens": 2510 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4508, + "estimated_input_tokens": 1127, + "estimated_eval_tokens": 2510 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These are related only at a high level (expert-parallel loading), but they fix different bugs in different code paths: one is config/ZeRO-3 launch handling, the other is shard-preservation during EP/FSDP2 parameter loading.", + "confidence": 0.91, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 45662 is the stronger representative because it fixes the actual parameter-assignment path and spans the EP/FSDP2 loading logic where the corruption occurs; PR 45548 is narrower and mostly about config serialization/feature detection for a different loading scenario.", + "best_issue_reason": null, + "best_pr_reason": "PR 45662 is the best overall fit for the cluster because it addresses a concrete runtime failure mode in loading/sharding behavior, with broader and more central code-path changes than 45548.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45548", + "right": "pull_request:45662", + "accept": false, + "reason": "Both concern expert-parallel loading, but 45548 targets distributed_config serialization and an EP capability flag for DeepSpeed ZeRO-3/accelerate launch, while 45662 fixes parameter wrapping during model loading to stop expert weights being overwritten under FSDP2. Different underlying bugs and not plausibly 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": 25, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "llm", + "cache_hit": false, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45639-2", + "nodes": [ + "pull_request:45639", + "pull_request:45645" + ], + "soft_pairs": [ + "pull_request:45639|pull_request:45645" + ], + "prepared_review_unit_hash": "1071af12868f40796d4fb774c7ec29d14db6adc3ed0714e94b323ad2b20b112b", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4590, + "estimated_input_tokens": 1148, + "estimated_eval_tokens": 2552 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4590, + "estimated_input_tokens": 1148, + "estimated_eval_tokens": 2552 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Both PRs target the same xdist-safe captured_info logging problem in testing_utils; 45645 is the broader, more complete fix and appears to subsume 45639.", + "confidence": 0.93, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 45645 is the stronger canonical choice because it contains the core testing_utils fix plus the CI/workflow handling needed to preserve and aggregate multiple captured_info files, making it the more complete implementation.", + "best_issue_reason": null, + "best_pr_reason": "PR 45645 is the best overall PR because it addresses the same underlying bug with the xdist-safe output path change and also updates downstream consumers, so it is more likely to fully resolve the issue.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45639", + "right": "pull_request:45645", + "accept": true, + "reason": "Same underlying captured_info xdist collision fix. 45639 is a narrower subset of the broader 45645 change set, and the overlapping testing_utils/test adjustments make them plausibly mergeable into one PR." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 26, + "component_count": 27, + "review_unit_index": 1, + "review_unit_count": 1, + "status": "reviewed", + "reason": null, + "source": "cache", + "cache_hit": true, + "model": "gpt-5.4-mini?service_tier=flex", + "cluster_id": "cluster-45737-2", + "nodes": [ + "pull_request:45737", + "pull_request:45738" + ], + "soft_pairs": [ + "pull_request:45737|pull_request:45738" + ], + "prepared_review_unit_hash": "cc2ade73a49ec6a8883a7a24c433988645588bcd049e8aad2bce56b1b24daa1c", + "component_budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4131, + "estimated_input_tokens": 1033, + "estimated_eval_tokens": 2322 + }, + "budget": { + "node_count": 2, + "item_count": 2, + "soft_pair_count": 1, + "serialized_chars": 4131, + "estimated_input_tokens": 1033, + "estimated_eval_tokens": 2322 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "Both PRs address the same MusicgenMelody cache initialization bug in `forward()`: replacing `EncoderDecoderCache(DynamicCache, DynamicCache)` with a single `DynamicCache`. PR #45737 is the more complete duplicate because it bundles the fix with a regression test, while #45738 is the fix-only variant.", + "confidence": 0.97, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR #45737 is the best canonical PR because it contains the same code fix as #45738 plus a regression test, making it the most complete representation of the bug fix.", + "best_issue_reason": null, + "best_pr_reason": "PR #45737 is the strongest overall candidate: it fixes the concrete cache bug and adds a targeted regression test, whereas #45738 only changes the implementation.", + "soft_edge_verdicts": [ + { + "left": "pull_request:45737", + "right": "pull_request:45738", + "accept": true, + "reason": "Same underlying code-path bug in `musicgen_melody` cache setup; the diffs match on the core fix and the test-only addition in #45737 could plausibly live in the same PR as #45738." + } + ] + }, + "evaluator_result": { + "accept": true, + "feedback": "The assessment is grounded in the packet: both PRs make the same `musicgen_melody` cache initialization change, and #45737 only adds a regression test on top of that fix. The soft-edge accept is conservative enough because the diffs are mergeable into one PR for the same bugfix." + }, + "evaluator_used": true, + "retried": false, + "accepted_nontrivial_soft_edge": true, + "error_kind": null, + "error_message": null + }, + { + "label": "pull_request", + "component_index": 27, + "component_count": 27, + "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-46069-3", + "nodes": [ + "pull_request:46069", + "pull_request:46073", + "pull_request:46154" + ], + "soft_pairs": [ + "pull_request:46069|pull_request:46073", + "pull_request:46069|pull_request:46154", + "pull_request:46073|pull_request:46154" + ], + "prepared_review_unit_hash": "a49e24056b0490895a8cdd3da402498368a40575ef0c42bbe31b2e91ce74d5d6", + "component_budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 3, + "serialized_chars": 7264, + "estimated_input_tokens": 1816, + "estimated_eval_tokens": 3888 + }, + "budget": { + "node_count": 3, + "item_count": 3, + "soft_pair_count": 3, + "serialized_chars": 7264, + "estimated_input_tokens": 1816, + "estimated_eval_tokens": 3888 + }, + "overflow_policy": "truncate_then_skip", + "trimmed": false, + "aggressively_trimmed": false, + "split": false, + "analyst_result": { + "summary": "These PRs are all model-specific additions for `attn_implementation`/SDPA dispatch, but they target different architectures and code paths. They are related in theme, not duplicates.", + "confidence": 0.95, + "canonical_issue_reason": null, + "canonical_pr_reason": "PR 46154 is the strongest representative: it has the clearest feature scope (`attn_implementation=sdpa` dispatch), touches model code plus multiple test files, and looks like the most complete implementation among the three.", + "best_issue_reason": null, + "best_pr_reason": "PR 46154 is the best single PR to represent this cluster because it is the most fully fleshed-out SDPA dispatch support change and has broader validation coverage.", + "soft_edge_verdicts": [ + { + "left": "pull_request:46069", + "right": "pull_request:46073", + "accept": false, + "reason": "Both add SDPA/attention dispatch support, but for different models (ShieldGemma2 vs CTRL) with different files and code paths, so they are not the same underlying change." + }, + { + "left": "pull_request:46069", + "right": "pull_request:46154", + "accept": false, + "reason": "Same broad feature area, but these are separate model implementations (ShieldGemma2 vs T5) and not a single mergeable code-path fix." + }, + { + "left": "pull_request:46073", + "right": "pull_request:46154", + "accept": false, + "reason": "Both are model-specific SDPA dispatch PRs, but they target different architectures and cannot plausibly be merged into one duplicate fix." + } + ] + }, + "evaluator_result": null, + "evaluator_used": false, + "retried": false, + "accepted_nontrivial_soft_edge": false, + "error_kind": null, + "error_message": null + } + ] +} diff --git a/snapshots/20260524T000134Z/analysis-runs/hybrid-model-20260524t000134z/manifest.json b/snapshots/20260524T000134Z/analysis-runs/hybrid-model-20260524t000134z/manifest.json new file mode 100644 index 0000000000000000000000000000000000000000..804fc9c6246d6631c9c02f409ad3d6848b3e0991 --- /dev/null +++ b/snapshots/20260524T000134Z/analysis-runs/hybrid-model-20260524t000134z/manifest.json @@ -0,0 +1,14 @@ +{ + "analysis_id": "hybrid-model-20260524t000134z", + "artifacts": { + "hybrid": "snapshots/20260524T000134Z/analysis-runs/hybrid-model-20260524t000134z/analysis-report-hybrid.json", + "hybrid_reviews": "snapshots/20260524T000134Z/analysis-runs/hybrid-model-20260524t000134z/analysis-report-hybrid.llm-reviews.json" + }, + "channel": "canonical", + "model": null, + "published_at": "2026-05-24T00:08:56Z", + "repo": "huggingface/transformers", + "schema_version": 1, + "snapshot_id": "20260524T000134Z", + "variant": "hybrid" +} diff --git a/snapshots/20260524T000134Z/manifest.json b/snapshots/20260524T000134Z/manifest.json index 32ff88e731262fb395ebbda97685f6f94716aa12..7df860afa909ec66172edfd944d7375619068ca5 100644 --- a/snapshots/20260524T000134Z/manifest.json +++ b/snapshots/20260524T000134Z/manifest.json @@ -32,6 +32,24 @@ "timeline_events": 101 }, "extracted_at": "2026-05-24T00:01:34Z", + "published_analysis": { + "canonical_analysis_id": "hybrid-model-20260524t000134z", + "runs": { + "hybrid-model-20260524t000134z": { + "analysis_id": "hybrid-model-20260524t000134z", + "artifacts": { + "hybrid": "snapshots/20260524T000134Z/analysis-runs/hybrid-model-20260524t000134z/analysis-report-hybrid.json", + "hybrid_reviews": "snapshots/20260524T000134Z/analysis-runs/hybrid-model-20260524t000134z/analysis-report-hybrid.llm-reviews.json" + }, + "channel": "canonical", + "manifest_path": "snapshots/20260524T000134Z/analysis-runs/hybrid-model-20260524t000134z/manifest.json", + "model": null, + "published_at": "2026-05-24T00:08:56Z", + "variant": "hybrid" + } + }, + "schema_version": 1 + }, "repo": "huggingface/transformers", "snapshot_id": "20260524T000134Z", "watermark": {